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.
20301 Commits
32 Branches
522MB
Tree: a2f4324bb6
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'a2f4324bb6'
${ noResults }
FFmpeg/libswscale/swscale.c

3608 lines
123KB
Raw Blame History 

  1. /*

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

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

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

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

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

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

  10.  *

  11.  * FFmpeg is distributed in the hope that it will be useful,

  12.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  13.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

  14.  * GNU General Public License for more details.

  15.  *

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

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

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

  19.  *

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

  21.  * under the LGPL license too

  22.  */

  23. 

  24. /*

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

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

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

  28. 

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

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

  31.   x -> x

  32.   YUV9 -> YV12

  33.   YUV9/YV12 -> Y800

  34.   Y800 -> YUV9/YV12

  35.   BGR24 -> BGR32 & RGB24 -> RGB32

  36.   BGR32 -> BGR24 & RGB32 -> RGB24

  37.   BGR15 -> BGR16

  38. */

  39. 

  40. /*

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

  42.  YV12 -> BGR16

  43.  YV12 -> YV12

  44.  BGR15 -> BGR16

  45.  BGR16 -> BGR16

  46.  YVU9 -> YV12

  47. 

  48. untested special converters

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

  50.   YV12/I420 -> YV12/I420

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

  52.   BGR24 -> BGR32 & RGB24 -> RGB32

  53.   BGR32 -> BGR24 & RGB32 -> RGB24

  54.   BGR24 -> YV12

  55. */

  56. 

  57. #define _SVID_SOURCE //needed for MAP_ANONYMOUS

  58. #include <inttypes.h>

  59. #include <string.h>

  60. #include <math.h>

  61. #include <stdio.h>

  62. #include "config.h"

  63. #include <assert.h>

  64. #if HAVE_SYS_MMAN_H

  65. #include <sys/mman.h>

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

  67. #define MAP_ANONYMOUS MAP_ANON

  68. #endif

  69. #endif

  70. #if HAVE_VIRTUALALLOC

  71. #define WIN32_LEAN_AND_MEAN

  72. #include <windows.h>

  73. #endif

  74. #include "swscale.h"

  75. #include "swscale_internal.h"

  76. #include "rgb2rgb.h"

  77. #include "libavutil/x86_cpu.h"

  78. #include "libavutil/bswap.h"

  79. 

  80. unsigned swscale_version(void)

  81. {

  82.     return LIBSWSCALE_VERSION_INT;

  83. }

  84. 

  85. #undef MOVNTQ

  86. #undef PAVGB

  87. 

  88. //#undef HAVE_MMX2

  89. //#define HAVE_AMD3DNOW

  90. //#undef HAVE_MMX

  91. //#undef ARCH_X86

  92. //#define WORDS_BIGENDIAN

  93. #define DITHER1XBPP

  94. 

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

  96. 

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

  98. 

  99. #ifdef M_PI

  100. #define PI M_PI

  101. #else

  102. #define PI 3.14159265358979323846

  103. #endif

  104. 

  105. #define isSupportedIn(x)    (       \

  106.            (x)==PIX_FMT_YUV420P     \

  107.         || (x)==PIX_FMT_YUVA420P    \

  108.         || (x)==PIX_FMT_YUYV422     \

  109.         || (x)==PIX_FMT_UYVY422     \

  110.         || (x)==PIX_FMT_RGB48BE     \

  111.         || (x)==PIX_FMT_RGB48LE     \

  112.         || (x)==PIX_FMT_RGB32       \

  113.         || (x)==PIX_FMT_RGB32_1     \

  114.         || (x)==PIX_FMT_BGR24       \

  115.         || (x)==PIX_FMT_BGR565      \

  116.         || (x)==PIX_FMT_BGR555      \

  117.         || (x)==PIX_FMT_BGR32       \

  118.         || (x)==PIX_FMT_BGR32_1     \

  119.         || (x)==PIX_FMT_RGB24       \

  120.         || (x)==PIX_FMT_RGB565      \

  121.         || (x)==PIX_FMT_RGB555      \

  122.         || (x)==PIX_FMT_GRAY8       \

  123.         || (x)==PIX_FMT_YUV410P     \

  124.         || (x)==PIX_FMT_YUV440P     \

  125.         || (x)==PIX_FMT_GRAY16BE    \

  126.         || (x)==PIX_FMT_GRAY16LE    \

  127.         || (x)==PIX_FMT_YUV444P     \

  128.         || (x)==PIX_FMT_YUV422P     \

  129.         || (x)==PIX_FMT_YUV411P     \

  130.         || (x)==PIX_FMT_PAL8        \

  131.         || (x)==PIX_FMT_BGR8        \

  132.         || (x)==PIX_FMT_RGB8        \

  133.         || (x)==PIX_FMT_BGR4_BYTE   \

  134.         || (x)==PIX_FMT_RGB4_BYTE   \

  135.         || (x)==PIX_FMT_YUV440P     \

  136.         || (x)==PIX_FMT_MONOWHITE   \

  137.         || (x)==PIX_FMT_MONOBLACK   \

  138.         || (x)==PIX_FMT_YUV420PLE   \

  139.         || (x)==PIX_FMT_YUV422PLE   \

  140.         || (x)==PIX_FMT_YUV444PLE   \

  141.         || (x)==PIX_FMT_YUV420PBE   \

  142.         || (x)==PIX_FMT_YUV422PBE   \

  143.         || (x)==PIX_FMT_YUV444PBE   \

  144.     )

  145. #define isSupportedOut(x)   (       \

  146.            (x)==PIX_FMT_YUV420P     \

  147.         || (x)==PIX_FMT_YUVA420P    \

  148.         || (x)==PIX_FMT_YUYV422     \

  149.         || (x)==PIX_FMT_UYVY422     \

  150.         || (x)==PIX_FMT_YUV444P     \

  151.         || (x)==PIX_FMT_YUV422P     \

  152.         || (x)==PIX_FMT_YUV411P     \

  153.         || isRGB(x)                 \

  154.         || isBGR(x)                 \

  155.         || (x)==PIX_FMT_NV12        \

  156.         || (x)==PIX_FMT_NV21        \

  157.         || (x)==PIX_FMT_GRAY16BE    \

  158.         || (x)==PIX_FMT_GRAY16LE    \

  159.         || (x)==PIX_FMT_GRAY8       \

  160.         || (x)==PIX_FMT_YUV410P     \

  161.         || (x)==PIX_FMT_YUV440P     \

  162.         || (x)==PIX_FMT_YUV420PLE   \

  163.         || (x)==PIX_FMT_YUV422PLE   \

  164.         || (x)==PIX_FMT_YUV444PLE   \

  165.         || (x)==PIX_FMT_YUV420PBE   \

  166.         || (x)==PIX_FMT_YUV422PBE   \

  167.         || (x)==PIX_FMT_YUV444PBE   \

  168.     )

  169. #define isPacked(x)         (       \

  170.            (x)==PIX_FMT_PAL8        \

  171.         || (x)==PIX_FMT_YUYV422     \

  172.         || (x)==PIX_FMT_UYVY422     \

  173.         || isRGB(x)                 \

  174.         || isBGR(x)                 \

  175.     )

  176. #define usePal(x)           (       \

  177.            (x)==PIX_FMT_PAL8        \

  178.         || (x)==PIX_FMT_BGR4_BYTE   \

  179.         || (x)==PIX_FMT_RGB4_BYTE   \

  180.         || (x)==PIX_FMT_BGR8        \

  181.         || (x)==PIX_FMT_RGB8        \

  182.     )

  183. 

  184. #define RGB2YUV_SHIFT 15

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

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

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

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

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

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

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

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

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

  194. 

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

  196. 

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

  198.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},

  199.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},

  200.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  201.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  202.     {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC

  203.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  204.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M

  205.     {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M

  206. };

  207. 

  208. /*

  209. NOTES

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

  211. 

  212. TODO

  213. more intelligent misalignment avoidance for the horizontal scaler

  214. write special vertical cubic upscale version

  215. optimize C code (YV12 / minmax)

  216. add support for packed pixel YUV input & output

  217. add support for Y8 output

  218. optimize BGR24 & BGR32

  219. add BGR4 output support

  220. write special BGR->BGR scaler

  221. */

  222. 

  223. #if ARCH_X86 && CONFIG_GPL

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

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

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

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

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

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

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

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

  232. 

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

  234.         0x0103010301030103LL,

  235.         0x0200020002000200LL,};

  236. 

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

  238.         0x0602060206020602LL,

  239.         0x0004000400040004LL,};

  240. 

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

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

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

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

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

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

  247. 

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

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

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

  251. 

  252. #ifdef FAST_BGR2YV12

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

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

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

  256. #else

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

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

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

  260. #endif /* FAST_BGR2YV12 */

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

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

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

  264. 

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

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

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

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

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

  270. 

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

  272.     {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},

  273.     {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},

  274. };

  275. 

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

  277. 

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

  279. 

  280. // clipping helper table for C implementations:

  281. static unsigned char clip_table[768];

  282. 

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

  284. 

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

  286. {  1,   3,   1,   3,   1,   3,   1,   3, },

  287. {  2,   0,   2,   0,   2,   0,   2,   0, },

  288. };

  289. 

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

  291. {  6,   2,   6,   2,   6,   2,   6,   2, },

  292. {  0,   4,   0,   4,   0,   4,   0,   4, },

  293. };

  294. 

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

  296. { 17,   9,  23,  15,  16,   8,  22,  14, },

  297. {  5,  29,   3,  27,   4,  28,   2,  26, },

  298. { 21,  13,  19,  11,  20,  12,  18,  10, },

  299. {  0,  24,   6,  30,   1,  25,   7,  31, },

  300. { 16,   8,  22,  14,  17,   9,  23,  15, },

  301. {  4,  28,   2,  26,   5,  29,   3,  27, },

  302. { 20,  12,  18,  10,  21,  13,  19,  11, },

  303. {  1,  25,   7,  31,   0,  24,   6,  30, },

  304. };

  305. 

  306. #if 0

  307. DECLARE_ALIGNED(8, const uint8_t, dither_8x8_64[8][8])={

  308. {  0,  48,  12,  60,   3,  51,  15,  63, },

  309. { 32,  16,  44,  28,  35,  19,  47,  31, },

  310. {  8,  56,   4,  52,  11,  59,   7,  55, },

  311. { 40,  24,  36,  20,  43,  27,  39,  23, },

  312. {  2,  50,  14,  62,   1,  49,  13,  61, },

  313. { 34,  18,  46,  30,  33,  17,  45,  29, },

  314. { 10,  58,   6,  54,   9,  57,   5,  53, },

  315. { 42,  26,  38,  22,  41,  25,  37,  21, },

  316. };

  317. #endif

  318. 

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

  320. {  0,  55,  14,  68,   3,  58,  17,  72, },

  321. { 37,  18,  50,  32,  40,  22,  54,  35, },

  322. {  9,  64,   5,  59,  13,  67,   8,  63, },

  323. { 46,  27,  41,  23,  49,  31,  44,  26, },

  324. {  2,  57,  16,  71,   1,  56,  15,  70, },

  325. { 39,  21,  52,  34,  38,  19,  51,  33, },

  326. { 11,  66,   7,  62,  10,  65,   6,  60, },

  327. { 48,  30,  43,  25,  47,  29,  42,  24, },

  328. };

  329. 

  330. #if 0

  331. DECLARE_ALIGNED(8, const uint8_t, dither_8x8_128[8][8])={

  332. { 68,  36,  92,  60,  66,  34,  90,  58, },

  333. { 20, 116,  12, 108,  18, 114,  10, 106, },

  334. { 84,  52,  76,  44,  82,  50,  74,  42, },

  335. {  0,  96,  24, 120,   6, 102,  30, 126, },

  336. { 64,  32,  88,  56,  70,  38,  94,  62, },

  337. { 16, 112,   8, 104,  22, 118,  14, 110, },

  338. { 80,  48,  72,  40,  86,  54,  78,  46, },

  339. {  4, 100,  28, 124,   2,  98,  26, 122, },

  340. };

  341. #endif

  342. 

  343. #if 1

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

  345. {117,  62, 158, 103, 113,  58, 155, 100, },

  346. { 34, 199,  21, 186,  31, 196,  17, 182, },

  347. {144,  89, 131,  76, 141,  86, 127,  72, },

  348. {  0, 165,  41, 206,  10, 175,  52, 217, },

  349. {110,  55, 151,  96, 120,  65, 162, 107, },

  350. { 28, 193,  14, 179,  38, 203,  24, 189, },

  351. {138,  83, 124,  69, 148,  93, 134,  79, },

  352. {  7, 172,  48, 213,   3, 168,  45, 210, },

  353. };

  354. #elif 1

  355. // tries to correct a gamma of 1.5

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

  357. {  0, 143,  18, 200,   2, 156,  25, 215, },

  358. { 78,  28, 125,  64,  89,  36, 138,  74, },

  359. { 10, 180,   3, 161,  16, 195,   8, 175, },

  360. {109,  51,  93,  38, 121,  60, 105,  47, },

  361. {  1, 152,  23, 210,   0, 147,  20, 205, },

  362. { 85,  33, 134,  71,  81,  30, 130,  67, },

  363. { 14, 190,   6, 171,  12, 185,   5, 166, },

  364. {117,  57, 101,  44, 113,  54,  97,  41, },

  365. };

  366. #elif 1

  367. // tries to correct a gamma of 2.0

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

  369. {  0, 124,   8, 193,   0, 140,  12, 213, },

  370. { 55,  14, 104,  42,  66,  19, 119,  52, },

  371. {  3, 168,   1, 145,   6, 187,   3, 162, },

  372. { 86,  31,  70,  21,  99,  39,  82,  28, },

  373. {  0, 134,  11, 206,   0, 129,   9, 200, },

  374. { 62,  17, 114,  48,  58,  16, 109,  45, },

  375. {  5, 181,   2, 157,   4, 175,   1, 151, },

  376. { 95,  36,  78,  26,  90,  34,  74,  24, },

  377. };

  378. #else

  379. // tries to correct a gamma of 2.5

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

  381. {  0, 107,   3, 187,   0, 125,   6, 212, },

  382. { 39,   7,  86,  28,  49,  11, 102,  36, },

  383. {  1, 158,   0, 131,   3, 180,   1, 151, },

  384. { 68,  19,  52,  12,  81,  25,  64,  17, },

  385. {  0, 119,   5, 203,   0, 113,   4, 195, },

  386. { 45,   9,  96,  33,  42,   8,  91,  30, },

  387. {  2, 172,   1, 144,   2, 165,   0, 137, },

  388. { 77,  23,  60,  15,  72,  21,  56,  14, },

  389. };

  390. #endif

  391. 

  392. const char *sws_format_name(enum PixelFormat format)

  393. {

  394.     switch (format) {

  395.         case PIX_FMT_YUV420P:

  396.             return "yuv420p";

  397.         case PIX_FMT_YUVA420P:

  398.             return "yuva420p";

  399.         case PIX_FMT_YUYV422:

  400.             return "yuyv422";

  401.         case PIX_FMT_RGB24:

  402.             return "rgb24";

  403.         case PIX_FMT_BGR24:

  404.             return "bgr24";

  405.         case PIX_FMT_YUV422P:

  406.             return "yuv422p";

  407.         case PIX_FMT_YUV444P:

  408.             return "yuv444p";

  409.         case PIX_FMT_RGB32:

  410.             return "rgb32";

  411.         case PIX_FMT_YUV410P:

  412.             return "yuv410p";

  413.         case PIX_FMT_YUV411P:

  414.             return "yuv411p";

  415.         case PIX_FMT_RGB565:

  416.             return "rgb565";

  417.         case PIX_FMT_RGB555:

  418.             return "rgb555";

  419.         case PIX_FMT_GRAY16BE:

  420.             return "gray16be";

  421.         case PIX_FMT_GRAY16LE:

  422.             return "gray16le";

  423.         case PIX_FMT_GRAY8:

  424.             return "gray8";

  425.         case PIX_FMT_MONOWHITE:

  426.             return "mono white";

  427.         case PIX_FMT_MONOBLACK:

  428.             return "mono black";

  429.         case PIX_FMT_PAL8:

  430.             return "Palette";

  431.         case PIX_FMT_YUVJ420P:

  432.             return "yuvj420p";

  433.         case PIX_FMT_YUVJ422P:

  434.             return "yuvj422p";

  435.         case PIX_FMT_YUVJ444P:

  436.             return "yuvj444p";

  437.         case PIX_FMT_XVMC_MPEG2_MC:

  438.             return "xvmc_mpeg2_mc";

  439.         case PIX_FMT_XVMC_MPEG2_IDCT:

  440.             return "xvmc_mpeg2_idct";

  441.         case PIX_FMT_UYVY422:

  442.             return "uyvy422";

  443.         case PIX_FMT_UYYVYY411:

  444.             return "uyyvyy411";

  445.         case PIX_FMT_RGB32_1:

  446.             return "rgb32x";

  447.         case PIX_FMT_BGR32_1:

  448.             return "bgr32x";

  449.         case PIX_FMT_BGR32:

  450.             return "bgr32";

  451.         case PIX_FMT_BGR565:

  452.             return "bgr565";

  453.         case PIX_FMT_BGR555:

  454.             return "bgr555";

  455.         case PIX_FMT_BGR8:

  456.             return "bgr8";

  457.         case PIX_FMT_BGR4:

  458.             return "bgr4";

  459.         case PIX_FMT_BGR4_BYTE:

  460.             return "bgr4 byte";

  461.         case PIX_FMT_RGB8:

  462.             return "rgb8";

  463.         case PIX_FMT_RGB4:

  464.             return "rgb4";

  465.         case PIX_FMT_RGB4_BYTE:

  466.             return "rgb4 byte";

  467.         case PIX_FMT_RGB48BE:

  468.             return "rgb48be";

  469.         case PIX_FMT_RGB48LE:

  470.             return "rgb48le";

  471.         case PIX_FMT_NV12:

  472.             return "nv12";

  473.         case PIX_FMT_NV21:

  474.             return "nv21";

  475.         case PIX_FMT_YUV440P:

  476.             return "yuv440p";

  477.         case PIX_FMT_VDPAU_H264:

  478.             return "vdpau_h264";

  479.         case PIX_FMT_VDPAU_MPEG1:

  480.             return "vdpau_mpeg1";

  481.         case PIX_FMT_VDPAU_MPEG2:

  482.             return "vdpau_mpeg2";

  483.         case PIX_FMT_VDPAU_WMV3:

  484.             return "vdpau_wmv3";

  485.         case PIX_FMT_VDPAU_VC1:

  486.             return "vdpau_vc1";

  487.         case PIX_FMT_YUV420PLE:

  488.             return "yuv420ple";

  489.         case PIX_FMT_YUV422PLE:

  490.             return "yuv422ple";

  491.         case PIX_FMT_YUV444PLE:

  492.             return "yuv444ple";

  493.         case PIX_FMT_YUV420PBE:

  494.             return "yuv420pbe";

  495.         case PIX_FMT_YUV422PBE:

  496.             return "yuv422pbe";

  497.         case PIX_FMT_YUV444PBE:

  498.             return "yuv444pbe";

  499.         default:

  500.             return "Unknown format";

  501.     }

  502. }

  503. 

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

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

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

  507. {

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

  509.     int i;

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

  511.     {

  512.         int val=1<<18;

  513.         int j;

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

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

  516. 

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

  518.     }

  519. 

  520.     if (uDest)

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

  522.         {

  523.             int u=1<<18;

  524.             int v=1<<18;

  525.             int j;

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

  527.             {

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

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

  530.             }

  531. 

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

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

  534.         }

  535. 

  536.     if (CONFIG_SWSCALE_ALPHA && aDest)

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

  538.             int val=1<<18;

  539.             int j;

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

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

  542. 

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

  544.         }

  545. 

  546. }

  547. 

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

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

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

  551. {

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

  553.     int i;

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

  555.     {

  556.         int val=1<<18;

  557.         int j;

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

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

  560. 

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

  562.     }

  563. 

  564.     if (!uDest)

  565.         return;

  566. 

  567.     if (dstFormat == PIX_FMT_NV12)

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

  569.         {

  570.             int u=1<<18;

  571.             int v=1<<18;

  572.             int j;

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

  574.             {

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

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

  577.             }

  578. 

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

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

  581.         }

  582.     else

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

  584.         {

  585.             int u=1<<18;

  586.             int v=1<<18;

  587.             int j;

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

  589.             {

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

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

  592.             }

  593. 

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

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

  596.         }

  597. }

  598. 

  599. #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \

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

  601.         int j;\

  602.         int Y1 = 1<<18;\

  603.         int Y2 = 1<<18;\

  604.         int U  = 1<<18;\

  605.         int V  = 1<<18;\

  606.         int av_unused A1, A2;\

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

  608.         const int i2= 2*i;\

  609.         \

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

  611.         {\

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

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

  614.         }\

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

  616.         {\

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

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

  619.         }\

  620.         Y1>>=19;\

  621.         Y2>>=19;\

  622.         U >>=19;\

  623.         V >>=19;\

  624.         if (alpha){\

  625.             A1 = 1<<18;\

  626.             A2 = 1<<18;\

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

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

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

  630.             }\

  631.             A1>>=19;\

  632.             A2>>=19;\

  633.         }\

  634. 

  635. #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \

  636.         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\

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

  638.         {\

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

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

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

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

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

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

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

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

  647.         }\

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

  649.             A1=av_clip_uint8(A1);\

  650.             A2=av_clip_uint8(A2);\

  651.         }

  652. 

  653. #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \

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

  655.         int j;\

  656.         int Y = 0;\

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

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

  659.         int av_unused A;\

  660.         int R,G,B;\

  661.         \

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

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

  664.         }\

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

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

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

  668.         }\

  669.         Y >>=10;\

  670.         U >>=10;\

  671.         V >>=10;\

  672.         if (alpha){\

  673.             A = rnd;\

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

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

  676.             A >>=19;\

  677.             if (A&256)\

  678.                 A = av_clip_uint8(A);\

  679.         }\

  680. 

  681. #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \

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

  683.         Y-= c->yuv2rgb_y_offset;\

  684.         Y*= c->yuv2rgb_y_coeff;\

  685.         Y+= rnd;\

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

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

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

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

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

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

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

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

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

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

  696.         }\

  697. 

  698. 

  699. #define YSCALE_YUV_2_GRAY16_C \

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

  701.         int j;\

  702.         int Y1 = 1<<18;\

  703.         int Y2 = 1<<18;\

  704.         int U  = 1<<18;\

  705.         int V  = 1<<18;\

  706.         \

  707.         const int i2= 2*i;\

  708.         \

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

  710.         {\

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

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

  713.         }\

  714.         Y1>>=11;\

  715.         Y2>>=11;\

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

  717.         {\

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

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

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

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

  722.         }

  723. 

  724. #define YSCALE_YUV_2_RGBX_C(type,alpha) \

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

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

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

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

  729. 

  730. #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \

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

  732.         const int i2= 2*i;       \

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

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

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

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

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

  738.         int av_unused A1, A2;                                         \

  739.         if (alpha){\

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

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

  742.         }\

  743. 

  744. #define YSCALE_YUV_2_GRAY16_2_C   \

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

  746.         const int i2= 2*i;       \

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

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

  749. 

  750. #define YSCALE_YUV_2_RGB2_C(type,alpha) \

  751.     YSCALE_YUV_2_PACKED2_C(type,alpha)\

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

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

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

  755. 

  756. #define YSCALE_YUV_2_PACKED1_C(type,alpha) \

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

  758.         const int i2= 2*i;\

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

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

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

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

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

  764.         int av_unused A1, A2;\

  765.         if (alpha){\

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

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

  768.         }\

  769. 

  770. #define YSCALE_YUV_2_GRAY16_1_C \

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

  772.         const int i2= 2*i;\

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

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

  775. 

  776. #define YSCALE_YUV_2_RGB1_C(type,alpha) \

  777.     YSCALE_YUV_2_PACKED1_C(type,alpha)\

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

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

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

  781. 

  782. #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \

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

  784.         const int i2= 2*i;\

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

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

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

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

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

  790.         int av_unused A1, A2;\

  791.         if (alpha){\

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

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

  794.         }\

  795. 

  796. #define YSCALE_YUV_2_RGB1B_C(type,alpha) \

  797.     YSCALE_YUV_2_PACKED1B_C(type,alpha)\

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

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

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

  801. 

  802. #define YSCALE_YUV_2_MONO2_C \

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

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

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

  806.         int acc;\

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

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

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

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

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

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

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

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

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

  816.         dest++;\

  817.     }\

  818. 

  819. 

  820. #define YSCALE_YUV_2_MONOX_C \

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

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

  823.     int acc=0;\

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

  825.         int j;\

  826.         int Y1=1<<18;\

  827.         int Y2=1<<18;\

  828. \

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

  830.         {\

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

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

  833.         }\

  834.         Y1>>=19;\

  835.         Y2>>=19;\

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

  837.         {\

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

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

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

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

  842.         }\

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

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

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

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

  847.             dest++;\

  848.         }\

  849.     }

  850. 

  851. 

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

  853.     switch(c->dstFormat)\

  854.     {\

  855.     case PIX_FMT_RGB48BE:\

  856.     case PIX_FMT_RGB48LE:\

  857.         func(uint8_t,0)\

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

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

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

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

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

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

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

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

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

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

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

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

  870.             dest+=12;\

  871.         }\

  872.         break;\

  873.     case PIX_FMT_RGBA:\

  874.     case PIX_FMT_BGRA:\

  875.         if (CONFIG_SMALL){\

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

  877.             func(uint32_t,needAlpha)\

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

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

  880.             }\

  881.         }else{\

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

  883.                 func(uint32_t,1)\

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

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

  886.                 }\

  887.             }else{\

  888.                 func(uint32_t,0)\

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

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

  891.                 }\

  892.             }\

  893.         }\

  894.         break;\

  895.     case PIX_FMT_ARGB:\

  896.     case PIX_FMT_ABGR:\

  897.         if (CONFIG_SMALL){\

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

  899.             func(uint32_t,needAlpha)\

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

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

  902.             }\

  903.         }else{\

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

  905.                 func(uint32_t,1)\

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

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

  908.                 }\

  909.             }else{\

  910.                 func(uint32_t,0)\

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

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

  913.                 }\

  914.             }\

  915.         }                \

  916.         break;\

  917.     case PIX_FMT_RGB24:\

  918.         func(uint8_t,0)\

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

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

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

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

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

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

  925.             dest+=6;\

  926.         }\

  927.         break;\

  928.     case PIX_FMT_BGR24:\

  929.         func(uint8_t,0)\

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

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

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

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

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

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

  936.             dest+=6;\

  937.         }\

  938.         break;\

  939.     case PIX_FMT_RGB565:\

  940.     case PIX_FMT_BGR565:\

  941.         {\

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

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

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

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

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

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

  948.             func(uint16_t,0)\

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

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

  951.             }\

  952.         }\

  953.         break;\

  954.     case PIX_FMT_RGB555:\

  955.     case PIX_FMT_BGR555:\

  956.         {\

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

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

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

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

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

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

  963.             func(uint16_t,0)\

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

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

  966.             }\

  967.         }\

  968.         break;\

  969.     case PIX_FMT_RGB8:\

  970.     case PIX_FMT_BGR8:\

  971.         {\

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

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

  974.             func(uint8_t,0)\

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

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

  977.             }\

  978.         }\

  979.         break;\

  980.     case PIX_FMT_RGB4:\

  981.     case PIX_FMT_BGR4:\

  982.         {\

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

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

  985.             func(uint8_t,0)\

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

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

  988.             }\

  989.         }\

  990.         break;\

  991.     case PIX_FMT_RGB4_BYTE:\

  992.     case PIX_FMT_BGR4_BYTE:\

  993.         {\

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

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

  996.             func(uint8_t,0)\

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

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

  999.             }\

  1000.         }\

  1001.         break;\

  1002.     case PIX_FMT_MONOBLACK:\

  1003.     case PIX_FMT_MONOWHITE:\

  1004.         {\

  1005.             func_monoblack\

  1006.         }\

  1007.         break;\

  1008.     case PIX_FMT_YUYV422:\

  1009.         func2\

  1010.             ((uint8_t*)dest)[2*i2+0]= Y1;\

  1011.             ((uint8_t*)dest)[2*i2+1]= U;\

  1012.             ((uint8_t*)dest)[2*i2+2]= Y2;\

  1013.             ((uint8_t*)dest)[2*i2+3]= V;\

  1014.         }                \

  1015.         break;\

  1016.     case PIX_FMT_UYVY422:\

  1017.         func2\

  1018.             ((uint8_t*)dest)[2*i2+0]= U;\

  1019.             ((uint8_t*)dest)[2*i2+1]= Y1;\

  1020.             ((uint8_t*)dest)[2*i2+2]= V;\

  1021.             ((uint8_t*)dest)[2*i2+3]= Y2;\

  1022.         }                \

  1023.         break;\

  1024.     case PIX_FMT_GRAY16BE:\

  1025.         func_g16\

  1026.             ((uint8_t*)dest)[2*i2+0]= Y1>>8;\

  1027.             ((uint8_t*)dest)[2*i2+1]= Y1;\

  1028.             ((uint8_t*)dest)[2*i2+2]= Y2>>8;\

  1029.             ((uint8_t*)dest)[2*i2+3]= Y2;\

  1030.         }                \

  1031.         break;\

  1032.     case PIX_FMT_GRAY16LE:\

  1033.         func_g16\

  1034.             ((uint8_t*)dest)[2*i2+0]= Y1;\

  1035.             ((uint8_t*)dest)[2*i2+1]= Y1>>8;\

  1036.             ((uint8_t*)dest)[2*i2+2]= Y2;\

  1037.             ((uint8_t*)dest)[2*i2+3]= Y2>>8;\

  1038.         }                \

  1039.         break;\

  1040.     }\

  1041. 

  1042. 

  1043. static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  1044.                                   const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  1045.                                   const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  1046. {

  1047.     int i;

  1048.     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)

  1049. }

  1050. 

  1051. static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  1052.                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  1053.                                     const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  1054. {

  1055.     int i;

  1056.     int step= fmt_depth(c->dstFormat)/8;

  1057.     int aidx= 3;

  1058. 

  1059.     switch(c->dstFormat){

  1060.     case PIX_FMT_ARGB:

  1061.         dest++;

  1062.         aidx= 0;

  1063.     case PIX_FMT_RGB24:

  1064.         aidx--;

  1065.     case PIX_FMT_RGBA:

  1066.         if (CONFIG_SMALL){

  1067.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  1068.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  1069.                 dest[aidx]= needAlpha ? A : 255;

  1070.                 dest[0]= R>>22;

  1071.                 dest[1]= G>>22;

  1072.                 dest[2]= B>>22;

  1073.                 dest+= step;

  1074.             }

  1075.         }else{

  1076.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){

  1077.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  1078.                     dest[aidx]= A;

  1079.                     dest[0]= R>>22;

  1080.                     dest[1]= G>>22;

  1081.                     dest[2]= B>>22;

  1082.                     dest+= step;

  1083.                 }

  1084.             }else{

  1085.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  1086.                     dest[aidx]= 255;

  1087.                     dest[0]= R>>22;

  1088.                     dest[1]= G>>22;

  1089.                     dest[2]= B>>22;

  1090.                     dest+= step;

  1091.                 }

  1092.             }

  1093.         }

  1094.         break;

  1095.     case PIX_FMT_ABGR:

  1096.         dest++;

  1097.         aidx= 0;

  1098.     case PIX_FMT_BGR24:

  1099.         aidx--;

  1100.     case PIX_FMT_BGRA:

  1101.         if (CONFIG_SMALL){

  1102.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  1103.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  1104.                 dest[aidx]= needAlpha ? A : 255;

  1105.                 dest[0]= B>>22;

  1106.                 dest[1]= G>>22;

  1107.                 dest[2]= R>>22;

  1108.                 dest+= step;

  1109.             }

  1110.         }else{

  1111.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){

  1112.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  1113.                     dest[aidx]= A;

  1114.                     dest[0]= B>>22;

  1115.                     dest[1]= G>>22;

  1116.                     dest[2]= R>>22;

  1117.                     dest+= step;

  1118.                 }

  1119.             }else{

  1120.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  1121.                     dest[aidx]= 255;

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

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

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

  1125.                     dest+= step;

  1126.                 }

  1127.             }

  1128.         }

  1129.         break;

  1130.     default:

  1131.         assert(0);

  1132.     }

  1133. }

  1134. 

  1135. static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val){

  1136.     int i;

  1137.     uint8_t *ptr = plane + stride*y;

  1138.     for (i=0; i<height; i++){

  1139.         memset(ptr, val, width);

  1140.         ptr += stride;

  1141.     }

  1142. }

  1143. 

  1144. static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width)

  1145. {

  1146.     int i;

  1147.     for (i = 0; i < width; i++) {

  1148.         int r = src[i*6+0];

  1149.         int g = src[i*6+2];

  1150.         int b = src[i*6+4];

  1151. 

  1152.         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1153.     }

  1154. }

  1155. 

  1156. static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,

  1157.                              uint8_t *src1, uint8_t *src2, int width)

  1158. {

  1159.     int i;

  1160.     assert(src1==src2);

  1161.     for (i = 0; i < width; i++) {

  1162.         int r = src1[6*i + 0];

  1163.         int g = src1[6*i + 2];

  1164.         int b = src1[6*i + 4];

  1165. 

  1166.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1167.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1168.     }

  1169. }

  1170. 

  1171. static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1172.                                   uint8_t *src1, uint8_t *src2, int width)

  1173. {

  1174.     int i;

  1175.     assert(src1==src2);

  1176.     for (i = 0; i < width; i++) {

  1177.         int r= src1[12*i + 0] + src1[12*i + 6];

  1178.         int g= src1[12*i + 2] + src1[12*i + 8];

  1179.         int b= src1[12*i + 4] + src1[12*i + 10];

  1180. 

  1181.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1182.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1183.     }

  1184. }

  1185. 

  1186. #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\

  1187. static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\

  1188. {\

  1189.     int i;\

  1190.     for (i=0; i<width; i++)\

  1191.     {\

  1192.         int b= (((const type*)src)[i]>>shb)&maskb;\

  1193.         int g= (((const type*)src)[i]>>shg)&maskg;\

  1194.         int r= (((const type*)src)[i]>>shr)&maskr;\

  1195. \

  1196.         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\

  1197.     }\

  1198. }

  1199. 

  1200. BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1201. BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1202. BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)

  1203. BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)

  1204. BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)

  1205. BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)

  1206. 

  1207. static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused){

  1208.     int i;

  1209.     for (i=0; i<width; i++){

  1210.         dst[i]= src[4*i];

  1211.     }

  1212. }

  1213. 

  1214. #define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\

  1215. static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1216. {\

  1217.     int i;\

  1218.     for (i=0; i<width; i++)\

  1219.     {\

  1220.         int b= (((const type*)src)[i]&maskb)>>shb;\

  1221.         int g= (((const type*)src)[i]&maskg)>>shg;\

  1222.         int r= (((const type*)src)[i]&maskr)>>shr;\

  1223. \

  1224.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\

  1225.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\

  1226.     }\

  1227. }\

  1228. static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1229. {\

  1230.     int i;\

  1231.     for (i=0; i<width; i++)\

  1232.     {\

  1233.         int pix0= ((const type*)src)[2*i+0];\

  1234.         int pix1= ((const type*)src)[2*i+1];\

  1235.         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\

  1236.         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\

  1237.         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\

  1238.         g&= maskg|(2*maskg);\

  1239. \

  1240.         g>>=shg;\

  1241. \

  1242.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\

  1243.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\

  1244.     }\

  1245. }

  1246. 

  1247. BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1248. BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1249. BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)

  1250. BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)

  1251. BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)

  1252. BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)

  1253. 

  1254. static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)

  1255. {

  1256.     int i;

  1257.     for (i=0; i<width; i++)

  1258.     {

  1259.         int d= src[i];

  1260. 

  1261.         dst[i]= pal[d] & 0xFF;

  1262.     }

  1263. }

  1264. 

  1265. static inline void palToUV(uint8_t *dstU, uint8_t *dstV,

  1266.                            const uint8_t *src1, const uint8_t *src2,

  1267.                            long width, uint32_t *pal)

  1268. {

  1269.     int i;

  1270.     assert(src1 == src2);

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

  1272.     {

  1273.         int p= pal[src1[i]];

  1274. 

  1275.         dstU[i]= p>>8;

  1276.         dstV[i]= p>>16;

  1277.     }

  1278. }

  1279. 

  1280. static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1281. {

  1282.     int i, j;

  1283.     for (i=0; i<width/8; i++){

  1284.         int d= ~src[i];

  1285.         for(j=0; j<8; j++)

  1286.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1287.     }

  1288. }

  1289. 

  1290. static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1291. {

  1292.     int i, j;

  1293.     for (i=0; i<width/8; i++){

  1294.         int d= src[i];

  1295.         for(j=0; j<8; j++)

  1296.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1297.     }

  1298. }

  1299. 

  1300. 

  1301. //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one

  1302. //Plain C versions

  1303. #if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT

  1304. #define COMPILE_C

  1305. #endif

  1306. 

  1307. #if ARCH_PPC

  1308. #if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT

  1309. #define COMPILE_ALTIVEC

  1310. #endif

  1311. #endif //ARCH_PPC

  1312. 

  1313. #if ARCH_X86

  1314. 

  1315. #if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1316. #define COMPILE_MMX

  1317. #endif

  1318. 

  1319. #if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1320. #define COMPILE_MMX2

  1321. #endif

  1322. 

  1323. #if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1324. #define COMPILE_3DNOW

  1325. #endif

  1326. #endif //ARCH_X86

  1327. 

  1328. #undef HAVE_MMX

  1329. #undef HAVE_MMX2

  1330. #undef HAVE_AMD3DNOW

  1331. #undef HAVE_ALTIVEC

  1332. #define HAVE_MMX 0

  1333. #define HAVE_MMX2 0

  1334. #define HAVE_AMD3DNOW 0

  1335. #define HAVE_ALTIVEC 0

  1336. 

  1337. #ifdef COMPILE_C

  1338. #define RENAME(a) a ## _C

  1339. #include "swscale_template.c"

  1340. #endif

  1341. 

  1342. #ifdef COMPILE_ALTIVEC

  1343. #undef RENAME

  1344. #undef HAVE_ALTIVEC

  1345. #define HAVE_ALTIVEC 1

  1346. #define RENAME(a) a ## _altivec

  1347. #include "swscale_template.c"

  1348. #endif

  1349. 

  1350. #if ARCH_X86

  1351. 

  1352. //MMX versions

  1353. #ifdef COMPILE_MMX

  1354. #undef RENAME

  1355. #undef HAVE_MMX

  1356. #undef HAVE_MMX2

  1357. #undef HAVE_AMD3DNOW

  1358. #define HAVE_MMX 1

  1359. #define HAVE_MMX2 0

  1360. #define HAVE_AMD3DNOW 0

  1361. #define RENAME(a) a ## _MMX

  1362. #include "swscale_template.c"

  1363. #endif

  1364. 

  1365. //MMX2 versions

  1366. #ifdef COMPILE_MMX2

  1367. #undef RENAME

  1368. #undef HAVE_MMX

  1369. #undef HAVE_MMX2

  1370. #undef HAVE_AMD3DNOW

  1371. #define HAVE_MMX 1

  1372. #define HAVE_MMX2 1

  1373. #define HAVE_AMD3DNOW 0

  1374. #define RENAME(a) a ## _MMX2

  1375. #include "swscale_template.c"

  1376. #endif

  1377. 

  1378. //3DNOW versions

  1379. #ifdef COMPILE_3DNOW

  1380. #undef RENAME

  1381. #undef HAVE_MMX

  1382. #undef HAVE_MMX2

  1383. #undef HAVE_AMD3DNOW

  1384. #define HAVE_MMX 1

  1385. #define HAVE_MMX2 0

  1386. #define HAVE_AMD3DNOW 1

  1387. #define RENAME(a) a ## _3DNow

  1388. #include "swscale_template.c"

  1389. #endif

  1390. 

  1391. #endif //ARCH_X86

  1392. 

  1393. // minor note: the HAVE_xyz are messed up after this line so don't use them

  1394. 

  1395. static double getSplineCoeff(double a, double b, double c, double d, double dist)

  1396. {

  1397. //    printf("%f %f %f %f %f\n", a,b,c,d,dist);

  1398.     if (dist<=1.0)      return ((d*dist + c)*dist + b)*dist +a;

  1399.     else                return getSplineCoeff(        0.0,

  1400.                                              b+ 2.0*c + 3.0*d,

  1401.                                                     c + 3.0*d,

  1402.                                             -b- 3.0*c - 6.0*d,

  1403.                                             dist-1.0);

  1404. }

  1405. 

  1406. static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,

  1407.                              int srcW, int dstW, int filterAlign, int one, int flags,

  1408.                              SwsVector *srcFilter, SwsVector *dstFilter, double param[2])

  1409. {

  1410.     int i;

  1411.     int filterSize;

  1412.     int filter2Size;

  1413.     int minFilterSize;

  1414.     int64_t *filter=NULL;

  1415.     int64_t *filter2=NULL;

  1416.     const int64_t fone= 1LL<<54;

  1417.     int ret= -1;

  1418. #if ARCH_X86

  1419.     if (flags & SWS_CPU_CAPS_MMX)

  1420.         __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)

  1421. #endif

  1422. 

  1423.     // NOTE: the +1 is for the MMX scaler which reads over the end

  1424.     *filterPos = av_malloc((dstW+1)*sizeof(int16_t));

  1425. 

  1426.     if (FFABS(xInc - 0x10000) <10) // unscaled

  1427.     {

  1428.         int i;

  1429.         filterSize= 1;

  1430.         filter= av_mallocz(dstW*sizeof(*filter)*filterSize);

  1431. 

  1432.         for (i=0; i<dstW; i++)

  1433.         {

  1434.             filter[i*filterSize]= fone;

  1435.             (*filterPos)[i]=i;

  1436.         }

  1437. 

  1438.     }

  1439.     else if (flags&SWS_POINT) // lame looking point sampling mode

  1440.     {

  1441.         int i;

  1442.         int xDstInSrc;

  1443.         filterSize= 1;

  1444.         filter= av_malloc(dstW*sizeof(*filter)*filterSize);

  1445. 

  1446.         xDstInSrc= xInc/2 - 0x8000;

  1447.         for (i=0; i<dstW; i++)

  1448.         {

  1449.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1450. 

  1451.             (*filterPos)[i]= xx;

  1452.             filter[i]= fone;

  1453.             xDstInSrc+= xInc;

  1454.         }

  1455.     }

  1456.     else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale

  1457.     {

  1458.         int i;

  1459.         int xDstInSrc;

  1460.         if      (flags&SWS_BICUBIC) filterSize= 4;

  1461.         else if (flags&SWS_X      ) filterSize= 4;

  1462.         else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA

  1463.         filter= av_malloc(dstW*sizeof(*filter)*filterSize);

  1464. 

  1465.         xDstInSrc= xInc/2 - 0x8000;

  1466.         for (i=0; i<dstW; i++)

  1467.         {

  1468.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1469.             int j;

  1470. 

  1471.             (*filterPos)[i]= xx;

  1472.                 //bilinear upscale / linear interpolate / area averaging

  1473.                 for (j=0; j<filterSize; j++)

  1474.                 {

  1475.                     int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);

  1476.                     if (coeff<0) coeff=0;

  1477.                     filter[i*filterSize + j]= coeff;

  1478.                     xx++;

  1479.                 }

  1480.             xDstInSrc+= xInc;

  1481.         }

  1482.     }

  1483.     else

  1484.     {

  1485.         int xDstInSrc;

  1486.         int sizeFactor;

  1487. 

  1488.         if      (flags&SWS_BICUBIC)      sizeFactor=  4;

  1489.         else if (flags&SWS_X)            sizeFactor=  8;

  1490.         else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear

  1491.         else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)

  1492.         else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;

  1493.         else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)

  1494.         else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)

  1495.         else if (flags&SWS_BILINEAR)     sizeFactor=  2;

  1496.         else {

  1497.             sizeFactor= 0; //GCC warning killer

  1498.             assert(0);

  1499.         }

  1500. 

  1501.         if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale

  1502.         else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;

  1503. 

  1504.         if (filterSize > srcW-2) filterSize=srcW-2;

  1505. 

  1506.         filter= av_malloc(dstW*sizeof(*filter)*filterSize);

  1507. 

  1508.         xDstInSrc= xInc - 0x10000;

  1509.         for (i=0; i<dstW; i++)

  1510.         {

  1511.             int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);

  1512.             int j;

  1513.             (*filterPos)[i]= xx;

  1514.             for (j=0; j<filterSize; j++)

  1515.             {

  1516.                 int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;

  1517.                 double floatd;

  1518.                 int64_t coeff;

  1519. 

  1520.                 if (xInc > 1<<16)

  1521.                     d= d*dstW/srcW;

  1522.                 floatd= d * (1.0/(1<<30));

  1523. 

  1524.                 if (flags & SWS_BICUBIC)

  1525.                 {

  1526.                     int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);

  1527.                     int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);

  1528.                     int64_t dd = ( d*d)>>30;

  1529.                     int64_t ddd= (dd*d)>>30;

  1530. 

  1531.                     if      (d < 1LL<<30)

  1532.                         coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);

  1533.                     else if (d < 1LL<<31)

  1534.                         coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);

  1535.                     else

  1536.                         coeff=0.0;

  1537.                     coeff *= fone>>(30+24);

  1538.                 }

  1539. /*                else if (flags & SWS_X)

  1540.                 {

  1541.                     double p= param ? param*0.01 : 0.3;

  1542.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1543.                     coeff*= pow(2.0, - p*d*d);

  1544.                 }*/

  1545.                 else if (flags & SWS_X)

  1546.                 {

  1547.                     double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

  1548.                     double c;

  1549. 

  1550.                     if (floatd<1.0)

  1551.                         c = cos(floatd*PI);

  1552.                     else

  1553.                         c=-1.0;

  1554.                     if (c<0.0)      c= -pow(-c, A);

  1555.                     else            c=  pow( c, A);

  1556.                     coeff= (c*0.5 + 0.5)*fone;

  1557.                 }

  1558.                 else if (flags & SWS_AREA)

  1559.                 {

  1560.                     int64_t d2= d - (1<<29);

  1561.                     if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));

  1562.                     else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));

  1563.                     else coeff=0.0;

  1564.                     coeff *= fone>>(30+16);

  1565.                 }

  1566.                 else if (flags & SWS_GAUSS)

  1567.                 {

  1568.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

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

  1570.                 }

  1571.                 else if (flags & SWS_SINC)

  1572.                 {

  1573.                     coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;

  1574.                 }

  1575.                 else if (flags & SWS_LANCZOS)

  1576.                 {

  1577.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1578.                     coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;

  1579.                     if (floatd>p) coeff=0;

  1580.                 }

  1581.                 else if (flags & SWS_BILINEAR)

  1582.                 {

  1583.                     coeff= (1<<30) - d;

  1584.                     if (coeff<0) coeff=0;

  1585.                     coeff *= fone >> 30;

  1586.                 }

  1587.                 else if (flags & SWS_SPLINE)

  1588.                 {

  1589.                     double p=-2.196152422706632;

  1590.                     coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;

  1591.                 }

  1592.                 else {

  1593.                     coeff= 0.0; //GCC warning killer

  1594.                     assert(0);

  1595.                 }

  1596. 

  1597.                 filter[i*filterSize + j]= coeff;

  1598.                 xx++;

  1599.             }

  1600.             xDstInSrc+= 2*xInc;

  1601.         }

  1602.     }

  1603. 

  1604.     /* apply src & dst Filter to filter -> filter2

  1605.        av_free(filter);

  1606.     */

  1607.     assert(filterSize>0);

  1608.     filter2Size= filterSize;

  1609.     if (srcFilter) filter2Size+= srcFilter->length - 1;

  1610.     if (dstFilter) filter2Size+= dstFilter->length - 1;

  1611.     assert(filter2Size>0);

  1612.     filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));

  1613. 

  1614.     for (i=0; i<dstW; i++)

  1615.     {

  1616.         int j, k;

  1617. 

  1618.         if(srcFilter){

  1619.             for (k=0; k<srcFilter->length; k++){

  1620.                 for (j=0; j<filterSize; j++)

  1621.                     filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];

  1622.             }

  1623.         }else{

  1624.             for (j=0; j<filterSize; j++)

  1625.                 filter2[i*filter2Size + j]= filter[i*filterSize + j];

  1626.         }

  1627.         //FIXME dstFilter

  1628. 

  1629.         (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

  1630.     }

  1631.     av_freep(&filter);

  1632. 

  1633.     /* try to reduce the filter-size (step1 find size and shift left) */

  1634.     // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).

  1635.     minFilterSize= 0;

  1636.     for (i=dstW-1; i>=0; i--)

  1637.     {

  1638.         int min= filter2Size;

  1639.         int j;

  1640.         int64_t cutOff=0.0;

  1641. 

  1642.         /* get rid off near zero elements on the left by shifting left */

  1643.         for (j=0; j<filter2Size; j++)

  1644.         {

  1645.             int k;

  1646.             cutOff += FFABS(filter2[i*filter2Size]);

  1647. 

  1648.             if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;

  1649. 

  1650.             /* preserve monotonicity because the core can't handle the filter otherwise */

  1651.             if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

  1652. 

  1653.             // move filter coefficients left

  1654.             for (k=1; k<filter2Size; k++)

  1655.                 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];

  1656.             filter2[i*filter2Size + k - 1]= 0;

  1657.             (*filterPos)[i]++;

  1658.         }

  1659. 

  1660.         cutOff=0;

  1661.         /* count near zeros on the right */

  1662.         for (j=filter2Size-1; j>0; j--)

  1663.         {

  1664.             cutOff += FFABS(filter2[i*filter2Size + j]);

  1665. 

  1666.             if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;

  1667.             min--;

  1668.         }

  1669. 

  1670.         if (min>minFilterSize) minFilterSize= min;

  1671.     }

  1672. 

  1673.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1674.         // we can handle the special case 4,

  1675.         // so we don't want to go to the full 8

  1676.         if (minFilterSize < 5)

  1677.             filterAlign = 4;

  1678. 

  1679.         // We really don't want to waste our time

  1680.         // doing useless computation, so fall back on

  1681.         // the scalar C code for very small filters.

  1682.         // Vectorizing is worth it only if you have a

  1683.         // decent-sized vector.

  1684.         if (minFilterSize < 3)

  1685.             filterAlign = 1;

  1686.     }

  1687. 

  1688.     if (flags & SWS_CPU_CAPS_MMX) {

  1689.         // special case for unscaled vertical filtering

  1690.         if (minFilterSize == 1 && filterAlign == 2)

  1691.             filterAlign= 1;

  1692.     }

  1693. 

  1694.     assert(minFilterSize > 0);

  1695.     filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));

  1696.     assert(filterSize > 0);

  1697.     filter= av_malloc(filterSize*dstW*sizeof(*filter));

  1698.     if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)

  1699.         goto error;

  1700.     *outFilterSize= filterSize;

  1701. 

  1702.     if (flags&SWS_PRINT_INFO)

  1703.         av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);

  1704.     /* try to reduce the filter-size (step2 reduce it) */

  1705.     for (i=0; i<dstW; i++)

  1706.     {

  1707.         int j;

  1708. 

  1709.         for (j=0; j<filterSize; j++)

  1710.         {

  1711.             if (j>=filter2Size) filter[i*filterSize + j]= 0;

  1712.             else               filter[i*filterSize + j]= filter2[i*filter2Size + j];

  1713.             if((flags & SWS_BITEXACT) && j>=minFilterSize)

  1714.                 filter[i*filterSize + j]= 0;

  1715.         }

  1716.     }

  1717. 

  1718. 

  1719.     //FIXME try to align filterPos if possible

  1720. 

  1721.     //fix borders

  1722.     for (i=0; i<dstW; i++)

  1723.     {

  1724.         int j;

  1725.         if ((*filterPos)[i] < 0)

  1726.         {

  1727.             // move filter coefficients left to compensate for filterPos

  1728.             for (j=1; j<filterSize; j++)

  1729.             {

  1730.                 int left= FFMAX(j + (*filterPos)[i], 0);

  1731.                 filter[i*filterSize + left] += filter[i*filterSize + j];

  1732.                 filter[i*filterSize + j]=0;

  1733.             }

  1734.             (*filterPos)[i]= 0;

  1735.         }

  1736. 

  1737.         if ((*filterPos)[i] + filterSize > srcW)

  1738.         {

  1739.             int shift= (*filterPos)[i] + filterSize - srcW;

  1740.             // move filter coefficients right to compensate for filterPos

  1741.             for (j=filterSize-2; j>=0; j--)

  1742.             {

  1743.                 int right= FFMIN(j + shift, filterSize-1);

  1744.                 filter[i*filterSize +right] += filter[i*filterSize +j];

  1745.                 filter[i*filterSize +j]=0;

  1746.             }

  1747.             (*filterPos)[i]= srcW - filterSize;

  1748.         }

  1749.     }

  1750. 

  1751.     // Note the +1 is for the MMX scaler which reads over the end

  1752.     /* align at 16 for AltiVec (needed by hScale_altivec_real) */

  1753.     *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));

  1754. 

  1755.     /* normalize & store in outFilter */

  1756.     for (i=0; i<dstW; i++)

  1757.     {

  1758.         int j;

  1759.         int64_t error=0;

  1760.         int64_t sum=0;

  1761. 

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

  1763.         {

  1764.             sum+= filter[i*filterSize + j];

  1765.         }

  1766.         sum= (sum + one/2)/ one;

  1767.         for (j=0; j<*outFilterSize; j++)

  1768.         {

  1769.             int64_t v= filter[i*filterSize + j] + error;

  1770.             int intV= ROUNDED_DIV(v, sum);

  1771.             (*outFilter)[i*(*outFilterSize) + j]= intV;

  1772.             error= v - intV*sum;

  1773.         }

  1774.     }

  1775. 

  1776.     (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end

  1777.     for (i=0; i<*outFilterSize; i++)

  1778.     {

  1779.         int j= dstW*(*outFilterSize);

  1780.         (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

  1781.     }

  1782. 

  1783.     ret=0;

  1784. error:

  1785.     av_free(filter);

  1786.     av_free(filter2);

  1787.     return ret;

  1788. }

  1789. 

  1790. #ifdef COMPILE_MMX2

  1791. static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)

  1792. {

  1793.     uint8_t *fragmentA;

  1794.     x86_reg imm8OfPShufW1A;

  1795.     x86_reg imm8OfPShufW2A;

  1796.     x86_reg fragmentLengthA;

  1797.     uint8_t *fragmentB;

  1798.     x86_reg imm8OfPShufW1B;

  1799.     x86_reg imm8OfPShufW2B;

  1800.     x86_reg fragmentLengthB;

  1801.     int fragmentPos;

  1802. 

  1803.     int xpos, i;

  1804. 

  1805.     // create an optimized horizontal scaling routine

  1806. 

  1807.     //code fragment

  1808. 

  1809.     __asm__ volatile(

  1810.         "jmp                         9f                 \n\t"

  1811.     // Begin

  1812.         "0:                                             \n\t"

  1813.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1814.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1815.         "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"

  1816.         "punpcklbw                %%mm7, %%mm1          \n\t"

  1817.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1818.         "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"

  1819.         "1:                                             \n\t"

  1820.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1821.         "2:                                             \n\t"

  1822.         "psubw                    %%mm1, %%mm0          \n\t"

  1823.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1824.         "pmullw                   %%mm3, %%mm0          \n\t"

  1825.         "psllw                       $7, %%mm1          \n\t"

  1826.         "paddw                    %%mm1, %%mm0          \n\t"

  1827. 

  1828.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1829. 

  1830.         "add                         $8, %%"REG_a"      \n\t"

  1831.     // End

  1832.         "9:                                             \n\t"

  1833. //        "int $3                                         \n\t"

  1834.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1835.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1836.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1837.         "dec                         %1                 \n\t"

  1838.         "dec                         %2                 \n\t"

  1839.         "sub                         %0, %1             \n\t"

  1840.         "sub                         %0, %2             \n\t"

  1841.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1842.         "sub                         %0, %3             \n\t"

  1843. 

  1844. 

  1845.         :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),

  1846.         "=r" (fragmentLengthA)

  1847.     );

  1848. 

  1849.     __asm__ volatile(

  1850.         "jmp                         9f                 \n\t"

  1851.     // Begin

  1852.         "0:                                             \n\t"

  1853.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1854.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1855.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1856.         "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"

  1857.         "1:                                             \n\t"

  1858.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1859.         "2:                                             \n\t"

  1860.         "psubw                    %%mm1, %%mm0          \n\t"

  1861.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1862.         "pmullw                   %%mm3, %%mm0          \n\t"

  1863.         "psllw                       $7, %%mm1          \n\t"

  1864.         "paddw                    %%mm1, %%mm0          \n\t"

  1865. 

  1866.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1867. 

  1868.         "add                         $8, %%"REG_a"      \n\t"

  1869.     // End

  1870.         "9:                                             \n\t"

  1871. //        "int                       $3                   \n\t"

  1872.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1873.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1874.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1875.         "dec                         %1                 \n\t"

  1876.         "dec                         %2                 \n\t"

  1877.         "sub                         %0, %1             \n\t"

  1878.         "sub                         %0, %2             \n\t"

  1879.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1880.         "sub                         %0, %3             \n\t"

  1881. 

  1882. 

  1883.         :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),

  1884.         "=r" (fragmentLengthB)

  1885.     );

  1886. 

  1887.     xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers

  1888.     fragmentPos=0;

  1889. 

  1890.     for (i=0; i<dstW/numSplits; i++)

  1891.     {

  1892.         int xx=xpos>>16;

  1893. 

  1894.         if ((i&3) == 0)

  1895.         {

  1896.             int a=0;

  1897.             int b=((xpos+xInc)>>16) - xx;

  1898.             int c=((xpos+xInc*2)>>16) - xx;

  1899.             int d=((xpos+xInc*3)>>16) - xx;

  1900. 

  1901.             filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;

  1902.             filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;

  1903.             filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;

  1904.             filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;

  1905.             filterPos[i/2]= xx;

  1906. 

  1907.             if (d+1<4)

  1908.             {

  1909.                 int maxShift= 3-(d+1);

  1910.                 int shift=0;

  1911. 

  1912.                 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);

  1913. 

  1914.                 funnyCode[fragmentPos + imm8OfPShufW1B]=

  1915.                     (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);

  1916.                 funnyCode[fragmentPos + imm8OfPShufW2B]=

  1917.                     a | (b<<2) | (c<<4) | (d<<6);

  1918. 

  1919.                 if (i+3>=dstW) shift=maxShift; //avoid overread

  1920.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

  1921. 

  1922.                 if (shift && i>=shift)

  1923.                 {

  1924.                     funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;

  1925.                     funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;

  1926.                     filterPos[i/2]-=shift;

  1927.                 }

  1928. 

  1929.                 fragmentPos+= fragmentLengthB;

  1930.             }

  1931.             else

  1932.             {

  1933.                 int maxShift= 3-d;

  1934.                 int shift=0;

  1935. 

  1936.                 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);

  1937. 

  1938.                 funnyCode[fragmentPos + imm8OfPShufW1A]=

  1939.                 funnyCode[fragmentPos + imm8OfPShufW2A]=

  1940.                     a | (b<<2) | (c<<4) | (d<<6);

  1941. 

  1942.                 if (i+4>=dstW) shift=maxShift; //avoid overread

  1943.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align

  1944. 

  1945.                 if (shift && i>=shift)

  1946.                 {

  1947.                     funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;

  1948.                     funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;

  1949.                     filterPos[i/2]-=shift;

  1950.                 }

  1951. 

  1952.                 fragmentPos+= fragmentLengthA;

  1953.             }

  1954. 

  1955.             funnyCode[fragmentPos]= RET;

  1956.         }

  1957.         xpos+=xInc;

  1958.     }

  1959.     filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part

  1960. }

  1961. #endif /* COMPILE_MMX2 */

  1962. 

  1963. static void globalInit(void){

  1964.     // generating tables:

  1965.     int i;

  1966.     for (i=0; i<768; i++){

  1967.         int c= av_clip_uint8(i-256);

  1968.         clip_table[i]=c;

  1969.     }

  1970. }

  1971. 

  1972. static SwsFunc getSwsFunc(SwsContext *c)

  1973. {

  1974. #if CONFIG_RUNTIME_CPUDETECT

  1975.     int flags = c->flags;

  1976. 

  1977. #if ARCH_X86 && CONFIG_GPL

  1978.     // ordered per speed fastest first

  1979.     if (flags & SWS_CPU_CAPS_MMX2) {

  1980.         sws_init_swScale_MMX2(c);

  1981.         return swScale_MMX2;

  1982.     } else if (flags & SWS_CPU_CAPS_3DNOW) {

  1983.         sws_init_swScale_3DNow(c);

  1984.         return swScale_3DNow;

  1985.     } else if (flags & SWS_CPU_CAPS_MMX) {

  1986.         sws_init_swScale_MMX(c);

  1987.         return swScale_MMX;

  1988.     } else {

  1989.         sws_init_swScale_C(c);

  1990.         return swScale_C;

  1991.     }

  1992. 

  1993. #else

  1994. #if ARCH_PPC

  1995.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1996.         sws_init_swScale_altivec(c);

  1997.         return swScale_altivec;

  1998.     } else {

  1999.         sws_init_swScale_C(c);

  2000.         return swScale_C;

  2001.     }

  2002. #endif

  2003.     sws_init_swScale_C(c);

  2004.     return swScale_C;

  2005. #endif /* ARCH_X86 && CONFIG_GPL */

  2006. #else //CONFIG_RUNTIME_CPUDETECT

  2007. #if   HAVE_MMX2

  2008.     sws_init_swScale_MMX2(c);

  2009.     return swScale_MMX2;

  2010. #elif HAVE_AMD3DNOW

  2011.     sws_init_swScale_3DNow(c);

  2012.     return swScale_3DNow;

  2013. #elif HAVE_MMX

  2014.     sws_init_swScale_MMX(c);

  2015.     return swScale_MMX;

  2016. #elif HAVE_ALTIVEC

  2017.     sws_init_swScale_altivec(c);

  2018.     return swScale_altivec;

  2019. #else

  2020.     sws_init_swScale_C(c);

  2021.     return swScale_C;

  2022. #endif

  2023. #endif //!CONFIG_RUNTIME_CPUDETECT

  2024. }

  2025. 

  2026. static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2027.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  2028.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2029.     /* Copy Y plane */

  2030.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  2031.         memcpy(dst, src[0], srcSliceH*dstStride[0]);

  2032.     else

  2033.     {

  2034.         int i;

  2035.         const uint8_t *srcPtr= src[0];

  2036.         uint8_t *dstPtr= dst;

  2037.         for (i=0; i<srcSliceH; i++)

  2038.         {

  2039.             memcpy(dstPtr, srcPtr, c->srcW);

  2040.             srcPtr+= srcStride[0];

  2041.             dstPtr+= dstStride[0];

  2042.         }

  2043.     }

  2044.     dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  2045.     if (c->dstFormat == PIX_FMT_NV12)

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

  2047.     else

  2048.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  2049. 

  2050.     return srcSliceH;

  2051. }

  2052. 

  2053. static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2054.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  2055.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2056. 

  2057.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  2058. 

  2059.     return srcSliceH;

  2060. }

  2061. 

  2062. static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2063.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

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

  2065. 

  2066.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  2067. 

  2068.     return srcSliceH;

  2069. }

  2070. 

  2071. static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

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

  2073.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2074. 

  2075.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  2076. 

  2077.     return srcSliceH;

  2078. }

  2079. 

  2080. static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2081.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  2082.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2083. 

  2084.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  2085. 

  2086.     return srcSliceH;

  2087. }

  2088. 

  2089. static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2090.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  2091.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2092.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  2093.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  2094. 

  2095.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2096. 

  2097.     if (dstParam[3])

  2098.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2099. 

  2100.     return srcSliceH;

  2101. }

  2102. 

  2103. static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2104.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  2105.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2106.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  2107.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  2108. 

  2109.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2110. 

  2111.     return srcSliceH;

  2112. }

  2113. 

  2114. static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2115.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  2116.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2117.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  2118.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  2119. 

  2120.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2121. 

  2122.     if (dstParam[3])

  2123.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2124. 

  2125.     return srcSliceH;

  2126. }

  2127. 

  2128. static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2129.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  2130.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2131.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  2132.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  2133. 

  2134.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2135. 

  2136.     return srcSliceH;

  2137. }

  2138. 

  2139. static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2140.                           int srcSliceH, uint8_t* dst[], int dstStride[]){

  2141.     const enum PixelFormat srcFormat= c->srcFormat;

  2142.     const enum PixelFormat dstFormat= c->dstFormat;

  2143.     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,

  2144.                  const uint8_t *palette)=NULL;

  2145.     int i;

  2146.     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2147.     uint8_t *srcPtr= src[0];

  2148. 

  2149.     if (!usePal(srcFormat))

  2150.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2151.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  2152. 

  2153.     switch(dstFormat){

  2154.     case PIX_FMT_RGB32  : conv = palette8topacked32; break;

  2155.     case PIX_FMT_BGR32  : conv = palette8topacked32; break;

  2156.     case PIX_FMT_BGR32_1: conv = palette8topacked32; break;

  2157.     case PIX_FMT_RGB32_1: conv = palette8topacked32; break;

  2158.     case PIX_FMT_RGB24  : conv = palette8topacked24; break;

  2159.     case PIX_FMT_BGR24  : conv = palette8topacked24; break;

  2160.     default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2161.                     sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2162.     }

  2163. 

  2164. 

  2165.     for (i=0; i<srcSliceH; i++) {

  2166.         conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  2167.         srcPtr+= srcStride[0];

  2168.         dstPtr+= dstStride[0];

  2169.     }

  2170. 

  2171.     return srcSliceH;

  2172. }

  2173. 

  2174. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  2175. static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2176.                           int srcSliceH, uint8_t* dst[], int dstStride[]){

  2177.     const enum PixelFormat srcFormat= c->srcFormat;

  2178.     const enum PixelFormat dstFormat= c->dstFormat;

  2179.     const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;

  2180.     const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;

  2181.     const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  2182.     const int dstId= fmt_depth(dstFormat) >> 2;

  2183.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  2184. 

  2185.     /* BGR -> BGR */

  2186.     if (  (isBGR(srcFormat) && isBGR(dstFormat))

  2187.        || (isRGB(srcFormat) && isRGB(dstFormat))){

  2188.         switch(srcId | (dstId<<4)){

  2189.         case 0x34: conv= rgb16to15; break;

  2190.         case 0x36: conv= rgb24to15; break;

  2191.         case 0x38: conv= rgb32to15; break;

  2192.         case 0x43: conv= rgb15to16; break;

  2193.         case 0x46: conv= rgb24to16; break;

  2194.         case 0x48: conv= rgb32to16; break;

  2195.         case 0x63: conv= rgb15to24; break;

  2196.         case 0x64: conv= rgb16to24; break;

  2197.         case 0x68: conv= rgb32to24; break;

  2198.         case 0x83: conv= rgb15to32; break;

  2199.         case 0x84: conv= rgb16to32; break;

  2200.         case 0x86: conv= rgb24to32; break;

  2201.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2202.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2203.         }

  2204.     }else if (  (isBGR(srcFormat) && isRGB(dstFormat))

  2205.              || (isRGB(srcFormat) && isBGR(dstFormat))){

  2206.         switch(srcId | (dstId<<4)){

  2207.         case 0x33: conv= rgb15tobgr15; break;

  2208.         case 0x34: conv= rgb16tobgr15; break;

  2209.         case 0x36: conv= rgb24tobgr15; break;

  2210.         case 0x38: conv= rgb32tobgr15; break;

  2211.         case 0x43: conv= rgb15tobgr16; break;

  2212.         case 0x44: conv= rgb16tobgr16; break;

  2213.         case 0x46: conv= rgb24tobgr16; break;

  2214.         case 0x48: conv= rgb32tobgr16; break;

  2215.         case 0x63: conv= rgb15tobgr24; break;

  2216.         case 0x64: conv= rgb16tobgr24; break;

  2217.         case 0x66: conv= rgb24tobgr24; break;

  2218.         case 0x68: conv= rgb32tobgr24; break;

  2219.         case 0x83: conv= rgb15tobgr32; break;

  2220.         case 0x84: conv= rgb16tobgr32; break;

  2221.         case 0x86: conv= rgb24tobgr32; break;

  2222.         case 0x88: conv= rgb32tobgr32; break;

  2223.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2224.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2225.         }

  2226.     }else{

  2227.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2228.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  2229.     }

  2230. 

  2231.     if(conv)

  2232.     {

  2233.         uint8_t *srcPtr= src[0];

  2234.         if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)

  2235.             srcPtr += ALT32_CORR;

  2236. 

  2237.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  2238.             conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  2239.         else

  2240.         {

  2241.             int i;

  2242.             uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2243. 

  2244.             for (i=0; i<srcSliceH; i++)

  2245.             {

  2246.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  2247.                 srcPtr+= srcStride[0];

  2248.                 dstPtr+= dstStride[0];

  2249.             }

  2250.         }

  2251.     }

  2252.     return srcSliceH;

  2253. }

  2254. 

  2255. static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2256.                               int srcSliceH, uint8_t* dst[], int dstStride[]){

  2257. 

  2258.     rgb24toyv12(

  2259.         src[0],

  2260.         dst[0]+ srcSliceY    *dstStride[0],

  2261.         dst[1]+(srcSliceY>>1)*dstStride[1],

  2262.         dst[2]+(srcSliceY>>1)*dstStride[2],

  2263.         c->srcW, srcSliceH,

  2264.         dstStride[0], dstStride[1], srcStride[0]);

  2265.     if (dst[3])

  2266.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2267.     return srcSliceH;

  2268. }

  2269. 

  2270. static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2271.                              int srcSliceH, uint8_t* dst[], int dstStride[]){

  2272.     int i;

  2273. 

  2274.     /* copy Y */

  2275.     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)

  2276.         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);

  2277.     else{

  2278.         uint8_t *srcPtr= src[0];

  2279.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2280. 

  2281.         for (i=0; i<srcSliceH; i++)

  2282.         {

  2283.             memcpy(dstPtr, srcPtr, c->srcW);

  2284.             srcPtr+= srcStride[0];

  2285.             dstPtr+= dstStride[0];

  2286.         }

  2287.     }

  2288. 

  2289.     if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P){

  2290.         planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  2291.                  srcSliceH >> 2, srcStride[1], dstStride[1]);

  2292.         planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  2293.                  srcSliceH >> 2, srcStride[2], dstStride[2]);

  2294.     }else{

  2295.         planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  2296.                  srcSliceH >> 2, srcStride[1], dstStride[2]);

  2297.         planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  2298.                  srcSliceH >> 2, srcStride[2], dstStride[1]);

  2299.     }

  2300.     if (dst[3])

  2301.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2302.     return srcSliceH;

  2303. }

  2304. 

  2305. /* unscaled copy like stuff (assumes nearly identical formats) */

  2306. static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2307.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  2308. {

  2309.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  2310.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  2311.     else

  2312.     {

  2313.         int i;

  2314.         uint8_t *srcPtr= src[0];

  2315.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2316.         int length=0;

  2317. 

  2318.         /* universal length finder */

  2319.         while(length+c->srcW <= FFABS(dstStride[0])

  2320.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  2321.         assert(length!=0);

  2322. 

  2323.         for (i=0; i<srcSliceH; i++)

  2324.         {

  2325.             memcpy(dstPtr, srcPtr, length);

  2326.             srcPtr+= srcStride[0];

  2327.             dstPtr+= dstStride[0];

  2328.         }

  2329.     }

  2330.     return srcSliceH;

  2331. }

  2332. 

  2333. static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2334.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  2335. {

  2336.     int plane, i, j;

  2337.     for (plane=0; plane<4; plane++)

  2338.     {

  2339.         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  2340.         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  2341.         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  2342.         uint8_t *srcPtr= src[plane];

  2343.         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  2344. 

  2345.         if (!dst[plane]) continue;

  2346.         // ignore palette for GRAY8

  2347.         if (plane == 1 && !dst[2]) continue;

  2348.         if (!src[plane] || (plane == 1 && !src[2])){

  2349.             if(is16BPS(c->dstFormat))

  2350.                 length*=2;

  2351.             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);

  2352.         }else

  2353.         {

  2354.             if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)){

  2355.                 if (!isBE(c->srcFormat)) srcPtr++;

  2356.                 for (i=0; i<height; i++){

  2357.                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  2358.                     srcPtr+= srcStride[plane];

  2359.                     dstPtr+= dstStride[plane];

  2360.                 }

  2361.             }else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)){

  2362.                 for (i=0; i<height; i++){

  2363.                     for (j=0; j<length; j++){

  2364.                         dstPtr[ j<<1   ] = srcPtr[j];

  2365.                         dstPtr[(j<<1)+1] = srcPtr[j];

  2366.                     }

  2367.                     srcPtr+= srcStride[plane];

  2368.                     dstPtr+= dstStride[plane];

  2369.                 }

  2370.             }else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)

  2371.                   && isBE(c->srcFormat) != isBE(c->dstFormat)){

  2372. 

  2373.                 for (i=0; i<height; i++){

  2374.                     for (j=0; j<length; j++)

  2375.                         ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);

  2376.                     srcPtr+= srcStride[plane];

  2377.                     dstPtr+= dstStride[plane];

  2378.                 }

  2379.             } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)

  2380.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);

  2381.             else

  2382.             {

  2383.                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))

  2384.                     length*=2;

  2385.                 for (i=0; i<height; i++)

  2386.                 {

  2387.                     memcpy(dstPtr, srcPtr, length);

  2388.                     srcPtr+= srcStride[plane];

  2389.                     dstPtr+= dstStride[plane];

  2390.                 }

  2391.             }

  2392.         }

  2393.     }

  2394.     return srcSliceH;

  2395. }

  2396. 

  2397. 

  2398. static void getSubSampleFactors(int *h, int *v, int format){

  2399.     switch(format){

  2400.     case PIX_FMT_UYVY422:

  2401.     case PIX_FMT_YUYV422:

  2402.         *h=1;

  2403.         *v=0;

  2404.         break;

  2405.     case PIX_FMT_YUV420P:

  2406.     case PIX_FMT_YUV420PLE:

  2407.     case PIX_FMT_YUV420PBE:

  2408.     case PIX_FMT_YUVA420P:

  2409.     case PIX_FMT_GRAY16BE:

  2410.     case PIX_FMT_GRAY16LE:

  2411.     case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented

  2412.     case PIX_FMT_NV12:

  2413.     case PIX_FMT_NV21:

  2414.         *h=1;

  2415.         *v=1;

  2416.         break;

  2417.     case PIX_FMT_YUV440P:

  2418.         *h=0;

  2419.         *v=1;

  2420.         break;

  2421.     case PIX_FMT_YUV410P:

  2422.         *h=2;

  2423.         *v=2;

  2424.         break;

  2425.     case PIX_FMT_YUV444P:

  2426.     case PIX_FMT_YUV444PLE:

  2427.     case PIX_FMT_YUV444PBE:

  2428.         *h=0;

  2429.         *v=0;

  2430.         break;

  2431.     case PIX_FMT_YUV422P:

  2432.     case PIX_FMT_YUV422PLE:

  2433.     case PIX_FMT_YUV422PBE:

  2434.         *h=1;

  2435.         *v=0;

  2436.         break;

  2437.     case PIX_FMT_YUV411P:

  2438.         *h=2;

  2439.         *v=0;

  2440.         break;

  2441.     default:

  2442.         *h=0;

  2443.         *v=0;

  2444.         break;

  2445.     }

  2446. }

  2447. 

  2448. static uint16_t roundToInt16(int64_t f){

  2449.     int r= (f + (1<<15))>>16;

  2450.          if (r<-0x7FFF) return 0x8000;

  2451.     else if (r> 0x7FFF) return 0x7FFF;

  2452.     else                return r;

  2453. }

  2454. 

  2455. int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){

  2456.     int64_t crv =  inv_table[0];

  2457.     int64_t cbu =  inv_table[1];

  2458.     int64_t cgu = -inv_table[2];

  2459.     int64_t cgv = -inv_table[3];

  2460.     int64_t cy  = 1<<16;

  2461.     int64_t oy  = 0;

  2462. 

  2463.     memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);

  2464.     memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);

  2465. 

  2466.     c->brightness= brightness;

  2467.     c->contrast  = contrast;

  2468.     c->saturation= saturation;

  2469.     c->srcRange  = srcRange;

  2470.     c->dstRange  = dstRange;

  2471.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  2472. 

  2473.     c->uOffset=   0x0400040004000400LL;

  2474.     c->vOffset=   0x0400040004000400LL;

  2475. 

  2476.     if (!srcRange){

  2477.         cy= (cy*255) / 219;

  2478.         oy= 16<<16;

  2479.     }else{

  2480.         crv= (crv*224) / 255;

  2481.         cbu= (cbu*224) / 255;

  2482.         cgu= (cgu*224) / 255;

  2483.         cgv= (cgv*224) / 255;

  2484.     }

  2485. 

  2486.     cy = (cy *contrast             )>>16;

  2487.     crv= (crv*contrast * saturation)>>32;

  2488.     cbu= (cbu*contrast * saturation)>>32;

  2489.     cgu= (cgu*contrast * saturation)>>32;

  2490.     cgv= (cgv*contrast * saturation)>>32;

  2491. 

  2492.     oy -= 256*brightness;

  2493. 

  2494.     c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;

  2495.     c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;

  2496.     c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;

  2497.     c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;

  2498.     c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;

  2499.     c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;

  2500. 

  2501.     c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);

  2502.     c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);

  2503.     c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);

  2504.     c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);

  2505.     c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);

  2506.     c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);

  2507. 

  2508.     ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);

  2509.     //FIXME factorize

  2510. 

  2511. #ifdef COMPILE_ALTIVEC

  2512.     if (c->flags & SWS_CPU_CAPS_ALTIVEC)

  2513.         ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);

  2514. #endif

  2515.     return 0;

  2516. }

  2517. 

  2518. int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){

  2519.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  2520. 

  2521.     *inv_table = c->srcColorspaceTable;

  2522.     *table     = c->dstColorspaceTable;

  2523.     *srcRange  = c->srcRange;

  2524.     *dstRange  = c->dstRange;

  2525.     *brightness= c->brightness;

  2526.     *contrast  = c->contrast;

  2527.     *saturation= c->saturation;

  2528. 

  2529.     return 0;

  2530. }

  2531. 

  2532. static int handle_jpeg(enum PixelFormat *format)

  2533. {

  2534.     switch (*format) {

  2535.         case PIX_FMT_YUVJ420P:

  2536.             *format = PIX_FMT_YUV420P;

  2537.             return 1;

  2538.         case PIX_FMT_YUVJ422P:

  2539.             *format = PIX_FMT_YUV422P;

  2540.             return 1;

  2541.         case PIX_FMT_YUVJ444P:

  2542.             *format = PIX_FMT_YUV444P;

  2543.             return 1;

  2544.         case PIX_FMT_YUVJ440P:

  2545.             *format = PIX_FMT_YUV440P;

  2546.             return 1;

  2547.         default:

  2548.             return 0;

  2549.     }

  2550. }

  2551. 

  2552. SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,

  2553.                            SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)

  2554. {

  2555. 

  2556.     SwsContext *c;

  2557.     int i;

  2558.     int usesVFilter, usesHFilter;

  2559.     int unscaled, needsDither;

  2560.     int srcRange, dstRange;

  2561.     SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};

  2562. #if ARCH_X86

  2563.     if (flags & SWS_CPU_CAPS_MMX)

  2564.         __asm__ volatile("emms\n\t"::: "memory");

  2565. #endif

  2566. 

  2567. #if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off

  2568.     flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);

  2569. #if   HAVE_MMX2

  2570.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  2571. #elif HAVE_AMD3DNOW

  2572.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  2573. #elif HAVE_MMX

  2574.     flags |= SWS_CPU_CAPS_MMX;

  2575. #elif HAVE_ALTIVEC

  2576.     flags |= SWS_CPU_CAPS_ALTIVEC;

  2577. #elif ARCH_BFIN

  2578.     flags |= SWS_CPU_CAPS_BFIN;

  2579. #endif

  2580. #endif /* CONFIG_RUNTIME_CPUDETECT */

  2581.     if (clip_table[512] != 255) globalInit();

  2582.     if (!rgb15to16) sws_rgb2rgb_init(flags);

  2583. 

  2584.     unscaled = (srcW == dstW && srcH == dstH);

  2585.     needsDither= (isBGR(dstFormat) || isRGB(dstFormat))

  2586.         && (fmt_depth(dstFormat))<24

  2587.         && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

  2588. 

  2589.     srcRange = handle_jpeg(&srcFormat);

  2590.     dstRange = handle_jpeg(&dstFormat);

  2591. 

  2592.     if (!isSupportedIn(srcFormat))

  2593.     {

  2594.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));

  2595.         return NULL;

  2596.     }

  2597.     if (!isSupportedOut(dstFormat))

  2598.     {

  2599.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));

  2600.         return NULL;

  2601.     }

  2602. 

  2603.     i= flags & ( SWS_POINT

  2604.                 |SWS_AREA

  2605.                 |SWS_BILINEAR

  2606.                 |SWS_FAST_BILINEAR

  2607.                 |SWS_BICUBIC

  2608.                 |SWS_X

  2609.                 |SWS_GAUSS

  2610.                 |SWS_LANCZOS

  2611.                 |SWS_SINC

  2612.                 |SWS_SPLINE

  2613.                 |SWS_BICUBLIN);

  2614.     if(!i || (i & (i-1)))

  2615.     {

  2616.         av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");

  2617.         return NULL;

  2618.     }

  2619. 

  2620.     /* sanity check */

  2621.     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

  2622.     {

  2623.         av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",

  2624.                srcW, srcH, dstW, dstH);

  2625.         return NULL;

  2626.     }

  2627.     if(srcW > VOFW || dstW > VOFW){

  2628.         av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");

  2629.         return NULL;

  2630.     }

  2631. 

  2632.     if (!dstFilter) dstFilter= &dummyFilter;

  2633.     if (!srcFilter) srcFilter= &dummyFilter;

  2634. 

  2635.     c= av_mallocz(sizeof(SwsContext));

  2636. 

  2637.     c->av_class = &sws_context_class;

  2638.     c->srcW= srcW;

  2639.     c->srcH= srcH;

  2640.     c->dstW= dstW;

  2641.     c->dstH= dstH;

  2642.     c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;

  2643.     c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;

  2644.     c->flags= flags;

  2645.     c->dstFormat= dstFormat;

  2646.     c->srcFormat= srcFormat;

  2647.     c->vRounder= 4* 0x0001000100010001ULL;

  2648. 

  2649.     usesHFilter= usesVFilter= 0;

  2650.     if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;

  2651.     if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;

  2652.     if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;

  2653.     if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;

  2654.     if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;

  2655.     if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;

  2656.     if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;

  2657.     if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;

  2658. 

  2659.     getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);

  2660.     getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

  2661. 

  2662.     // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation

  2663.     if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

  2664. 

  2665.     // drop some chroma lines if the user wants it

  2666.     c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;

  2667.     c->chrSrcVSubSample+= c->vChrDrop;

  2668. 

  2669.     // drop every other pixel for chroma calculation unless user wants full chroma

  2670.     if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)

  2671.       && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8

  2672.       && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4

  2673.       && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE

  2674.       && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2675.         c->chrSrcHSubSample=1;

  2676. 

  2677.     if (param){

  2678.         c->param[0] = param[0];

  2679.         c->param[1] = param[1];

  2680.     }else{

  2681.         c->param[0] =

  2682.         c->param[1] = SWS_PARAM_DEFAULT;

  2683.     }

  2684. 

  2685.     // Note the -((-x)>>y) is so that we always round toward +inf.

  2686.     c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);

  2687.     c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);

  2688.     c->chrDstW= -((-dstW) >> c->chrDstHSubSample);

  2689.     c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

  2690. 

  2691.     sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);

  2692. 

  2693.     /* unscaled special cases */

  2694.     if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))

  2695.     {

  2696.         /* yv12_to_nv12 */

  2697.         if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))

  2698.         {

  2699.             c->swScale= PlanarToNV12Wrapper;

  2700.         }

  2701.         /* yuv2bgr */

  2702.         if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))

  2703.             && !(flags & SWS_ACCURATE_RND) && !(dstH&1))

  2704.         {

  2705.             c->swScale= ff_yuv2rgb_get_func_ptr(c);

  2706.         }

  2707. 

  2708.         if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT))

  2709.         {

  2710.             c->swScale= yvu9toyv12Wrapper;

  2711.         }

  2712. 

  2713.         /* bgr24toYV12 */

  2714.         if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))

  2715.             c->swScale= bgr24toyv12Wrapper;

  2716. 

  2717.         /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  2718.         if (  (isBGR(srcFormat) || isRGB(srcFormat))

  2719.            && (isBGR(dstFormat) || isRGB(dstFormat))

  2720.            && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  2721.            && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  2722.            && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  2723.            && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  2724.            && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  2725.            && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  2726.            && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  2727.            && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE

  2728.                                              && dstFormat != PIX_FMT_RGB32_1

  2729.                                              && dstFormat != PIX_FMT_BGR32_1

  2730.            && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE

  2731.            && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE

  2732.            && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2733.              c->swScale= rgb2rgbWrapper;

  2734. 

  2735.         if ((usePal(srcFormat) && (

  2736.                  dstFormat == PIX_FMT_RGB32   ||

  2737.                  dstFormat == PIX_FMT_RGB32_1 ||

  2738.                  dstFormat == PIX_FMT_RGB24   ||

  2739.                  dstFormat == PIX_FMT_BGR32   ||

  2740.                  dstFormat == PIX_FMT_BGR32_1 ||

  2741.                  dstFormat == PIX_FMT_BGR24)))

  2742.              c->swScale= pal2rgbWrapper;

  2743. 

  2744.         if (srcFormat == PIX_FMT_YUV422P)

  2745.         {

  2746.             if (dstFormat == PIX_FMT_YUYV422)

  2747.                 c->swScale= YUV422PToYuy2Wrapper;

  2748.             else if (dstFormat == PIX_FMT_UYVY422)

  2749.                 c->swScale= YUV422PToUyvyWrapper;

  2750.         }

  2751. 

  2752.         /* LQ converters if -sws 0 or -sws 4*/

  2753.         if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){

  2754.             /* yv12_to_yuy2 */

  2755.             if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P)

  2756.             {

  2757.                 if (dstFormat == PIX_FMT_YUYV422)

  2758.                     c->swScale= PlanarToYuy2Wrapper;

  2759.                 else if (dstFormat == PIX_FMT_UYVY422)

  2760.                     c->swScale= PlanarToUyvyWrapper;

  2761.             }

  2762.         }

  2763.         if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  2764.             c->swScale= YUYV2YUV420Wrapper;

  2765.         if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  2766.             c->swScale= UYVY2YUV420Wrapper;

  2767.         if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)

  2768.             c->swScale= YUYV2YUV422Wrapper;

  2769.         if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)

  2770.             c->swScale= UYVY2YUV422Wrapper;

  2771. 

  2772. #ifdef COMPILE_ALTIVEC

  2773.         if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  2774.             !(c->flags & SWS_BITEXACT) &&

  2775.             srcFormat == PIX_FMT_YUV420P) {

  2776.           // unscaled YV12 -> packed YUV, we want speed

  2777.           if (dstFormat == PIX_FMT_YUYV422)

  2778.               c->swScale= yv12toyuy2_unscaled_altivec;

  2779.           else if (dstFormat == PIX_FMT_UYVY422)

  2780.               c->swScale= yv12touyvy_unscaled_altivec;

  2781.         }

  2782. #endif

  2783. 

  2784.         /* simple copy */

  2785.         if (  srcFormat == dstFormat

  2786.             || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)

  2787.             || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)

  2788.             || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  2789.             || (isPlanarYUV(dstFormat) && isGray(srcFormat))

  2790.             || (isGray(dstFormat) && isGray(srcFormat))

  2791.             || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)

  2792.                 && c->chrDstHSubSample == c->chrSrcHSubSample

  2793.                 && c->chrDstVSubSample == c->chrSrcVSubSample))

  2794.         {

  2795.             if (isPacked(c->srcFormat))

  2796.                 c->swScale= packedCopy;

  2797.             else /* Planar YUV or gray */

  2798.                 c->swScale= planarCopy;

  2799.         }

  2800. #if ARCH_BFIN

  2801.         if (flags & SWS_CPU_CAPS_BFIN)

  2802.             ff_bfin_get_unscaled_swscale (c);

  2803. #endif

  2804. 

  2805.         if (c->swScale){

  2806.             if (flags&SWS_PRINT_INFO)

  2807.                 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",

  2808.                                 sws_format_name(srcFormat), sws_format_name(dstFormat));

  2809.             return c;

  2810.         }

  2811.     }

  2812. 

  2813.     if (flags & SWS_CPU_CAPS_MMX2)

  2814.     {

  2815.         c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;

  2816.         if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))

  2817.         {

  2818.             if (flags&SWS_PRINT_INFO)

  2819.                 av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");

  2820.         }

  2821.         if (usesHFilter) c->canMMX2BeUsed=0;

  2822.     }

  2823.     else

  2824.         c->canMMX2BeUsed=0;

  2825. 

  2826.     c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;

  2827.     c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

  2828. 

  2829.     // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst

  2830.     // but only for the FAST_BILINEAR mode otherwise do correct scaling

  2831.     // n-2 is the last chrominance sample available

  2832.     // this is not perfect, but no one should notice the difference, the more correct variant

  2833.     // would be like the vertical one, but that would require some special code for the

  2834.     // first and last pixel

  2835.     if (flags&SWS_FAST_BILINEAR)

  2836.     {

  2837.         if (c->canMMX2BeUsed)

  2838.         {

  2839.             c->lumXInc+= 20;

  2840.             c->chrXInc+= 20;

  2841.         }

  2842.         //we don't use the x86 asm scaler if MMX is available

  2843.         else if (flags & SWS_CPU_CAPS_MMX)

  2844.         {

  2845.             c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;

  2846.             c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;

  2847.         }

  2848.     }

  2849. 

  2850.     /* precalculate horizontal scaler filter coefficients */

  2851.     {

  2852.         const int filterAlign=

  2853.             (flags & SWS_CPU_CAPS_MMX) ? 4 :

  2854.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2855.             1;

  2856. 

  2857.         initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,

  2858.                    srcW      ,       dstW, filterAlign, 1<<14,

  2859.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2860.                    srcFilter->lumH, dstFilter->lumH, c->param);

  2861.         initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,

  2862.                    c->chrSrcW, c->chrDstW, filterAlign, 1<<14,

  2863.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2864.                    srcFilter->chrH, dstFilter->chrH, c->param);

  2865. 

  2866. #define MAX_FUNNY_CODE_SIZE 10000

  2867. #if defined(COMPILE_MMX2)

  2868. // can't downscale !!!

  2869.         if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))

  2870.         {

  2871. #ifdef MAP_ANONYMOUS

  2872.             c->funnyYCode  = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2873.             c->funnyUVCode = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2874. #elif HAVE_VIRTUALALLOC

  2875.             c->funnyYCode  = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);

  2876.             c->funnyUVCode = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);

  2877. #else

  2878.             c->funnyYCode  = av_malloc(MAX_FUNNY_CODE_SIZE);

  2879.             c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2880. #endif

  2881. 

  2882.             c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));

  2883.             c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));

  2884.             c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));

  2885.             c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));

  2886. 

  2887.             initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);

  2888.             initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);

  2889.         }

  2890. #endif /* defined(COMPILE_MMX2) */

  2891.     } // initialize horizontal stuff

  2892. 

  2893. 

  2894. 

  2895.     /* precalculate vertical scaler filter coefficients */

  2896.     {

  2897.         const int filterAlign=

  2898.             (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :

  2899.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2900.             1;

  2901. 

  2902.         initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,

  2903.                    srcH      ,        dstH, filterAlign, (1<<12),

  2904.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2905.                    srcFilter->lumV, dstFilter->lumV, c->param);

  2906.         initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,

  2907.                    c->chrSrcH, c->chrDstH, filterAlign, (1<<12),

  2908.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2909.                    srcFilter->chrV, dstFilter->chrV, c->param);

  2910. 

  2911. #if HAVE_ALTIVEC

  2912.         c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);

  2913.         c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);

  2914. 

  2915.         for (i=0;i<c->vLumFilterSize*c->dstH;i++) {

  2916.             int j;

  2917.             short *p = (short *)&c->vYCoeffsBank[i];

  2918.             for (j=0;j<8;j++)

  2919.                 p[j] = c->vLumFilter[i];

  2920.         }

  2921. 

  2922.         for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {

  2923.             int j;

  2924.             short *p = (short *)&c->vCCoeffsBank[i];

  2925.             for (j=0;j<8;j++)

  2926.                 p[j] = c->vChrFilter[i];

  2927.         }

  2928. #endif

  2929.     }

  2930. 

  2931.     // calculate buffer sizes so that they won't run out while handling these damn slices

  2932.     c->vLumBufSize= c->vLumFilterSize;

  2933.     c->vChrBufSize= c->vChrFilterSize;

  2934.     for (i=0; i<dstH; i++)

  2935.     {

  2936.         int chrI= i*c->chrDstH / dstH;

  2937.         int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,

  2938.                            ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

  2939. 

  2940.         nextSlice>>= c->chrSrcVSubSample;

  2941.         nextSlice<<= c->chrSrcVSubSample;

  2942.         if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)

  2943.             c->vLumBufSize= nextSlice - c->vLumFilterPos[i];

  2944.         if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))

  2945.             c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];

  2946.     }

  2947. 

  2948.     // allocate pixbufs (we use dynamic allocation because otherwise we would need to

  2949.     c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));

  2950.     c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));

  2951.     if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))

  2952.         c->alpPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));

  2953.     //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)

  2954.     /* align at 16 bytes for AltiVec */

  2955.     for (i=0; i<c->vLumBufSize; i++)

  2956.         c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);

  2957.     for (i=0; i<c->vChrBufSize; i++)

  2958.         c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);

  2959.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)

  2960.         for (i=0; i<c->vLumBufSize; i++)

  2961.             c->alpPixBuf[i]= c->alpPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);

  2962. 

  2963.     //try to avoid drawing green stuff between the right end and the stride end

  2964.     for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);

  2965. 

  2966.     assert(2*VOFW == VOF);

  2967. 

  2968.     assert(c->chrDstH <= dstH);

  2969. 

  2970.     if (flags&SWS_PRINT_INFO)

  2971.     {

  2972. #ifdef DITHER1XBPP

  2973.         const char *dither= " dithered";

  2974. #else

  2975.         const char *dither= "";

  2976. #endif

  2977.         if (flags&SWS_FAST_BILINEAR)

  2978.             av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");

  2979.         else if (flags&SWS_BILINEAR)

  2980.             av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");

  2981.         else if (flags&SWS_BICUBIC)

  2982.             av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");

  2983.         else if (flags&SWS_X)

  2984.             av_log(c, AV_LOG_INFO, "Experimental scaler, ");

  2985.         else if (flags&SWS_POINT)

  2986.             av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");

  2987.         else if (flags&SWS_AREA)

  2988.             av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");

  2989.         else if (flags&SWS_BICUBLIN)

  2990.             av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");

  2991.         else if (flags&SWS_GAUSS)

  2992.             av_log(c, AV_LOG_INFO, "Gaussian scaler, ");

  2993.         else if (flags&SWS_SINC)

  2994.             av_log(c, AV_LOG_INFO, "Sinc scaler, ");

  2995.         else if (flags&SWS_LANCZOS)

  2996.             av_log(c, AV_LOG_INFO, "Lanczos scaler, ");

  2997.         else if (flags&SWS_SPLINE)

  2998.             av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");

  2999.         else

  3000.             av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");

  3001. 

  3002.         if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)

  3003.             av_log(c, AV_LOG_INFO, "from %s to%s %s ",

  3004.                    sws_format_name(srcFormat), dither, sws_format_name(dstFormat));

  3005.         else

  3006.             av_log(c, AV_LOG_INFO, "from %s to %s ",

  3007.                    sws_format_name(srcFormat), sws_format_name(dstFormat));

  3008. 

  3009.         if (flags & SWS_CPU_CAPS_MMX2)

  3010.             av_log(c, AV_LOG_INFO, "using MMX2\n");

  3011.         else if (flags & SWS_CPU_CAPS_3DNOW)

  3012.             av_log(c, AV_LOG_INFO, "using 3DNOW\n");

  3013.         else if (flags & SWS_CPU_CAPS_MMX)

  3014.             av_log(c, AV_LOG_INFO, "using MMX\n");

  3015.         else if (flags & SWS_CPU_CAPS_ALTIVEC)

  3016.             av_log(c, AV_LOG_INFO, "using AltiVec\n");

  3017.         else

  3018.             av_log(c, AV_LOG_INFO, "using C\n");

  3019.     }

  3020. 

  3021.     if (flags & SWS_PRINT_INFO)

  3022.     {

  3023.         if (flags & SWS_CPU_CAPS_MMX)

  3024.         {

  3025.             if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))

  3026.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");

  3027.             else

  3028.             {

  3029.                 if (c->hLumFilterSize==4)

  3030.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");

  3031.                 else if (c->hLumFilterSize==8)

  3032.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");

  3033.                 else

  3034.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");

  3035. 

  3036.                 if (c->hChrFilterSize==4)

  3037.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");

  3038.                 else if (c->hChrFilterSize==8)

  3039.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");

  3040.                 else

  3041.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");

  3042.             }

  3043.         }

  3044.         else

  3045.         {

  3046. #if ARCH_X86

  3047.             av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");

  3048. #else

  3049.             if (flags & SWS_FAST_BILINEAR)

  3050.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");

  3051.             else

  3052.                 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");

  3053. #endif

  3054.         }

  3055.         if (isPlanarYUV(dstFormat))

  3056.         {

  3057.             if (c->vLumFilterSize==1)

  3058.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3059.             else

  3060.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3061.         }

  3062.         else

  3063.         {

  3064.             if (c->vLumFilterSize==1 && c->vChrFilterSize==2)

  3065.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"

  3066.                        "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3067.             else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)

  3068.                 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3069.             else

  3070.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3071.         }

  3072. 

  3073.         if (dstFormat==PIX_FMT_BGR24)

  3074.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",

  3075.                    (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));

  3076.         else if (dstFormat==PIX_FMT_RGB32)

  3077.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3078.         else if (dstFormat==PIX_FMT_BGR565)

  3079.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3080.         else if (dstFormat==PIX_FMT_BGR555)

  3081.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3082. 

  3083.         av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);

  3084.     }

  3085.     if (flags & SWS_PRINT_INFO)

  3086.     {

  3087.         av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  3088.                c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);

  3089.         av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  3090.                c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);

  3091.     }

  3092. 

  3093.     c->swScale= getSwsFunc(c);

  3094.     return c;

  3095. }

  3096. 

  3097. static void reset_ptr(uint8_t* src[], int format){

  3098.     if(!isALPHA(format))

  3099.         src[3]=NULL;

  3100.     if(!isPlanarYUV(format)){

  3101.         src[3]=src[2]=NULL;

  3102.         if(   format != PIX_FMT_PAL8

  3103.            && format != PIX_FMT_RGB8

  3104.            && format != PIX_FMT_BGR8

  3105.            && format != PIX_FMT_RGB4_BYTE

  3106.            && format != PIX_FMT_BGR4_BYTE

  3107.           )

  3108.             src[1]= NULL;

  3109.     }

  3110. }

  3111. 

  3112. /**

  3113.  * swscale wrapper, so we don't need to export the SwsContext.

  3114.  * Assumes planar YUV to be in YUV order instead of YVU.

  3115.  */

  3116. int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  3117.               int srcSliceH, uint8_t* dst[], int dstStride[]){

  3118.     int i;

  3119.     uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};

  3120.     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};

  3121. 

  3122.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  3123.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  3124.         return 0;

  3125.     }

  3126.     if (c->sliceDir == 0) {

  3127.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  3128.     }

  3129. 

  3130.     if (usePal(c->srcFormat)){

  3131.         for (i=0; i<256; i++){

  3132.             int p, r, g, b,y,u,v;

  3133.             if(c->srcFormat == PIX_FMT_PAL8){

  3134.                 p=((uint32_t*)(src[1]))[i];

  3135.                 r= (p>>16)&0xFF;

  3136.                 g= (p>> 8)&0xFF;

  3137.                 b=  p     &0xFF;

  3138.             }else if(c->srcFormat == PIX_FMT_RGB8){

  3139.                 r= (i>>5    )*36;

  3140.                 g= ((i>>2)&7)*36;

  3141.                 b= (i&3     )*85;

  3142.             }else if(c->srcFormat == PIX_FMT_BGR8){

  3143.                 b= (i>>6    )*85;

  3144.                 g= ((i>>3)&7)*36;

  3145.                 r= (i&7     )*36;

  3146.             }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){

  3147.                 r= (i>>3    )*255;

  3148.                 g= ((i>>1)&3)*85;

  3149.                 b= (i&1     )*255;

  3150.             }else {

  3151.                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);

  3152.                 b= (i>>3    )*255;

  3153.                 g= ((i>>1)&3)*85;

  3154.                 r= (i&1     )*255;

  3155.             }

  3156.             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3157.             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3158.             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3159.             c->pal_yuv[i]= y + (u<<8) + (v<<16);

  3160. 

  3161. 

  3162.             switch(c->dstFormat) {

  3163.             case PIX_FMT_BGR32:

  3164. #ifndef WORDS_BIGENDIAN

  3165.             case PIX_FMT_RGB24:

  3166. #endif

  3167.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);

  3168.                 break;

  3169.             case PIX_FMT_BGR32_1:

  3170. #ifdef  WORDS_BIGENDIAN

  3171.             case PIX_FMT_BGR24:

  3172. #endif

  3173.                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;

  3174.                 break;

  3175.             case PIX_FMT_RGB32_1:

  3176. #ifdef  WORDS_BIGENDIAN

  3177.             case PIX_FMT_RGB24:

  3178. #endif

  3179.                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;

  3180.                 break;

  3181.             case PIX_FMT_RGB32:

  3182. #ifndef WORDS_BIGENDIAN

  3183.             case PIX_FMT_BGR24:

  3184. #endif

  3185.             default:

  3186.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);

  3187.             }

  3188.         }

  3189.     }

  3190. 

  3191.     // copy strides, so they can safely be modified

  3192.     if (c->sliceDir == 1) {

  3193.         // slices go from top to bottom

  3194.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};

  3195.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};

  3196. 

  3197.         reset_ptr(src2, c->srcFormat);

  3198.         reset_ptr(dst2, c->dstFormat);

  3199. 

  3200.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);

  3201.     } else {

  3202.         // slices go from bottom to top => we flip the image internally

  3203.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};

  3204.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};

  3205. 

  3206.         src2[0] += (srcSliceH-1)*srcStride[0];

  3207.         if (!usePal(c->srcFormat))

  3208.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  3209.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  3210.         src2[3] += (srcSliceH-1)*srcStride[3];

  3211.         dst2[0] += ( c->dstH                      -1)*dstStride[0];

  3212.         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];

  3213.         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];

  3214.         dst2[3] += ( c->dstH                      -1)*dstStride[3];

  3215. 

  3216.         reset_ptr(src2, c->srcFormat);

  3217.         reset_ptr(dst2, c->dstFormat);

  3218. 

  3219.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  3220.     }

  3221. }

  3222. 

  3223. #if LIBSWSCALE_VERSION_MAJOR < 1

  3224. int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  3225.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  3226.     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);

  3227. }

  3228. #endif

  3229. 

  3230. SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,

  3231.                                 float lumaSharpen, float chromaSharpen,

  3232.                                 float chromaHShift, float chromaVShift,

  3233.                                 int verbose)

  3234. {

  3235.     SwsFilter *filter= av_malloc(sizeof(SwsFilter));

  3236. 

  3237.     if (lumaGBlur!=0.0){

  3238.         filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);

  3239.         filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);

  3240.     }else{

  3241.         filter->lumH= sws_getIdentityVec();

  3242.         filter->lumV= sws_getIdentityVec();

  3243.     }

  3244. 

  3245.     if (chromaGBlur!=0.0){

  3246.         filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);

  3247.         filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);

  3248.     }else{

  3249.         filter->chrH= sws_getIdentityVec();

  3250.         filter->chrV= sws_getIdentityVec();

  3251.     }

  3252. 

  3253.     if (chromaSharpen!=0.0){

  3254.         SwsVector *id= sws_getIdentityVec();

  3255.         sws_scaleVec(filter->chrH, -chromaSharpen);

  3256.         sws_scaleVec(filter->chrV, -chromaSharpen);

  3257.         sws_addVec(filter->chrH, id);

  3258.         sws_addVec(filter->chrV, id);

  3259.         sws_freeVec(id);

  3260.     }

  3261. 

  3262.     if (lumaSharpen!=0.0){

  3263.         SwsVector *id= sws_getIdentityVec();

  3264.         sws_scaleVec(filter->lumH, -lumaSharpen);

  3265.         sws_scaleVec(filter->lumV, -lumaSharpen);

  3266.         sws_addVec(filter->lumH, id);

  3267.         sws_addVec(filter->lumV, id);

  3268.         sws_freeVec(id);

  3269.     }

  3270. 

  3271.     if (chromaHShift != 0.0)

  3272.         sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));

  3273. 

  3274.     if (chromaVShift != 0.0)

  3275.         sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));

  3276. 

  3277.     sws_normalizeVec(filter->chrH, 1.0);

  3278.     sws_normalizeVec(filter->chrV, 1.0);

  3279.     sws_normalizeVec(filter->lumH, 1.0);

  3280.     sws_normalizeVec(filter->lumV, 1.0);

  3281. 

  3282.     if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);

  3283.     if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);

  3284. 

  3285.     return filter;

  3286. }

  3287. 

  3288. SwsVector *sws_getGaussianVec(double variance, double quality){

  3289.     const int length= (int)(variance*quality + 0.5) | 1;

  3290.     int i;

  3291.     double *coeff= av_malloc(length*sizeof(double));

  3292.     double middle= (length-1)*0.5;

  3293.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3294. 

  3295.     vec->coeff= coeff;

  3296.     vec->length= length;

  3297. 

  3298.     for (i=0; i<length; i++)

  3299.     {

  3300.         double dist= i-middle;

  3301.         coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);

  3302.     }

  3303. 

  3304.     sws_normalizeVec(vec, 1.0);

  3305. 

  3306.     return vec;

  3307. }

  3308. 

  3309. SwsVector *sws_getConstVec(double c, int length){

  3310.     int i;

  3311.     double *coeff= av_malloc(length*sizeof(double));

  3312.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3313. 

  3314.     vec->coeff= coeff;

  3315.     vec->length= length;

  3316. 

  3317.     for (i=0; i<length; i++)

  3318.         coeff[i]= c;

  3319. 

  3320.     return vec;

  3321. }

  3322. 

  3323. 

  3324. SwsVector *sws_getIdentityVec(void){

  3325.     return sws_getConstVec(1.0, 1);

  3326. }

  3327. 

  3328. double sws_dcVec(SwsVector *a){

  3329.     int i;

  3330.     double sum=0;

  3331. 

  3332.     for (i=0; i<a->length; i++)

  3333.         sum+= a->coeff[i];

  3334. 

  3335.     return sum;

  3336. }

  3337. 

  3338. void sws_scaleVec(SwsVector *a, double scalar){

  3339.     int i;

  3340. 

  3341.     for (i=0; i<a->length; i++)

  3342.         a->coeff[i]*= scalar;

  3343. }

  3344. 

  3345. void sws_normalizeVec(SwsVector *a, double height){

  3346.     sws_scaleVec(a, height/sws_dcVec(a));

  3347. }

  3348. 

  3349. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){

  3350.     int length= a->length + b->length - 1;

  3351.     double *coeff= av_malloc(length*sizeof(double));

  3352.     int i, j;

  3353.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3354. 

  3355.     vec->coeff= coeff;

  3356.     vec->length= length;

  3357. 

  3358.     for (i=0; i<length; i++) coeff[i]= 0.0;

  3359. 

  3360.     for (i=0; i<a->length; i++)

  3361.     {

  3362.         for (j=0; j<b->length; j++)

  3363.         {

  3364.             coeff[i+j]+= a->coeff[i]*b->coeff[j];

  3365.         }

  3366.     }

  3367. 

  3368.     return vec;

  3369. }

  3370. 

  3371. static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){

  3372.     int length= FFMAX(a->length, b->length);

  3373.     double *coeff= av_malloc(length*sizeof(double));

  3374.     int i;

  3375.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3376. 

  3377.     vec->coeff= coeff;

  3378.     vec->length= length;

  3379. 

  3380.     for (i=0; i<length; i++) coeff[i]= 0.0;

  3381. 

  3382.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  3383.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];

  3384. 

  3385.     return vec;

  3386. }

  3387. 

  3388. static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){

  3389.     int length= FFMAX(a->length, b->length);

  3390.     double *coeff= av_malloc(length*sizeof(double));

  3391.     int i;

  3392.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3393. 

  3394.     vec->coeff= coeff;

  3395.     vec->length= length;

  3396. 

  3397.     for (i=0; i<length; i++) coeff[i]= 0.0;

  3398. 

  3399.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  3400.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];

  3401. 

  3402.     return vec;

  3403. }

  3404. 

  3405. /* shift left / or right if "shift" is negative */

  3406. static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){

  3407.     int length= a->length + FFABS(shift)*2;

  3408.     double *coeff= av_malloc(length*sizeof(double));

  3409.     int i;

  3410.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3411. 

  3412.     vec->coeff= coeff;

  3413.     vec->length= length;

  3414. 

  3415.     for (i=0; i<length; i++) coeff[i]= 0.0;

  3416. 

  3417.     for (i=0; i<a->length; i++)

  3418.     {

  3419.         coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];

  3420.     }

  3421. 

  3422.     return vec;

  3423. }

  3424. 

  3425. void sws_shiftVec(SwsVector *a, int shift){

  3426.     SwsVector *shifted= sws_getShiftedVec(a, shift);

  3427.     av_free(a->coeff);

  3428.     a->coeff= shifted->coeff;

  3429.     a->length= shifted->length;

  3430.     av_free(shifted);

  3431. }

  3432. 

  3433. void sws_addVec(SwsVector *a, SwsVector *b){

  3434.     SwsVector *sum= sws_sumVec(a, b);

  3435.     av_free(a->coeff);

  3436.     a->coeff= sum->coeff;

  3437.     a->length= sum->length;

  3438.     av_free(sum);

  3439. }

  3440. 

  3441. void sws_subVec(SwsVector *a, SwsVector *b){

  3442.     SwsVector *diff= sws_diffVec(a, b);

  3443.     av_free(a->coeff);

  3444.     a->coeff= diff->coeff;

  3445.     a->length= diff->length;

  3446.     av_free(diff);

  3447. }

  3448. 

  3449. void sws_convVec(SwsVector *a, SwsVector *b){

  3450.     SwsVector *conv= sws_getConvVec(a, b);

  3451.     av_free(a->coeff);

  3452.     a->coeff= conv->coeff;

  3453.     a->length= conv->length;

  3454.     av_free(conv);

  3455. }

  3456. 

  3457. SwsVector *sws_cloneVec(SwsVector *a){

  3458.     double *coeff= av_malloc(a->length*sizeof(double));

  3459.     int i;

  3460.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3461. 

  3462.     vec->coeff= coeff;

  3463.     vec->length= a->length;

  3464. 

  3465.     for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];

  3466. 

  3467.     return vec;

  3468. }

  3469. 

  3470. void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level){

  3471.     int i;

  3472.     double max=0;

  3473.     double min=0;

  3474.     double range;

  3475. 

  3476.     for (i=0; i<a->length; i++)

  3477.         if (a->coeff[i]>max) max= a->coeff[i];

  3478. 

  3479.     for (i=0; i<a->length; i++)

  3480.         if (a->coeff[i]<min) min= a->coeff[i];

  3481. 

  3482.     range= max - min;

  3483. 

  3484.     for (i=0; i<a->length; i++)

  3485.     {

  3486.         int x= (int)((a->coeff[i]-min)*60.0/range +0.5);

  3487.         av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);

  3488.         for (;x>0; x--) av_log(log_ctx, log_level, " ");

  3489.         av_log(log_ctx, log_level, "|\n");

  3490.     }

  3491. }

  3492. 

  3493. #if LIBSWSCALE_VERSION_MAJOR < 1

  3494. void sws_printVec(SwsVector *a){

  3495.     sws_printVec2(a, NULL, AV_LOG_DEBUG);

  3496. }

  3497. #endif

  3498. 

  3499. void sws_freeVec(SwsVector *a){

  3500.     if (!a) return;

  3501.     av_freep(&a->coeff);

  3502.     a->length=0;

  3503.     av_free(a);

  3504. }

  3505. 

  3506. void sws_freeFilter(SwsFilter *filter){

  3507.     if (!filter) return;

  3508. 

  3509.     if (filter->lumH) sws_freeVec(filter->lumH);

  3510.     if (filter->lumV) sws_freeVec(filter->lumV);

  3511.     if (filter->chrH) sws_freeVec(filter->chrH);

  3512.     if (filter->chrV) sws_freeVec(filter->chrV);

  3513.     av_free(filter);

  3514. }

  3515. 

  3516. 

  3517. void sws_freeContext(SwsContext *c){

  3518.     int i;

  3519.     if (!c) return;

  3520. 

  3521.     if (c->lumPixBuf)

  3522.     {

  3523.         for (i=0; i<c->vLumBufSize; i++)

  3524.             av_freep(&c->lumPixBuf[i]);

  3525.         av_freep(&c->lumPixBuf);

  3526.     }

  3527. 

  3528.     if (c->chrPixBuf)

  3529.     {

  3530.         for (i=0; i<c->vChrBufSize; i++)

  3531.             av_freep(&c->chrPixBuf[i]);

  3532.         av_freep(&c->chrPixBuf);

  3533.     }

  3534. 

  3535.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){

  3536.         for (i=0; i<c->vLumBufSize; i++)

  3537.             av_freep(&c->alpPixBuf[i]);

  3538.         av_freep(&c->alpPixBuf);

  3539.     }

  3540. 

  3541.     av_freep(&c->vLumFilter);

  3542.     av_freep(&c->vChrFilter);

  3543.     av_freep(&c->hLumFilter);

  3544.     av_freep(&c->hChrFilter);

  3545. #if HAVE_ALTIVEC

  3546.     av_freep(&c->vYCoeffsBank);

  3547.     av_freep(&c->vCCoeffsBank);

  3548. #endif

  3549. 

  3550.     av_freep(&c->vLumFilterPos);

  3551.     av_freep(&c->vChrFilterPos);

  3552.     av_freep(&c->hLumFilterPos);

  3553.     av_freep(&c->hChrFilterPos);

  3554. 

  3555. #if ARCH_X86 && CONFIG_GPL

  3556. #ifdef MAP_ANONYMOUS

  3557.     if (c->funnyYCode ) munmap(c->funnyYCode , MAX_FUNNY_CODE_SIZE);

  3558.     if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);

  3559. #elif HAVE_VIRTUALALLOC

  3560.     if (c->funnyYCode ) VirtualFree(c->funnyYCode , MAX_FUNNY_CODE_SIZE, MEM_RELEASE);

  3561.     if (c->funnyUVCode) VirtualFree(c->funnyUVCode, MAX_FUNNY_CODE_SIZE, MEM_RELEASE);

  3562. #else

  3563.     av_free(c->funnyYCode );

  3564.     av_free(c->funnyUVCode);

  3565. #endif

  3566.     c->funnyYCode=NULL;

  3567.     c->funnyUVCode=NULL;

  3568. #endif /* ARCH_X86 && CONFIG_GPL */

  3569. 

  3570.     av_freep(&c->lumMmx2Filter);

  3571.     av_freep(&c->chrMmx2Filter);

  3572.     av_freep(&c->lumMmx2FilterPos);

  3573.     av_freep(&c->chrMmx2FilterPos);

  3574.     av_freep(&c->yuvTable);

  3575. 

  3576.     av_free(c);

  3577. }

  3578. 

  3579. struct SwsContext *sws_getCachedContext(struct SwsContext *context,

  3580.                                         int srcW, int srcH, enum PixelFormat srcFormat,

  3581.                                         int dstW, int dstH, enum PixelFormat dstFormat, int flags,

  3582.                                         SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)

  3583. {

  3584.     static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};

  3585. 

  3586.     if (!param)

  3587.         param = default_param;

  3588. 

  3589.     if (context) {

  3590.         if (context->srcW != srcW || context->srcH != srcH ||

  3591.             context->srcFormat != srcFormat ||

  3592.             context->dstW != dstW || context->dstH != dstH ||

  3593.             context->dstFormat != dstFormat || context->flags != flags ||

  3594.             context->param[0] != param[0] || context->param[1] != param[1])

  3595.         {

  3596.             sws_freeContext(context);

  3597.             context = NULL;

  3598.         }

  3599.     }

  3600.     if (!context) {

  3601.         return sws_getContext(srcW, srcH, srcFormat,

  3602.                               dstW, dstH, dstFormat, flags,

  3603.                               srcFilter, dstFilter, param);

  3604.     }

  3605.     return context;

  3606. }