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.
22060 Commits
32 Branches
1.1GB
Tree: 08f8b51f69
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '08f8b51f69'
${ noResults }
FFmpeg/libswscale/swscale.c

3480 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/intreadwrite.h"

  78. #include "libavutil/x86_cpu.h"

  79. #include "libavutil/avutil.h"

  80. #include "libavutil/bswap.h"

  81. #include "libavutil/pixdesc.h"

  82. 

  83. unsigned swscale_version(void)

  84. {

  85.     return LIBSWSCALE_VERSION_INT;

  86. }

  87. 

  88. const char *swscale_configuration(void)

  89. {

  90.     return FFMPEG_CONFIGURATION;

  91. }

  92. 

  93. const char *swscale_license(void)

  94. {

  95. #define LICENSE_PREFIX "libswscale license: "

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

  97. }

  98. 

  99. #undef MOVNTQ

  100. #undef PAVGB

  101. 

  102. //#undef HAVE_MMX2

  103. //#define HAVE_AMD3DNOW

  104. //#undef HAVE_MMX

  105. //#undef ARCH_X86

  106. #define DITHER1XBPP

  107. 

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

  109. 

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

  111. 

  112. #ifdef M_PI

  113. #define PI M_PI

  114. #else

  115. #define PI 3.14159265358979323846

  116. #endif

  117. 

  118. #define isSupportedIn(x)    (       \

  119.            (x)==PIX_FMT_YUV420P     \

  120.         || (x)==PIX_FMT_YUVA420P    \

  121.         || (x)==PIX_FMT_YUYV422     \

  122.         || (x)==PIX_FMT_UYVY422     \

  123.         || (x)==PIX_FMT_RGB48BE     \

  124.         || (x)==PIX_FMT_RGB48LE     \

  125.         || (x)==PIX_FMT_RGB32       \

  126.         || (x)==PIX_FMT_RGB32_1     \

  127.         || (x)==PIX_FMT_BGR24       \

  128.         || (x)==PIX_FMT_BGR565      \

  129.         || (x)==PIX_FMT_BGR555      \

  130.         || (x)==PIX_FMT_BGR32       \

  131.         || (x)==PIX_FMT_BGR32_1     \

  132.         || (x)==PIX_FMT_RGB24       \

  133.         || (x)==PIX_FMT_RGB565      \

  134.         || (x)==PIX_FMT_RGB555      \

  135.         || (x)==PIX_FMT_GRAY8       \

  136.         || (x)==PIX_FMT_YUV410P     \

  137.         || (x)==PIX_FMT_YUV440P     \

  138.         || (x)==PIX_FMT_NV12        \

  139.         || (x)==PIX_FMT_NV21        \

  140.         || (x)==PIX_FMT_GRAY16BE    \

  141.         || (x)==PIX_FMT_GRAY16LE    \

  142.         || (x)==PIX_FMT_YUV444P     \

  143.         || (x)==PIX_FMT_YUV422P     \

  144.         || (x)==PIX_FMT_YUV411P     \

  145.         || (x)==PIX_FMT_PAL8        \

  146.         || (x)==PIX_FMT_BGR8        \

  147.         || (x)==PIX_FMT_RGB8        \

  148.         || (x)==PIX_FMT_BGR4_BYTE   \

  149.         || (x)==PIX_FMT_RGB4_BYTE   \

  150.         || (x)==PIX_FMT_YUV440P     \

  151.         || (x)==PIX_FMT_MONOWHITE   \

  152.         || (x)==PIX_FMT_MONOBLACK   \

  153.         || (x)==PIX_FMT_YUV420P16LE   \

  154.         || (x)==PIX_FMT_YUV422P16LE   \

  155.         || (x)==PIX_FMT_YUV444P16LE   \

  156.         || (x)==PIX_FMT_YUV420P16BE   \

  157.         || (x)==PIX_FMT_YUV422P16BE   \

  158.         || (x)==PIX_FMT_YUV444P16BE   \

  159.     )

  160. 

  161. int sws_isSupportedInput(enum PixelFormat pix_fmt)

  162. {

  163.     return isSupportedIn(pix_fmt);

  164. }

  165. 

  166. #define isSupportedOut(x)   (       \

  167.            (x)==PIX_FMT_YUV420P     \

  168.         || (x)==PIX_FMT_YUVA420P    \

  169.         || (x)==PIX_FMT_YUYV422     \

  170.         || (x)==PIX_FMT_UYVY422     \

  171.         || (x)==PIX_FMT_YUV444P     \

  172.         || (x)==PIX_FMT_YUV422P     \

  173.         || (x)==PIX_FMT_YUV411P     \

  174.         || isRGB(x)                 \

  175.         || isBGR(x)                 \

  176.         || (x)==PIX_FMT_NV12        \

  177.         || (x)==PIX_FMT_NV21        \

  178.         || (x)==PIX_FMT_GRAY16BE    \

  179.         || (x)==PIX_FMT_GRAY16LE    \

  180.         || (x)==PIX_FMT_GRAY8       \

  181.         || (x)==PIX_FMT_YUV410P     \

  182.         || (x)==PIX_FMT_YUV440P     \

  183.         || (x)==PIX_FMT_YUV420P16LE   \

  184.         || (x)==PIX_FMT_YUV422P16LE   \

  185.         || (x)==PIX_FMT_YUV444P16LE   \

  186.         || (x)==PIX_FMT_YUV420P16BE   \

  187.         || (x)==PIX_FMT_YUV422P16BE   \

  188.         || (x)==PIX_FMT_YUV444P16BE   \

  189.     )

  190. 

  191. int sws_isSupportedOutput(enum PixelFormat pix_fmt)

  192. {

  193.     return isSupportedOut(pix_fmt);

  194. }

  195. 

  196. #define isPacked(x)         (       \

  197.            (x)==PIX_FMT_PAL8        \

  198.         || (x)==PIX_FMT_YUYV422     \

  199.         || (x)==PIX_FMT_UYVY422     \

  200.         || isRGB(x)                 \

  201.         || isBGR(x)                 \

  202.     )

  203. #define usePal(x) (av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL)

  204. 

  205. #define RGB2YUV_SHIFT 15

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

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

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

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

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

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

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

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

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

  215. 

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

  217. 

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

  219.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},

  220.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},

  221.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  222.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  223.     {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC

  224.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  225.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M

  226.     {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M

  227. };

  228. 

  229. /*

  230. NOTES

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

  232. 

  233. TODO

  234. more intelligent misalignment avoidance for the horizontal scaler

  235. write special vertical cubic upscale version

  236. optimize C code (YV12 / minmax)

  237. add support for packed pixel YUV input & output

  238. add support for Y8 output

  239. optimize BGR24 & BGR32

  240. add BGR4 output support

  241. write special BGR->BGR scaler

  242. */

  243. 

  244. #if ARCH_X86 && CONFIG_GPL

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

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

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

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

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

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

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

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

  253. 

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

  255.         0x0103010301030103LL,

  256.         0x0200020002000200LL,};

  257. 

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

  259.         0x0602060206020602LL,

  260.         0x0004000400040004LL,};

  261. 

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

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

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

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

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

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

  268. 

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

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

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

  272. 

  273. #ifdef FAST_BGR2YV12

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

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

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

  277. #else

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

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

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

  281. #endif /* FAST_BGR2YV12 */

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

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

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

  285. 

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

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

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

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

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

  291. 

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

  293.     {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},

  294.     {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},

  295. };

  296. 

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

  298. 

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

  300. 

  301. // clipping helper table for C implementations:

  302. static unsigned char clip_table[768];

  303. 

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

  305. 

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

  307. {  1,   3,   1,   3,   1,   3,   1,   3, },

  308. {  2,   0,   2,   0,   2,   0,   2,   0, },

  309. };

  310. 

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

  312. {  6,   2,   6,   2,   6,   2,   6,   2, },

  313. {  0,   4,   0,   4,   0,   4,   0,   4, },

  314. };

  315. 

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

  317. { 17,   9,  23,  15,  16,   8,  22,  14, },

  318. {  5,  29,   3,  27,   4,  28,   2,  26, },

  319. { 21,  13,  19,  11,  20,  12,  18,  10, },

  320. {  0,  24,   6,  30,   1,  25,   7,  31, },

  321. { 16,   8,  22,  14,  17,   9,  23,  15, },

  322. {  4,  28,   2,  26,   5,  29,   3,  27, },

  323. { 20,  12,  18,  10,  21,  13,  19,  11, },

  324. {  1,  25,   7,  31,   0,  24,   6,  30, },

  325. };

  326. 

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

  328. {  0,  55,  14,  68,   3,  58,  17,  72, },

  329. { 37,  18,  50,  32,  40,  22,  54,  35, },

  330. {  9,  64,   5,  59,  13,  67,   8,  63, },

  331. { 46,  27,  41,  23,  49,  31,  44,  26, },

  332. {  2,  57,  16,  71,   1,  56,  15,  70, },

  333. { 39,  21,  52,  34,  38,  19,  51,  33, },

  334. { 11,  66,   7,  62,  10,  65,   6,  60, },

  335. { 48,  30,  43,  25,  47,  29,  42,  24, },

  336. };

  337. 

  338. #if 1

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

  340. {117,  62, 158, 103, 113,  58, 155, 100, },

  341. { 34, 199,  21, 186,  31, 196,  17, 182, },

  342. {144,  89, 131,  76, 141,  86, 127,  72, },

  343. {  0, 165,  41, 206,  10, 175,  52, 217, },

  344. {110,  55, 151,  96, 120,  65, 162, 107, },

  345. { 28, 193,  14, 179,  38, 203,  24, 189, },

  346. {138,  83, 124,  69, 148,  93, 134,  79, },

  347. {  7, 172,  48, 213,   3, 168,  45, 210, },

  348. };

  349. #elif 1

  350. // tries to correct a gamma of 1.5

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

  352. {  0, 143,  18, 200,   2, 156,  25, 215, },

  353. { 78,  28, 125,  64,  89,  36, 138,  74, },

  354. { 10, 180,   3, 161,  16, 195,   8, 175, },

  355. {109,  51,  93,  38, 121,  60, 105,  47, },

  356. {  1, 152,  23, 210,   0, 147,  20, 205, },

  357. { 85,  33, 134,  71,  81,  30, 130,  67, },

  358. { 14, 190,   6, 171,  12, 185,   5, 166, },

  359. {117,  57, 101,  44, 113,  54,  97,  41, },

  360. };

  361. #elif 1

  362. // tries to correct a gamma of 2.0

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

  364. {  0, 124,   8, 193,   0, 140,  12, 213, },

  365. { 55,  14, 104,  42,  66,  19, 119,  52, },

  366. {  3, 168,   1, 145,   6, 187,   3, 162, },

  367. { 86,  31,  70,  21,  99,  39,  82,  28, },

  368. {  0, 134,  11, 206,   0, 129,   9, 200, },

  369. { 62,  17, 114,  48,  58,  16, 109,  45, },

  370. {  5, 181,   2, 157,   4, 175,   1, 151, },

  371. { 95,  36,  78,  26,  90,  34,  74,  24, },

  372. };

  373. #else

  374. // tries to correct a gamma of 2.5

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

  376. {  0, 107,   3, 187,   0, 125,   6, 212, },

  377. { 39,   7,  86,  28,  49,  11, 102,  36, },

  378. {  1, 158,   0, 131,   3, 180,   1, 151, },

  379. { 68,  19,  52,  12,  81,  25,  64,  17, },

  380. {  0, 119,   5, 203,   0, 113,   4, 195, },

  381. { 45,   9,  96,  33,  42,   8,  91,  30, },

  382. {  2, 172,   1, 144,   2, 165,   0, 137, },

  383. { 77,  23,  60,  15,  72,  21,  56,  14, },

  384. };

  385. #endif

  386. 

  387. const char *sws_format_name(enum PixelFormat format)

  388. {

  389.     if ((unsigned)format < PIX_FMT_NB && av_pix_fmt_descriptors[format].name)

  390.         return av_pix_fmt_descriptors[format].name;

  391.     else

  392.         return "Unknown format";

  393. }

  394. 

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

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

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

  398.                                                     int dstW, int chrDstW, int big_endian)

  399. {

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

  401.     int i;

  402. 

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

  404.         int val = 1 << 10;

  405.         int j;

  406. 

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

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

  409. 

  410.         if (big_endian) {

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

  412.         } else {

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

  414.         }

  415.     }

  416. 

  417.     if (uDest) {

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

  419.             int u = 1 << 10;

  420.             int v = 1 << 10;

  421.             int j;

  422. 

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

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

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

  426.             }

  427. 

  428.             if (big_endian) {

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

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

  431.             } else {

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

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

  434.             }

  435.         }

  436.     }

  437. 

  438.     if (CONFIG_SWSCALE_ALPHA && aDest) {

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

  440.             int val = 1 << 10;

  441.             int j;

  442. 

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

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

  445. 

  446.             if (big_endian) {

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

  448.             } else {

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

  450.             }

  451.         }

  452.     }

  453. }

  454. 

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

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

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

  458.                                  enum PixelFormat dstFormat)

  459. {

  460.     if (isBE(dstFormat)) {

  461.         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,

  462.                                chrFilter, chrSrc, chrFilterSize,

  463.                                alpSrc,

  464.                                dest, uDest, vDest, aDest,

  465.                                dstW, chrDstW, 1);

  466.     } else {

  467.         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,

  468.                                chrFilter, chrSrc, chrFilterSize,

  469.                                alpSrc,

  470.                                dest, uDest, vDest, aDest,

  471.                                dstW, chrDstW, 0);

  472.     }

  473. }

  474. 

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

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

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

  478. {

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

  480.     int i;

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

  482.         int val=1<<18;

  483.         int j;

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

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

  486. 

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

  488.     }

  489. 

  490.     if (uDest)

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

  492.             int u=1<<18;

  493.             int v=1<<18;

  494.             int j;

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

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

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

  498.             }

  499. 

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

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

  502.         }

  503. 

  504.     if (CONFIG_SWSCALE_ALPHA && aDest)

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

  506.             int val=1<<18;

  507.             int j;

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

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

  510. 

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

  512.         }

  513. 

  514. }

  515. 

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

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

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

  519. {

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

  521.     int i;

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

  523.         int val=1<<18;

  524.         int j;

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

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

  527. 

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

  529.     }

  530. 

  531.     if (!uDest)

  532.         return;

  533. 

  534.     if (dstFormat == PIX_FMT_NV12)

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

  536.             int u=1<<18;

  537.             int v=1<<18;

  538.             int j;

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

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

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

  542.             }

  543. 

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

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

  546.         }

  547.     else

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

  549.             int u=1<<18;

  550.             int v=1<<18;

  551.             int j;

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

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

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

  555.             }

  556. 

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

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

  559.         }

  560. }

  561. 

  562. #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \

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

  564.         int j;\

  565.         int Y1 = 1<<18;\

  566.         int Y2 = 1<<18;\

  567.         int U  = 1<<18;\

  568.         int V  = 1<<18;\

  569.         int av_unused A1, A2;\

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

  571.         const int i2= 2*i;\

  572.         \

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

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

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

  576.         }\

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

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

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

  580.         }\

  581.         Y1>>=19;\

  582.         Y2>>=19;\

  583.         U >>=19;\

  584.         V >>=19;\

  585.         if (alpha) {\

  586.             A1 = 1<<18;\

  587.             A2 = 1<<18;\

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

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

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

  591.             }\

  592.             A1>>=19;\

  593.             A2>>=19;\

  594.         }\

  595. 

  596. #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \

  597.         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\

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

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

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

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

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

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

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

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

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

  607.         }\

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

  609.             A1=av_clip_uint8(A1);\

  610.             A2=av_clip_uint8(A2);\

  611.         }

  612. 

  613. #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \

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

  615.         int j;\

  616.         int Y = 0;\

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

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

  619.         int av_unused A;\

  620.         int R,G,B;\

  621.         \

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

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

  624.         }\

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

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

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

  628.         }\

  629.         Y >>=10;\

  630.         U >>=10;\

  631.         V >>=10;\

  632.         if (alpha) {\

  633.             A = rnd;\

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

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

  636.             A >>=19;\

  637.             if (A&256)\

  638.                 A = av_clip_uint8(A);\

  639.         }\

  640. 

  641. #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \

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

  643.         Y-= c->yuv2rgb_y_offset;\

  644.         Y*= c->yuv2rgb_y_coeff;\

  645.         Y+= rnd;\

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

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

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

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

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

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

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

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

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

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

  656.         }\

  657. 

  658. 

  659. #define YSCALE_YUV_2_GRAY16_C \

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

  661.         int j;\

  662.         int Y1 = 1<<18;\

  663.         int Y2 = 1<<18;\

  664.         int U  = 1<<18;\

  665.         int V  = 1<<18;\

  666.         \

  667.         const int i2= 2*i;\

  668.         \

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

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

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

  672.         }\

  673.         Y1>>=11;\

  674.         Y2>>=11;\

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

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

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

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

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

  680.         }

  681. 

  682. #define YSCALE_YUV_2_RGBX_C(type,alpha) \

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

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

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

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

  687. 

  688. #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \

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

  690.         const int i2= 2*i;       \

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

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

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

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

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

  696.         int av_unused A1, A2;                                         \

  697.         if (alpha) {\

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

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

  700.         }\

  701. 

  702. #define YSCALE_YUV_2_GRAY16_2_C   \

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

  704.         const int i2= 2*i;       \

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

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

  707. 

  708. #define YSCALE_YUV_2_RGB2_C(type,alpha) \

  709.     YSCALE_YUV_2_PACKED2_C(type,alpha)\

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

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

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

  713. 

  714. #define YSCALE_YUV_2_PACKED1_C(type,alpha) \

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

  716.         const int i2= 2*i;\

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

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

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

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

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

  722.         int av_unused A1, A2;\

  723.         if (alpha) {\

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

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

  726.         }\

  727. 

  728. #define YSCALE_YUV_2_GRAY16_1_C \

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

  730.         const int i2= 2*i;\

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

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

  733. 

  734. #define YSCALE_YUV_2_RGB1_C(type,alpha) \

  735.     YSCALE_YUV_2_PACKED1_C(type,alpha)\

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

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

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

  739. 

  740. #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \

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

  742.         const int i2= 2*i;\

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

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

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

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

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

  748.         int av_unused A1, A2;\

  749.         if (alpha) {\

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

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

  752.         }\

  753. 

  754. #define YSCALE_YUV_2_RGB1B_C(type,alpha) \

  755.     YSCALE_YUV_2_PACKED1B_C(type,alpha)\

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

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

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

  759. 

  760. #define YSCALE_YUV_2_MONO2_C \

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

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

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

  764.         int acc;\

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

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

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

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

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

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

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

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

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

  774.         dest++;\

  775.     }\

  776. 

  777. 

  778. #define YSCALE_YUV_2_MONOX_C \

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

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

  781.     int acc=0;\

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

  783.         int j;\

  784.         int Y1=1<<18;\

  785.         int Y2=1<<18;\

  786. \

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

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

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

  790.         }\

  791.         Y1>>=19;\

  792.         Y2>>=19;\

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

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

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

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

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

  798.         }\

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

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

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

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

  803.             dest++;\

  804.         }\

  805.     }

  806. 

  807. 

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

  809.     switch(c->dstFormat) {\

  810.     case PIX_FMT_RGB48BE:\

  811.     case PIX_FMT_RGB48LE:\

  812.         func(uint8_t,0)\

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

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

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

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

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

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

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

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

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

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

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

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

  825.             dest+=12;\

  826.         }\

  827.         break;\

  828.     case PIX_FMT_RGBA:\

  829.     case PIX_FMT_BGRA:\

  830.         if (CONFIG_SMALL) {\

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

  832.             func(uint32_t,needAlpha)\

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

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

  835.             }\

  836.         } else {\

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

  838.                 func(uint32_t,1)\

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

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

  841.                 }\

  842.             } else {\

  843.                 func(uint32_t,0)\

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

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

  846.                 }\

  847.             }\

  848.         }\

  849.         break;\

  850.     case PIX_FMT_ARGB:\

  851.     case PIX_FMT_ABGR:\

  852.         if (CONFIG_SMALL) {\

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

  854.             func(uint32_t,needAlpha)\

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

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

  857.             }\

  858.         } else {\

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

  860.                 func(uint32_t,1)\

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

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

  863.                 }\

  864.             } else {\

  865.                 func(uint32_t,0)\

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

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

  868.                 }\

  869.             }\

  870.         }                \

  871.         break;\

  872.     case PIX_FMT_RGB24:\

  873.         func(uint8_t,0)\

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

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

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

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

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

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

  880.             dest+=6;\

  881.         }\

  882.         break;\

  883.     case PIX_FMT_BGR24:\

  884.         func(uint8_t,0)\

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

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

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

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

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

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

  891.             dest+=6;\

  892.         }\

  893.         break;\

  894.     case PIX_FMT_RGB565:\

  895.     case PIX_FMT_BGR565:\

  896.         {\

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

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

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

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

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

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

  903.             func(uint16_t,0)\

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

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

  906.             }\

  907.         }\

  908.         break;\

  909.     case PIX_FMT_RGB555:\

  910.     case PIX_FMT_BGR555:\

  911.         {\

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

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

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

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

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

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

  918.             func(uint16_t,0)\

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

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

  921.             }\

  922.         }\

  923.         break;\

  924.     case PIX_FMT_RGB8:\

  925.     case PIX_FMT_BGR8:\

  926.         {\

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

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

  929.             func(uint8_t,0)\

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

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

  932.             }\

  933.         }\

  934.         break;\

  935.     case PIX_FMT_RGB4:\

  936.     case PIX_FMT_BGR4:\

  937.         {\

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

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

  940.             func(uint8_t,0)\

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

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

  943.             }\

  944.         }\

  945.         break;\

  946.     case PIX_FMT_RGB4_BYTE:\

  947.     case PIX_FMT_BGR4_BYTE:\

  948.         {\

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

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

  951.             func(uint8_t,0)\

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

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

  954.             }\

  955.         }\

  956.         break;\

  957.     case PIX_FMT_MONOBLACK:\

  958.     case PIX_FMT_MONOWHITE:\

  959.         {\

  960.             func_monoblack\

  961.         }\

  962.         break;\

  963.     case PIX_FMT_YUYV422:\

  964.         func2\

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

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

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

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

  969.         }                \

  970.         break;\

  971.     case PIX_FMT_UYVY422:\

  972.         func2\

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

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

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

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

  977.         }                \

  978.         break;\

  979.     case PIX_FMT_GRAY16BE:\

  980.         func_g16\

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

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

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

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

  985.         }                \

  986.         break;\

  987.     case PIX_FMT_GRAY16LE:\

  988.         func_g16\

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

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

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

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

  993.         }                \

  994.         break;\

  995.     }\

  996. 

  997. 

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

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

  1000.                                   const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  1001. {

  1002.     int i;

  1003.     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)

  1004. }

  1005. 

  1006. static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  1007.                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  1008.                                     const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  1009. {

  1010.     int i;

  1011.     int step= fmt_depth(c->dstFormat)/8;

  1012.     int aidx= 3;

  1013. 

  1014.     switch(c->dstFormat) {

  1015.     case PIX_FMT_ARGB:

  1016.         dest++;

  1017.         aidx= 0;

  1018.     case PIX_FMT_RGB24:

  1019.         aidx--;

  1020.     case PIX_FMT_RGBA:

  1021.         if (CONFIG_SMALL) {

  1022.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  1023.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  1024.                 dest[aidx]= needAlpha ? A : 255;

  1025.                 dest[0]= R>>22;

  1026.                 dest[1]= G>>22;

  1027.                 dest[2]= B>>22;

  1028.                 dest+= step;

  1029.             }

  1030.         } else {

  1031.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1032.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  1033.                     dest[aidx]= A;

  1034.                     dest[0]= R>>22;

  1035.                     dest[1]= G>>22;

  1036.                     dest[2]= B>>22;

  1037.                     dest+= step;

  1038.                 }

  1039.             } else {

  1040.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  1041.                     dest[aidx]= 255;

  1042.                     dest[0]= R>>22;

  1043.                     dest[1]= G>>22;

  1044.                     dest[2]= B>>22;

  1045.                     dest+= step;

  1046.                 }

  1047.             }

  1048.         }

  1049.         break;

  1050.     case PIX_FMT_ABGR:

  1051.         dest++;

  1052.         aidx= 0;

  1053.     case PIX_FMT_BGR24:

  1054.         aidx--;

  1055.     case PIX_FMT_BGRA:

  1056.         if (CONFIG_SMALL) {

  1057.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  1058.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  1059.                 dest[aidx]= needAlpha ? A : 255;

  1060.                 dest[0]= B>>22;

  1061.                 dest[1]= G>>22;

  1062.                 dest[2]= R>>22;

  1063.                 dest+= step;

  1064.             }

  1065.         } else {

  1066.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1067.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  1068.                     dest[aidx]= A;

  1069.                     dest[0]= B>>22;

  1070.                     dest[1]= G>>22;

  1071.                     dest[2]= R>>22;

  1072.                     dest+= step;

  1073.                 }

  1074.             } else {

  1075.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  1076.                     dest[aidx]= 255;

  1077.                     dest[0]= B>>22;

  1078.                     dest[1]= G>>22;

  1079.                     dest[2]= R>>22;

  1080.                     dest+= step;

  1081.                 }

  1082.             }

  1083.         }

  1084.         break;

  1085.     default:

  1086.         assert(0);

  1087.     }

  1088. }

  1089. 

  1090. static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)

  1091. {

  1092.     int i;

  1093.     uint8_t *ptr = plane + stride*y;

  1094.     for (i=0; i<height; i++) {

  1095.         memset(ptr, val, width);

  1096.         ptr += stride;

  1097.     }

  1098. }

  1099. 

  1100. static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width)

  1101. {

  1102.     int i;

  1103.     for (i = 0; i < width; i++) {

  1104.         int r = src[i*6+0];

  1105.         int g = src[i*6+2];

  1106.         int b = src[i*6+4];

  1107. 

  1108.         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1109.     }

  1110. }

  1111. 

  1112. static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,

  1113.                              uint8_t *src1, uint8_t *src2, int width)

  1114. {

  1115.     int i;

  1116.     assert(src1==src2);

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

  1118.         int r = src1[6*i + 0];

  1119.         int g = src1[6*i + 2];

  1120.         int b = src1[6*i + 4];

  1121. 

  1122.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1123.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1124.     }

  1125. }

  1126. 

  1127. static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1128.                                   uint8_t *src1, uint8_t *src2, int width)

  1129. {

  1130.     int i;

  1131.     assert(src1==src2);

  1132.     for (i = 0; i < width; i++) {

  1133.         int r= src1[12*i + 0] + src1[12*i + 6];

  1134.         int g= src1[12*i + 2] + src1[12*i + 8];

  1135.         int b= src1[12*i + 4] + src1[12*i + 10];

  1136. 

  1137.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1138.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1139.     }

  1140. }

  1141. 

  1142. #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\

  1143. static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\

  1144. {\

  1145.     int i;\

  1146.     for (i=0; i<width; i++) {\

  1147.         int b= (((const type*)src)[i]>>shb)&maskb;\

  1148.         int g= (((const type*)src)[i]>>shg)&maskg;\

  1149.         int r= (((const type*)src)[i]>>shr)&maskr;\

  1150. \

  1151.         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\

  1152.     }\

  1153. }

  1154. 

  1155. BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1156. BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1157. BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)

  1158. BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)

  1159. BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)

  1160. BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)

  1161. 

  1162. static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1163. {

  1164.     int i;

  1165.     for (i=0; i<width; i++) {

  1166.         dst[i]= src[4*i];

  1167.     }

  1168. }

  1169. 

  1170. #define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\

  1171. static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1172. {\

  1173.     int i;\

  1174.     for (i=0; i<width; i++) {\

  1175.         int b= (((const type*)src)[i]&maskb)>>shb;\

  1176.         int g= (((const type*)src)[i]&maskg)>>shg;\

  1177.         int r= (((const type*)src)[i]&maskr)>>shr;\

  1178. \

  1179.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\

  1180.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\

  1181.     }\

  1182. }\

  1183. static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1184. {\

  1185.     int i;\

  1186.     for (i=0; i<width; i++) {\

  1187.         int pix0= ((const type*)src)[2*i+0];\

  1188.         int pix1= ((const type*)src)[2*i+1];\

  1189.         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\

  1190.         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\

  1191.         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\

  1192.         g&= maskg|(2*maskg);\

  1193. \

  1194.         g>>=shg;\

  1195. \

  1196.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\

  1197.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\

  1198.     }\

  1199. }

  1200. 

  1201. BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1202. BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1203. BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)

  1204. BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)

  1205. BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)

  1206. BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)

  1207. 

  1208. static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)

  1209. {

  1210.     int i;

  1211.     for (i=0; i<width; i++) {

  1212.         int d= src[i];

  1213. 

  1214.         dst[i]= pal[d] & 0xFF;

  1215.     }

  1216. }

  1217. 

  1218. static inline void palToUV(uint8_t *dstU, uint8_t *dstV,

  1219.                            const uint8_t *src1, const uint8_t *src2,

  1220.                            long width, uint32_t *pal)

  1221. {

  1222.     int i;

  1223.     assert(src1 == src2);

  1224.     for (i=0; i<width; i++) {

  1225.         int p= pal[src1[i]];

  1226. 

  1227.         dstU[i]= p>>8;

  1228.         dstV[i]= p>>16;

  1229.     }

  1230. }

  1231. 

  1232. static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1233. {

  1234.     int i, j;

  1235.     for (i=0; i<width/8; i++) {

  1236.         int d= ~src[i];

  1237.         for(j=0; j<8; j++)

  1238.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1239.     }

  1240. }

  1241. 

  1242. static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1243. {

  1244.     int i, j;

  1245.     for (i=0; i<width/8; i++) {

  1246.         int d= src[i];

  1247.         for(j=0; j<8; j++)

  1248.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1249.     }

  1250. }

  1251. 

  1252. 

  1253. //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one

  1254. //Plain C versions

  1255. #if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT

  1256. #define COMPILE_C

  1257. #endif

  1258. 

  1259. #if ARCH_PPC

  1260. #if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT

  1261. #define COMPILE_ALTIVEC

  1262. #endif

  1263. #endif //ARCH_PPC

  1264. 

  1265. #if ARCH_X86

  1266. 

  1267. #if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1268. #define COMPILE_MMX

  1269. #endif

  1270. 

  1271. #if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1272. #define COMPILE_MMX2

  1273. #endif

  1274. 

  1275. #if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1276. #define COMPILE_3DNOW

  1277. #endif

  1278. #endif //ARCH_X86

  1279. 

  1280. #define COMPILE_TEMPLATE_MMX 0

  1281. #define COMPILE_TEMPLATE_MMX2 0

  1282. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1283. #define COMPILE_TEMPLATE_ALTIVEC 0

  1284. 

  1285. #ifdef COMPILE_C

  1286. #define RENAME(a) a ## _C

  1287. #include "swscale_template.c"

  1288. #endif

  1289. 

  1290. #ifdef COMPILE_ALTIVEC

  1291. #undef RENAME

  1292. #undef COMPILE_TEMPLATE_ALTIVEC

  1293. #define COMPILE_TEMPLATE_ALTIVEC 1

  1294. #define RENAME(a) a ## _altivec

  1295. #include "swscale_template.c"

  1296. #endif

  1297. 

  1298. #if ARCH_X86

  1299. 

  1300. //MMX versions

  1301. #ifdef COMPILE_MMX

  1302. #undef RENAME

  1303. #undef COMPILE_TEMPLATE_MMX

  1304. #undef COMPILE_TEMPLATE_MMX2

  1305. #undef COMPILE_TEMPLATE_AMD3DNOW

  1306. #define COMPILE_TEMPLATE_MMX 1

  1307. #define COMPILE_TEMPLATE_MMX2 0

  1308. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1309. #define RENAME(a) a ## _MMX

  1310. #include "swscale_template.c"

  1311. #endif

  1312. 

  1313. //MMX2 versions

  1314. #ifdef COMPILE_MMX2

  1315. #undef RENAME

  1316. #undef COMPILE_TEMPLATE_MMX

  1317. #undef COMPILE_TEMPLATE_MMX2

  1318. #undef COMPILE_TEMPLATE_AMD3DNOW

  1319. #define COMPILE_TEMPLATE_MMX 1

  1320. #define COMPILE_TEMPLATE_MMX2 1

  1321. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1322. #define RENAME(a) a ## _MMX2

  1323. #include "swscale_template.c"

  1324. #endif

  1325. 

  1326. //3DNOW versions

  1327. #ifdef COMPILE_3DNOW

  1328. #undef RENAME

  1329. #undef COMPILE_TEMPLATE_MMX

  1330. #undef COMPILE_TEMPLATE_MMX2

  1331. #undef COMPILE_TEMPLATE_AMD3DNOW

  1332. #define COMPILE_TEMPLATE_MMX 1

  1333. #define COMPILE_TEMPLATE_MMX2 0

  1334. #define COMPILE_TEMPLATE_AMD3DNOW 1

  1335. #define RENAME(a) a ## _3DNow

  1336. #include "swscale_template.c"

  1337. #endif

  1338. 

  1339. #endif //ARCH_X86

  1340. 

  1341. static double getSplineCoeff(double a, double b, double c, double d, double dist)

  1342. {

  1343. //    printf("%f %f %f %f %f\n", a,b,c,d,dist);

  1344.     if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;

  1345.     else           return getSplineCoeff(        0.0,

  1346.                                           b+ 2.0*c + 3.0*d,

  1347.                                                  c + 3.0*d,

  1348.                                          -b- 3.0*c - 6.0*d,

  1349.                                          dist-1.0);

  1350. }

  1351. 

  1352. static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,

  1353.                              int srcW, int dstW, int filterAlign, int one, int flags,

  1354.                              SwsVector *srcFilter, SwsVector *dstFilter, double param[2])

  1355. {

  1356.     int i;

  1357.     int filterSize;

  1358.     int filter2Size;

  1359.     int minFilterSize;

  1360.     int64_t *filter=NULL;

  1361.     int64_t *filter2=NULL;

  1362.     const int64_t fone= 1LL<<54;

  1363.     int ret= -1;

  1364. #if ARCH_X86

  1365.     if (flags & SWS_CPU_CAPS_MMX)

  1366.         __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)

  1367. #endif

  1368. 

  1369.     // NOTE: the +1 is for the MMX scaler which reads over the end

  1370.     FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW+1)*sizeof(int16_t), fail);

  1371. 

  1372.     if (FFABS(xInc - 0x10000) <10) { // unscaled

  1373.         int i;

  1374.         filterSize= 1;

  1375.         FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);

  1376. 

  1377.         for (i=0; i<dstW; i++) {

  1378.             filter[i*filterSize]= fone;

  1379.             (*filterPos)[i]=i;

  1380.         }

  1381. 

  1382.     } else if (flags&SWS_POINT) { // lame looking point sampling mode

  1383.         int i;

  1384.         int xDstInSrc;

  1385.         filterSize= 1;

  1386.         FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);

  1387. 

  1388.         xDstInSrc= xInc/2 - 0x8000;

  1389.         for (i=0; i<dstW; i++) {

  1390.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1391. 

  1392.             (*filterPos)[i]= xx;

  1393.             filter[i]= fone;

  1394.             xDstInSrc+= xInc;

  1395.         }

  1396.     } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale

  1397.         int i;

  1398.         int xDstInSrc;

  1399.         filterSize= 2;

  1400.         FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);

  1401. 

  1402.         xDstInSrc= xInc/2 - 0x8000;

  1403.         for (i=0; i<dstW; i++) {

  1404.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1405.             int j;

  1406. 

  1407.             (*filterPos)[i]= xx;

  1408.             //bilinear upscale / linear interpolate / area averaging

  1409.             for (j=0; j<filterSize; j++) {

  1410.                 int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);

  1411.                 if (coeff<0) coeff=0;

  1412.                 filter[i*filterSize + j]= coeff;

  1413.                 xx++;

  1414.             }

  1415.             xDstInSrc+= xInc;

  1416.         }

  1417.     } else {

  1418.         int xDstInSrc;

  1419.         int sizeFactor;

  1420. 

  1421.         if      (flags&SWS_BICUBIC)      sizeFactor=  4;

  1422.         else if (flags&SWS_X)            sizeFactor=  8;

  1423.         else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear

  1424.         else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)

  1425.         else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;

  1426.         else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)

  1427.         else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)

  1428.         else if (flags&SWS_BILINEAR)     sizeFactor=  2;

  1429.         else {

  1430.             sizeFactor= 0; //GCC warning killer

  1431.             assert(0);

  1432.         }

  1433. 

  1434.         if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale

  1435.         else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;

  1436. 

  1437.         if (filterSize > srcW-2) filterSize=srcW-2;

  1438. 

  1439.         FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);

  1440. 

  1441.         xDstInSrc= xInc - 0x10000;

  1442.         for (i=0; i<dstW; i++) {

  1443.             int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);

  1444.             int j;

  1445.             (*filterPos)[i]= xx;

  1446.             for (j=0; j<filterSize; j++) {

  1447.                 int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;

  1448.                 double floatd;

  1449.                 int64_t coeff;

  1450. 

  1451.                 if (xInc > 1<<16)

  1452.                     d= d*dstW/srcW;

  1453.                 floatd= d * (1.0/(1<<30));

  1454. 

  1455.                 if (flags & SWS_BICUBIC) {

  1456.                     int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);

  1457.                     int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);

  1458.                     int64_t dd = ( d*d)>>30;

  1459.                     int64_t ddd= (dd*d)>>30;

  1460. 

  1461.                     if      (d < 1LL<<30)

  1462.                         coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);

  1463.                     else if (d < 1LL<<31)

  1464.                         coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);

  1465.                     else

  1466.                         coeff=0.0;

  1467.                     coeff *= fone>>(30+24);

  1468.                 }

  1469. /*                else if (flags & SWS_X) {

  1470.                     double p= param ? param*0.01 : 0.3;

  1471.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1472.                     coeff*= pow(2.0, - p*d*d);

  1473.                 }*/

  1474.                 else if (flags & SWS_X) {

  1475.                     double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

  1476.                     double c;

  1477. 

  1478.                     if (floatd<1.0)

  1479.                         c = cos(floatd*PI);

  1480.                     else

  1481.                         c=-1.0;

  1482.                     if (c<0.0)      c= -pow(-c, A);

  1483.                     else            c=  pow( c, A);

  1484.                     coeff= (c*0.5 + 0.5)*fone;

  1485.                 } else if (flags & SWS_AREA) {

  1486.                     int64_t d2= d - (1<<29);

  1487.                     if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));

  1488.                     else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));

  1489.                     else coeff=0.0;

  1490.                     coeff *= fone>>(30+16);

  1491.                 } else if (flags & SWS_GAUSS) {

  1492.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1493.                     coeff = (pow(2.0, - p*floatd*floatd))*fone;

  1494.                 } else if (flags & SWS_SINC) {

  1495.                     coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;

  1496.                 } else if (flags & SWS_LANCZOS) {

  1497.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1498.                     coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;

  1499.                     if (floatd>p) coeff=0;

  1500.                 } else if (flags & SWS_BILINEAR) {

  1501.                     coeff= (1<<30) - d;

  1502.                     if (coeff<0) coeff=0;

  1503.                     coeff *= fone >> 30;

  1504.                 } else if (flags & SWS_SPLINE) {

  1505.                     double p=-2.196152422706632;

  1506.                     coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;

  1507.                 } else {

  1508.                     coeff= 0.0; //GCC warning killer

  1509.                     assert(0);

  1510.                 }

  1511. 

  1512.                 filter[i*filterSize + j]= coeff;

  1513.                 xx++;

  1514.             }

  1515.             xDstInSrc+= 2*xInc;

  1516.         }

  1517.     }

  1518. 

  1519.     /* apply src & dst Filter to filter -> filter2

  1520.        av_free(filter);

  1521.     */

  1522.     assert(filterSize>0);

  1523.     filter2Size= filterSize;

  1524.     if (srcFilter) filter2Size+= srcFilter->length - 1;

  1525.     if (dstFilter) filter2Size+= dstFilter->length - 1;

  1526.     assert(filter2Size>0);

  1527.     FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail);

  1528. 

  1529.     for (i=0; i<dstW; i++) {

  1530.         int j, k;

  1531. 

  1532.         if(srcFilter) {

  1533.             for (k=0; k<srcFilter->length; k++) {

  1534.                 for (j=0; j<filterSize; j++)

  1535.                     filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];

  1536.             }

  1537.         } else {

  1538.             for (j=0; j<filterSize; j++)

  1539.                 filter2[i*filter2Size + j]= filter[i*filterSize + j];

  1540.         }

  1541.         //FIXME dstFilter

  1542. 

  1543.         (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

  1544.     }

  1545.     av_freep(&filter);

  1546. 

  1547.     /* try to reduce the filter-size (step1 find size and shift left) */

  1548.     // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).

  1549.     minFilterSize= 0;

  1550.     for (i=dstW-1; i>=0; i--) {

  1551.         int min= filter2Size;

  1552.         int j;

  1553.         int64_t cutOff=0.0;

  1554. 

  1555.         /* get rid off near zero elements on the left by shifting left */

  1556.         for (j=0; j<filter2Size; j++) {

  1557.             int k;

  1558.             cutOff += FFABS(filter2[i*filter2Size]);

  1559. 

  1560.             if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;

  1561. 

  1562.             /* preserve monotonicity because the core can't handle the filter otherwise */

  1563.             if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

  1564. 

  1565.             // move filter coefficients left

  1566.             for (k=1; k<filter2Size; k++)

  1567.                 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];

  1568.             filter2[i*filter2Size + k - 1]= 0;

  1569.             (*filterPos)[i]++;

  1570.         }

  1571. 

  1572.         cutOff=0;

  1573.         /* count near zeros on the right */

  1574.         for (j=filter2Size-1; j>0; j--) {

  1575.             cutOff += FFABS(filter2[i*filter2Size + j]);

  1576. 

  1577.             if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;

  1578.             min--;

  1579.         }

  1580. 

  1581.         if (min>minFilterSize) minFilterSize= min;

  1582.     }

  1583. 

  1584.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1585.         // we can handle the special case 4,

  1586.         // so we don't want to go to the full 8

  1587.         if (minFilterSize < 5)

  1588.             filterAlign = 4;

  1589. 

  1590.         // We really don't want to waste our time

  1591.         // doing useless computation, so fall back on

  1592.         // the scalar C code for very small filters.

  1593.         // Vectorizing is worth it only if you have a

  1594.         // decent-sized vector.

  1595.         if (minFilterSize < 3)

  1596.             filterAlign = 1;

  1597.     }

  1598. 

  1599.     if (flags & SWS_CPU_CAPS_MMX) {

  1600.         // special case for unscaled vertical filtering

  1601.         if (minFilterSize == 1 && filterAlign == 2)

  1602.             filterAlign= 1;

  1603.     }

  1604. 

  1605.     assert(minFilterSize > 0);

  1606.     filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));

  1607.     assert(filterSize > 0);

  1608.     filter= av_malloc(filterSize*dstW*sizeof(*filter));

  1609.     if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)

  1610.         goto fail;

  1611.     *outFilterSize= filterSize;

  1612. 

  1613.     if (flags&SWS_PRINT_INFO)

  1614.         av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);

  1615.     /* try to reduce the filter-size (step2 reduce it) */

  1616.     for (i=0; i<dstW; i++) {

  1617.         int j;

  1618. 

  1619.         for (j=0; j<filterSize; j++) {

  1620.             if (j>=filter2Size) filter[i*filterSize + j]= 0;

  1621.             else               filter[i*filterSize + j]= filter2[i*filter2Size + j];

  1622.             if((flags & SWS_BITEXACT) && j>=minFilterSize)

  1623.                 filter[i*filterSize + j]= 0;

  1624.         }

  1625.     }

  1626. 

  1627. 

  1628.     //FIXME try to align filterPos if possible

  1629. 

  1630.     //fix borders

  1631.     for (i=0; i<dstW; i++) {

  1632.         int j;

  1633.         if ((*filterPos)[i] < 0) {

  1634.             // move filter coefficients left to compensate for filterPos

  1635.             for (j=1; j<filterSize; j++) {

  1636.                 int left= FFMAX(j + (*filterPos)[i], 0);

  1637.                 filter[i*filterSize + left] += filter[i*filterSize + j];

  1638.                 filter[i*filterSize + j]=0;

  1639.             }

  1640.             (*filterPos)[i]= 0;

  1641.         }

  1642. 

  1643.         if ((*filterPos)[i] + filterSize > srcW) {

  1644.             int shift= (*filterPos)[i] + filterSize - srcW;

  1645.             // move filter coefficients right to compensate for filterPos

  1646.             for (j=filterSize-2; j>=0; j--) {

  1647.                 int right= FFMIN(j + shift, filterSize-1);

  1648.                 filter[i*filterSize +right] += filter[i*filterSize +j];

  1649.                 filter[i*filterSize +j]=0;

  1650.             }

  1651.             (*filterPos)[i]= srcW - filterSize;

  1652.         }

  1653.     }

  1654. 

  1655.     // Note the +1 is for the MMX scaler which reads over the end

  1656.     /* align at 16 for AltiVec (needed by hScale_altivec_real) */

  1657.     FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+1)*sizeof(int16_t), fail);

  1658. 

  1659.     /* normalize & store in outFilter */

  1660.     for (i=0; i<dstW; i++) {

  1661.         int j;

  1662.         int64_t error=0;

  1663.         int64_t sum=0;

  1664. 

  1665.         for (j=0; j<filterSize; j++) {

  1666.             sum+= filter[i*filterSize + j];

  1667.         }

  1668.         sum= (sum + one/2)/ one;

  1669.         for (j=0; j<*outFilterSize; j++) {

  1670.             int64_t v= filter[i*filterSize + j] + error;

  1671.             int intV= ROUNDED_DIV(v, sum);

  1672.             (*outFilter)[i*(*outFilterSize) + j]= intV;

  1673.             error= v - intV*sum;

  1674.         }

  1675.     }

  1676. 

  1677.     (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end

  1678.     for (i=0; i<*outFilterSize; i++) {

  1679.         int j= dstW*(*outFilterSize);

  1680.         (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

  1681.     }

  1682. 

  1683.     ret=0;

  1684. fail:

  1685.     av_free(filter);

  1686.     av_free(filter2);

  1687.     return ret;

  1688. }

  1689. 

  1690. #ifdef COMPILE_MMX2

  1691. static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)

  1692. {

  1693.     uint8_t *fragmentA;

  1694.     x86_reg imm8OfPShufW1A;

  1695.     x86_reg imm8OfPShufW2A;

  1696.     x86_reg fragmentLengthA;

  1697.     uint8_t *fragmentB;

  1698.     x86_reg imm8OfPShufW1B;

  1699.     x86_reg imm8OfPShufW2B;

  1700.     x86_reg fragmentLengthB;

  1701.     int fragmentPos;

  1702. 

  1703.     int xpos, i;

  1704. 

  1705.     // create an optimized horizontal scaling routine

  1706.     /* This scaler is made of runtime-generated MMX2 code using specially

  1707.      * tuned pshufw instructions. For every four output pixels, if four

  1708.      * input pixels are enough for the fast bilinear scaling, then a chunk

  1709.      * of fragmentB is used. If five input pixels are needed, then a chunk

  1710.      * of fragmentA is used.

  1711.      */

  1712. 

  1713.     //code fragment

  1714. 

  1715.     __asm__ volatile(

  1716.         "jmp                         9f                 \n\t"

  1717.     // Begin

  1718.         "0:                                             \n\t"

  1719.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1720.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1721.         "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"

  1722.         "punpcklbw                %%mm7, %%mm1          \n\t"

  1723.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1724.         "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"

  1725.         "1:                                             \n\t"

  1726.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1727.         "2:                                             \n\t"

  1728.         "psubw                    %%mm1, %%mm0          \n\t"

  1729.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1730.         "pmullw                   %%mm3, %%mm0          \n\t"

  1731.         "psllw                       $7, %%mm1          \n\t"

  1732.         "paddw                    %%mm1, %%mm0          \n\t"

  1733. 

  1734.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1735. 

  1736.         "add                         $8, %%"REG_a"      \n\t"

  1737.     // End

  1738.         "9:                                             \n\t"

  1739. //        "int $3                                         \n\t"

  1740.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1741.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1742.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1743.         "dec                         %1                 \n\t"

  1744.         "dec                         %2                 \n\t"

  1745.         "sub                         %0, %1             \n\t"

  1746.         "sub                         %0, %2             \n\t"

  1747.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1748.         "sub                         %0, %3             \n\t"

  1749. 

  1750. 

  1751.         :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),

  1752.         "=r" (fragmentLengthA)

  1753.     );

  1754. 

  1755.     __asm__ volatile(

  1756.         "jmp                         9f                 \n\t"

  1757.     // Begin

  1758.         "0:                                             \n\t"

  1759.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1760.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1761.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1762.         "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"

  1763.         "1:                                             \n\t"

  1764.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1765.         "2:                                             \n\t"

  1766.         "psubw                    %%mm1, %%mm0          \n\t"

  1767.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1768.         "pmullw                   %%mm3, %%mm0          \n\t"

  1769.         "psllw                       $7, %%mm1          \n\t"

  1770.         "paddw                    %%mm1, %%mm0          \n\t"

  1771. 

  1772.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1773. 

  1774.         "add                         $8, %%"REG_a"      \n\t"

  1775.     // End

  1776.         "9:                                             \n\t"

  1777. //        "int                       $3                   \n\t"

  1778.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1779.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1780.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1781.         "dec                         %1                 \n\t"

  1782.         "dec                         %2                 \n\t"

  1783.         "sub                         %0, %1             \n\t"

  1784.         "sub                         %0, %2             \n\t"

  1785.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1786.         "sub                         %0, %3             \n\t"

  1787. 

  1788. 

  1789.         :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),

  1790.         "=r" (fragmentLengthB)

  1791.     );

  1792. 

  1793.     xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers

  1794.     fragmentPos=0;

  1795. 

  1796.     for (i=0; i<dstW/numSplits; i++) {

  1797.         int xx=xpos>>16;

  1798. 

  1799.         if ((i&3) == 0) {

  1800.             int a=0;

  1801.             int b=((xpos+xInc)>>16) - xx;

  1802.             int c=((xpos+xInc*2)>>16) - xx;

  1803.             int d=((xpos+xInc*3)>>16) - xx;

  1804.             int inc                = (d+1<4);

  1805.             uint8_t *fragment      = (d+1<4) ? fragmentB       : fragmentA;

  1806.             x86_reg imm8OfPShufW1  = (d+1<4) ? imm8OfPShufW1B  : imm8OfPShufW1A;

  1807.             x86_reg imm8OfPShufW2  = (d+1<4) ? imm8OfPShufW2B  : imm8OfPShufW2A;

  1808.             x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;

  1809.             int maxShift= 3-(d+inc);

  1810.             int shift=0;

  1811. 

  1812.             if (filterCode) {

  1813.                 filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;

  1814.                 filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;

  1815.                 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;

  1816.                 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;

  1817.                 filterPos[i/2]= xx;

  1818. 

  1819.                 memcpy(filterCode + fragmentPos, fragment, fragmentLength);

  1820. 

  1821.                 filterCode[fragmentPos + imm8OfPShufW1]=

  1822.                     (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);

  1823.                 filterCode[fragmentPos + imm8OfPShufW2]=

  1824.                     a | (b<<2) | (c<<4) | (d<<6);

  1825. 

  1826.                 if (i+4-inc>=dstW) shift=maxShift; //avoid overread

  1827.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

  1828. 

  1829.                 if (shift && i>=shift) {

  1830.                     filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;

  1831.                     filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;

  1832.                     filterPos[i/2]-=shift;

  1833.                 }

  1834.             }

  1835. 

  1836.             fragmentPos+= fragmentLength;

  1837. 

  1838.             if (filterCode)

  1839.                 filterCode[fragmentPos]= RET;

  1840.         }

  1841.         xpos+=xInc;

  1842.     }

  1843.     if (filterCode)

  1844.         filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part

  1845. 

  1846.     return fragmentPos + 1;

  1847. }

  1848. #endif /* COMPILE_MMX2 */

  1849. 

  1850. static void globalInit(void)

  1851. {

  1852.     // generating tables:

  1853.     int i;

  1854.     for (i=0; i<768; i++) {

  1855.         int c= av_clip_uint8(i-256);

  1856.         clip_table[i]=c;

  1857.     }

  1858. }

  1859. 

  1860. static SwsFunc getSwsFunc(SwsContext *c)

  1861. {

  1862. #if CONFIG_RUNTIME_CPUDETECT

  1863.     int flags = c->flags;

  1864. 

  1865. #if ARCH_X86 && CONFIG_GPL

  1866.     // ordered per speed fastest first

  1867.     if (flags & SWS_CPU_CAPS_MMX2) {

  1868.         sws_init_swScale_MMX2(c);

  1869.         return swScale_MMX2;

  1870.     } else if (flags & SWS_CPU_CAPS_3DNOW) {

  1871.         sws_init_swScale_3DNow(c);

  1872.         return swScale_3DNow;

  1873.     } else if (flags & SWS_CPU_CAPS_MMX) {

  1874.         sws_init_swScale_MMX(c);

  1875.         return swScale_MMX;

  1876.     } else {

  1877.         sws_init_swScale_C(c);

  1878.         return swScale_C;

  1879.     }

  1880. 

  1881. #else

  1882. #if ARCH_PPC

  1883.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1884.         sws_init_swScale_altivec(c);

  1885.         return swScale_altivec;

  1886.     } else {

  1887.         sws_init_swScale_C(c);

  1888.         return swScale_C;

  1889.     }

  1890. #endif

  1891.     sws_init_swScale_C(c);

  1892.     return swScale_C;

  1893. #endif /* ARCH_X86 && CONFIG_GPL */

  1894. #else //CONFIG_RUNTIME_CPUDETECT

  1895. #if   COMPILE_TEMPLATE_MMX2

  1896.     sws_init_swScale_MMX2(c);

  1897.     return swScale_MMX2;

  1898. #elif COMPILE_TEMPLATE_AMD3DNOW

  1899.     sws_init_swScale_3DNow(c);

  1900.     return swScale_3DNow;

  1901. #elif COMPILE_TEMPLATE_MMX

  1902.     sws_init_swScale_MMX(c);

  1903.     return swScale_MMX;

  1904. #elif COMPILE_TEMPLATE_ALTIVEC

  1905.     sws_init_swScale_altivec(c);

  1906.     return swScale_altivec;

  1907. #else

  1908.     sws_init_swScale_C(c);

  1909.     return swScale_C;

  1910. #endif

  1911. #endif //!CONFIG_RUNTIME_CPUDETECT

  1912. }

  1913. 

  1914. static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1915.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1916. {

  1917.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1918.     /* Copy Y plane */

  1919.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1920.         memcpy(dst, src[0], srcSliceH*dstStride[0]);

  1921.     else {

  1922.         int i;

  1923.         const uint8_t *srcPtr= src[0];

  1924.         uint8_t *dstPtr= dst;

  1925.         for (i=0; i<srcSliceH; i++) {

  1926.             memcpy(dstPtr, srcPtr, c->srcW);

  1927.             srcPtr+= srcStride[0];

  1928.             dstPtr+= dstStride[0];

  1929.         }

  1930.     }

  1931.     dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1932.     if (c->dstFormat == PIX_FMT_NV12)

  1933.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  1934.     else

  1935.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  1936. 

  1937.     return srcSliceH;

  1938. }

  1939. 

  1940. static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1941.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1942. {

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

  1944. 

  1945.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1946. 

  1947.     return srcSliceH;

  1948. }

  1949. 

  1950. static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1951.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1952. {

  1953.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1954. 

  1955.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1956. 

  1957.     return srcSliceH;

  1958. }

  1959. 

  1960. static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1961.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1962. {

  1963.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1964. 

  1965.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1966. 

  1967.     return srcSliceH;

  1968. }

  1969. 

  1970. static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1971.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1972. {

  1973.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1974. 

  1975.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1976. 

  1977.     return srcSliceH;

  1978. }

  1979. 

  1980. static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1981.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1982. {

  1983.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1984.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1985.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1986. 

  1987.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1988. 

  1989.     if (dstParam[3])

  1990.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1991. 

  1992.     return srcSliceH;

  1993. }

  1994. 

  1995. static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1996.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1997. {

  1998.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1999.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  2000.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  2001. 

  2002.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2003. 

  2004.     return srcSliceH;

  2005. }

  2006. 

  2007. static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2008.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2009. {

  2010.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2011.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  2012.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  2013. 

  2014.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2015. 

  2016.     if (dstParam[3])

  2017.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2018. 

  2019.     return srcSliceH;

  2020. }

  2021. 

  2022. static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2023.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2024. {

  2025.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2026.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  2027.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  2028. 

  2029.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2030. 

  2031.     return srcSliceH;

  2032. }

  2033. 

  2034. static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2035.                           int srcSliceH, uint8_t* dst[], int dstStride[])

  2036. {

  2037.     const enum PixelFormat srcFormat= c->srcFormat;

  2038.     const enum PixelFormat dstFormat= c->dstFormat;

  2039.     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,

  2040.                  const uint8_t *palette)=NULL;

  2041.     int i;

  2042.     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2043.     uint8_t *srcPtr= src[0];

  2044. 

  2045.     if (!usePal(srcFormat))

  2046.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2047.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  2048. 

  2049.     switch(dstFormat) {

  2050.     case PIX_FMT_RGB32  : conv = palette8topacked32; break;

  2051.     case PIX_FMT_BGR32  : conv = palette8topacked32; break;

  2052.     case PIX_FMT_BGR32_1: conv = palette8topacked32; break;

  2053.     case PIX_FMT_RGB32_1: conv = palette8topacked32; break;

  2054.     case PIX_FMT_RGB24  : conv = palette8topacked24; break;

  2055.     case PIX_FMT_BGR24  : conv = palette8topacked24; break;

  2056.     default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2057.                     sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2058.     }

  2059. 

  2060. 

  2061.     for (i=0; i<srcSliceH; i++) {

  2062.         conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  2063.         srcPtr+= srcStride[0];

  2064.         dstPtr+= dstStride[0];

  2065.     }

  2066. 

  2067.     return srcSliceH;

  2068. }

  2069. 

  2070. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  2071. static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

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

  2073. {

  2074.     const enum PixelFormat srcFormat= c->srcFormat;

  2075.     const enum PixelFormat dstFormat= c->dstFormat;

  2076.     const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;

  2077.     const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;

  2078.     const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  2079.     const int dstId= fmt_depth(dstFormat) >> 2;

  2080.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  2081. 

  2082.     /* BGR -> BGR */

  2083.     if (  (isBGR(srcFormat) && isBGR(dstFormat))

  2084.        || (isRGB(srcFormat) && isRGB(dstFormat))) {

  2085.         switch(srcId | (dstId<<4)) {

  2086.         case 0x34: conv= rgb16to15; break;

  2087.         case 0x36: conv= rgb24to15; break;

  2088.         case 0x38: conv= rgb32to15; break;

  2089.         case 0x43: conv= rgb15to16; break;

  2090.         case 0x46: conv= rgb24to16; break;

  2091.         case 0x48: conv= rgb32to16; break;

  2092.         case 0x63: conv= rgb15to24; break;

  2093.         case 0x64: conv= rgb16to24; break;

  2094.         case 0x68: conv= rgb32to24; break;

  2095.         case 0x83: conv= rgb15to32; break;

  2096.         case 0x84: conv= rgb16to32; break;

  2097.         case 0x86: conv= rgb24to32; break;

  2098.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2099.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2100.         }

  2101.     } else if (  (isBGR(srcFormat) && isRGB(dstFormat))

  2102.              || (isRGB(srcFormat) && isBGR(dstFormat))) {

  2103.         switch(srcId | (dstId<<4)) {

  2104.         case 0x33: conv= rgb15tobgr15; break;

  2105.         case 0x34: conv= rgb16tobgr15; break;

  2106.         case 0x36: conv= rgb24tobgr15; break;

  2107.         case 0x38: conv= rgb32tobgr15; break;

  2108.         case 0x43: conv= rgb15tobgr16; break;

  2109.         case 0x44: conv= rgb16tobgr16; break;

  2110.         case 0x46: conv= rgb24tobgr16; break;

  2111.         case 0x48: conv= rgb32tobgr16; break;

  2112.         case 0x63: conv= rgb15tobgr24; break;

  2113.         case 0x64: conv= rgb16tobgr24; break;

  2114.         case 0x66: conv= rgb24tobgr24; break;

  2115.         case 0x68: conv= rgb32tobgr24; break;

  2116.         case 0x83: conv= rgb15tobgr32; break;

  2117.         case 0x84: conv= rgb16tobgr32; break;

  2118.         case 0x86: conv= rgb24tobgr32; break;

  2119.         case 0x88: conv= rgb32tobgr32; break;

  2120.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2121.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2122.         }

  2123.     } else {

  2124.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2125.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  2126.     }

  2127. 

  2128.     if(conv) {

  2129.         uint8_t *srcPtr= src[0];

  2130.         if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)

  2131.             srcPtr += ALT32_CORR;

  2132. 

  2133.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  2134.             conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  2135.         else {

  2136.             int i;

  2137.             uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2138. 

  2139.             for (i=0; i<srcSliceH; i++) {

  2140.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  2141.                 srcPtr+= srcStride[0];

  2142.                 dstPtr+= dstStride[0];

  2143.             }

  2144.         }

  2145.     }

  2146.     return srcSliceH;

  2147. }

  2148. 

  2149. static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2150.                               int srcSliceH, uint8_t* dst[], int dstStride[])

  2151. {

  2152. 

  2153.     rgb24toyv12(

  2154.         src[0],

  2155.         dst[0]+ srcSliceY    *dstStride[0],

  2156.         dst[1]+(srcSliceY>>1)*dstStride[1],

  2157.         dst[2]+(srcSliceY>>1)*dstStride[2],

  2158.         c->srcW, srcSliceH,

  2159.         dstStride[0], dstStride[1], srcStride[0]);

  2160.     if (dst[3])

  2161.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2162.     return srcSliceH;

  2163. }

  2164. 

  2165. static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2166.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  2167. {

  2168.     int i;

  2169. 

  2170.     /* copy Y */

  2171.     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)

  2172.         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);

  2173.     else {

  2174.         uint8_t *srcPtr= src[0];

  2175.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2176. 

  2177.         for (i=0; i<srcSliceH; i++) {

  2178.             memcpy(dstPtr, srcPtr, c->srcW);

  2179.             srcPtr+= srcStride[0];

  2180.             dstPtr+= dstStride[0];

  2181.         }

  2182.     }

  2183. 

  2184.     if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {

  2185.         planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  2186.                  srcSliceH >> 2, srcStride[1], dstStride[1]);

  2187.         planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  2188.                  srcSliceH >> 2, srcStride[2], dstStride[2]);

  2189.     } else {

  2190.         planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  2191.                  srcSliceH >> 2, srcStride[1], dstStride[2]);

  2192.         planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  2193.                  srcSliceH >> 2, srcStride[2], dstStride[1]);

  2194.     }

  2195.     if (dst[3])

  2196.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2197.     return srcSliceH;

  2198. }

  2199. 

  2200. /* unscaled copy like stuff (assumes nearly identical formats) */

  2201. static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2202.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  2203. {

  2204.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  2205.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  2206.     else {

  2207.         int i;

  2208.         uint8_t *srcPtr= src[0];

  2209.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2210.         int length=0;

  2211. 

  2212.         /* universal length finder */

  2213.         while(length+c->srcW <= FFABS(dstStride[0])

  2214.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  2215.         assert(length!=0);

  2216. 

  2217.         for (i=0; i<srcSliceH; i++) {

  2218.             memcpy(dstPtr, srcPtr, length);

  2219.             srcPtr+= srcStride[0];

  2220.             dstPtr+= dstStride[0];

  2221.         }

  2222.     }

  2223.     return srcSliceH;

  2224. }

  2225. 

  2226. static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2227.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  2228. {

  2229.     int plane, i, j;

  2230.     for (plane=0; plane<4; plane++) {

  2231.         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  2232.         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  2233.         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  2234.         uint8_t *srcPtr= src[plane];

  2235.         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  2236. 

  2237.         if (!dst[plane]) continue;

  2238.         // ignore palette for GRAY8

  2239.         if (plane == 1 && !dst[2]) continue;

  2240.         if (!src[plane] || (plane == 1 && !src[2])) {

  2241.             if(is16BPS(c->dstFormat))

  2242.                 length*=2;

  2243.             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);

  2244.         } else {

  2245.             if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {

  2246.                 if (!isBE(c->srcFormat)) srcPtr++;

  2247.                 for (i=0; i<height; i++) {

  2248.                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  2249.                     srcPtr+= srcStride[plane];

  2250.                     dstPtr+= dstStride[plane];

  2251.                 }

  2252.             } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {

  2253.                 for (i=0; i<height; i++) {

  2254.                     for (j=0; j<length; j++) {

  2255.                         dstPtr[ j<<1   ] = srcPtr[j];

  2256.                         dstPtr[(j<<1)+1] = srcPtr[j];

  2257.                     }

  2258.                     srcPtr+= srcStride[plane];

  2259.                     dstPtr+= dstStride[plane];

  2260.                 }

  2261.             } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)

  2262.                   && isBE(c->srcFormat) != isBE(c->dstFormat)) {

  2263. 

  2264.                 for (i=0; i<height; i++) {

  2265.                     for (j=0; j<length; j++)

  2266.                         ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);

  2267.                     srcPtr+= srcStride[plane];

  2268.                     dstPtr+= dstStride[plane];

  2269.                 }

  2270.             } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)

  2271.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);

  2272.             else {

  2273.                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))

  2274.                     length*=2;

  2275.                 for (i=0; i<height; i++) {

  2276.                     memcpy(dstPtr, srcPtr, length);

  2277.                     srcPtr+= srcStride[plane];

  2278.                     dstPtr+= dstStride[plane];

  2279.                 }

  2280.             }

  2281.         }

  2282.     }

  2283.     return srcSliceH;

  2284. }

  2285. 

  2286. 

  2287. static void getSubSampleFactors(int *h, int *v, enum PixelFormat format)

  2288. {

  2289.     *h = av_pix_fmt_descriptors[format].log2_chroma_w;

  2290.     *v = av_pix_fmt_descriptors[format].log2_chroma_h;

  2291. }

  2292. 

  2293. static uint16_t roundToInt16(int64_t f)

  2294. {

  2295.     int r= (f + (1<<15))>>16;

  2296.          if (r<-0x7FFF) return 0x8000;

  2297.     else if (r> 0x7FFF) return 0x7FFF;

  2298.     else                return r;

  2299. }

  2300. 

  2301. int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation)

  2302. {

  2303.     int64_t crv =  inv_table[0];

  2304.     int64_t cbu =  inv_table[1];

  2305.     int64_t cgu = -inv_table[2];

  2306.     int64_t cgv = -inv_table[3];

  2307.     int64_t cy  = 1<<16;

  2308.     int64_t oy  = 0;

  2309. 

  2310.     memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);

  2311.     memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);

  2312. 

  2313.     c->brightness= brightness;

  2314.     c->contrast  = contrast;

  2315.     c->saturation= saturation;

  2316.     c->srcRange  = srcRange;

  2317.     c->dstRange  = dstRange;

  2318.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  2319. 

  2320.     c->uOffset=   0x0400040004000400LL;

  2321.     c->vOffset=   0x0400040004000400LL;

  2322. 

  2323.     if (!srcRange) {

  2324.         cy= (cy*255) / 219;

  2325.         oy= 16<<16;

  2326.     } else {

  2327.         crv= (crv*224) / 255;

  2328.         cbu= (cbu*224) / 255;

  2329.         cgu= (cgu*224) / 255;

  2330.         cgv= (cgv*224) / 255;

  2331.     }

  2332. 

  2333.     cy = (cy *contrast             )>>16;

  2334.     crv= (crv*contrast * saturation)>>32;

  2335.     cbu= (cbu*contrast * saturation)>>32;

  2336.     cgu= (cgu*contrast * saturation)>>32;

  2337.     cgv= (cgv*contrast * saturation)>>32;

  2338. 

  2339.     oy -= 256*brightness;

  2340. 

  2341.     c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;

  2342.     c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;

  2343.     c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;

  2344.     c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;

  2345.     c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;

  2346.     c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;

  2347. 

  2348.     c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);

  2349.     c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);

  2350.     c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);

  2351.     c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);

  2352.     c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);

  2353.     c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);

  2354. 

  2355.     ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);

  2356.     //FIXME factorize

  2357. 

  2358. #ifdef COMPILE_ALTIVEC

  2359.     if (c->flags & SWS_CPU_CAPS_ALTIVEC)

  2360.         ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);

  2361. #endif

  2362.     return 0;

  2363. }

  2364. 

  2365. int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation)

  2366. {

  2367.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  2368. 

  2369.     *inv_table = c->srcColorspaceTable;

  2370.     *table     = c->dstColorspaceTable;

  2371.     *srcRange  = c->srcRange;

  2372.     *dstRange  = c->dstRange;

  2373.     *brightness= c->brightness;

  2374.     *contrast  = c->contrast;

  2375.     *saturation= c->saturation;

  2376. 

  2377.     return 0;

  2378. }

  2379. 

  2380. static int handle_jpeg(enum PixelFormat *format)

  2381. {

  2382.     switch (*format) {

  2383.     case PIX_FMT_YUVJ420P:

  2384.         *format = PIX_FMT_YUV420P;

  2385.         return 1;

  2386.     case PIX_FMT_YUVJ422P:

  2387.         *format = PIX_FMT_YUV422P;

  2388.         return 1;

  2389.     case PIX_FMT_YUVJ444P:

  2390.         *format = PIX_FMT_YUV444P;

  2391.         return 1;

  2392.     case PIX_FMT_YUVJ440P:

  2393.         *format = PIX_FMT_YUV440P;

  2394.         return 1;

  2395.     default:

  2396.         return 0;

  2397.     }

  2398. }

  2399. 

  2400. SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,

  2401.                            SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)

  2402. {

  2403. 

  2404.     SwsContext *c;

  2405.     int i;

  2406.     int usesVFilter, usesHFilter;

  2407.     int unscaled, needsDither;

  2408.     int srcRange, dstRange;

  2409.     SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};

  2410. #if ARCH_X86

  2411.     if (flags & SWS_CPU_CAPS_MMX)

  2412.         __asm__ volatile("emms\n\t"::: "memory");

  2413. #endif

  2414. 

  2415. #if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off

  2416.     flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);

  2417. #if   COMPILE_TEMPLATE_MMX2

  2418.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  2419. #elif COMPILE_TEMPLATE_AMD3DNOW

  2420.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  2421. #elif COMPILE_TEMPLATE_MMX

  2422.     flags |= SWS_CPU_CAPS_MMX;

  2423. #elif COMPILE_TEMPLATE_ALTIVEC

  2424.     flags |= SWS_CPU_CAPS_ALTIVEC;

  2425. #elif ARCH_BFIN

  2426.     flags |= SWS_CPU_CAPS_BFIN;

  2427. #endif

  2428. #endif /* CONFIG_RUNTIME_CPUDETECT */

  2429.     if (clip_table[512] != 255) globalInit();

  2430.     if (!rgb15to16) sws_rgb2rgb_init(flags);

  2431. 

  2432.     unscaled = (srcW == dstW && srcH == dstH);

  2433.     needsDither= (isBGR(dstFormat) || isRGB(dstFormat))

  2434.         && (fmt_depth(dstFormat))<24

  2435.         && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

  2436. 

  2437.     srcRange = handle_jpeg(&srcFormat);

  2438.     dstRange = handle_jpeg(&dstFormat);

  2439. 

  2440.     if (!isSupportedIn(srcFormat)) {

  2441.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));

  2442.         return NULL;

  2443.     }

  2444.     if (!isSupportedOut(dstFormat)) {

  2445.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));

  2446.         return NULL;

  2447.     }

  2448. 

  2449.     i= flags & ( SWS_POINT

  2450.                 |SWS_AREA

  2451.                 |SWS_BILINEAR

  2452.                 |SWS_FAST_BILINEAR

  2453.                 |SWS_BICUBIC

  2454.                 |SWS_X

  2455.                 |SWS_GAUSS

  2456.                 |SWS_LANCZOS

  2457.                 |SWS_SINC

  2458.                 |SWS_SPLINE

  2459.                 |SWS_BICUBLIN);

  2460.     if(!i || (i & (i-1))) {

  2461.         av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");

  2462.         return NULL;

  2463.     }

  2464. 

  2465.     /* sanity check */

  2466.     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

  2467.         av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",

  2468.                srcW, srcH, dstW, dstH);

  2469.         return NULL;

  2470.     }

  2471.     if(srcW > VOFW || dstW > VOFW) {

  2472.         av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");

  2473.         return NULL;

  2474.     }

  2475. 

  2476.     if (!dstFilter) dstFilter= &dummyFilter;

  2477.     if (!srcFilter) srcFilter= &dummyFilter;

  2478. 

  2479.     FF_ALLOCZ_OR_GOTO(NULL, c, sizeof(SwsContext), fail);

  2480. 

  2481.     c->av_class = &sws_context_class;

  2482.     c->srcW= srcW;

  2483.     c->srcH= srcH;

  2484.     c->dstW= dstW;

  2485.     c->dstH= dstH;

  2486.     c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;

  2487.     c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;

  2488.     c->flags= flags;

  2489.     c->dstFormat= dstFormat;

  2490.     c->srcFormat= srcFormat;

  2491.     c->vRounder= 4* 0x0001000100010001ULL;

  2492. 

  2493.     usesHFilter= usesVFilter= 0;

  2494.     if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;

  2495.     if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;

  2496.     if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;

  2497.     if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;

  2498.     if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;

  2499.     if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;

  2500.     if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;

  2501.     if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;

  2502. 

  2503.     getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);

  2504.     getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

  2505. 

  2506.     // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation

  2507.     if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

  2508. 

  2509.     // drop some chroma lines if the user wants it

  2510.     c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;

  2511.     c->chrSrcVSubSample+= c->vChrDrop;

  2512. 

  2513.     // drop every other pixel for chroma calculation unless user wants full chroma

  2514.     if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)

  2515.       && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8

  2516.       && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4

  2517.       && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE

  2518.       && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2519.         c->chrSrcHSubSample=1;

  2520. 

  2521.     if (param) {

  2522.         c->param[0] = param[0];

  2523.         c->param[1] = param[1];

  2524.     } else {

  2525.         c->param[0] =

  2526.         c->param[1] = SWS_PARAM_DEFAULT;

  2527.     }

  2528. 

  2529.     // Note the -((-x)>>y) is so that we always round toward +inf.

  2530.     c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);

  2531.     c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);

  2532.     c->chrDstW= -((-dstW) >> c->chrDstHSubSample);

  2533.     c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

  2534. 

  2535.     sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);

  2536. 

  2537.     /* unscaled special cases */

  2538.     if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat))) {

  2539.         /* yv12_to_nv12 */

  2540.         if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {

  2541.             c->swScale= PlanarToNV12Wrapper;

  2542.         }

  2543.         /* yuv2bgr */

  2544.         if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))

  2545.             && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {

  2546.             c->swScale= ff_yuv2rgb_get_func_ptr(c);

  2547.         }

  2548. 

  2549.         if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {

  2550.             c->swScale= yvu9toyv12Wrapper;

  2551.         }

  2552. 

  2553.         /* bgr24toYV12 */

  2554.         if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))

  2555.             c->swScale= bgr24toyv12Wrapper;

  2556. 

  2557.         /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  2558.         if (  (isBGR(srcFormat) || isRGB(srcFormat))

  2559.            && (isBGR(dstFormat) || isRGB(dstFormat))

  2560.            && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  2561.            && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  2562.            && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  2563.            && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  2564.            && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  2565.            && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  2566.            && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  2567.            && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE

  2568.                                              && dstFormat != PIX_FMT_RGB32_1

  2569.                                              && dstFormat != PIX_FMT_BGR32_1

  2570.            && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE

  2571.            && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE

  2572.            && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2573.              c->swScale= rgb2rgbWrapper;

  2574. 

  2575.         if ((usePal(srcFormat) && (

  2576.                  dstFormat == PIX_FMT_RGB32   ||

  2577.                  dstFormat == PIX_FMT_RGB32_1 ||

  2578.                  dstFormat == PIX_FMT_RGB24   ||

  2579.                  dstFormat == PIX_FMT_BGR32   ||

  2580.                  dstFormat == PIX_FMT_BGR32_1 ||

  2581.                  dstFormat == PIX_FMT_BGR24)))

  2582.              c->swScale= pal2rgbWrapper;

  2583. 

  2584.         if (srcFormat == PIX_FMT_YUV422P) {

  2585.             if (dstFormat == PIX_FMT_YUYV422)

  2586.                 c->swScale= YUV422PToYuy2Wrapper;

  2587.             else if (dstFormat == PIX_FMT_UYVY422)

  2588.                 c->swScale= YUV422PToUyvyWrapper;

  2589.         }

  2590. 

  2591.         /* LQ converters if -sws 0 or -sws 4*/

  2592.         if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {

  2593.             /* yv12_to_yuy2 */

  2594.             if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {

  2595.                 if (dstFormat == PIX_FMT_YUYV422)

  2596.                     c->swScale= PlanarToYuy2Wrapper;

  2597.                 else if (dstFormat == PIX_FMT_UYVY422)

  2598.                     c->swScale= PlanarToUyvyWrapper;

  2599.             }

  2600.         }

  2601.         if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  2602.             c->swScale= YUYV2YUV420Wrapper;

  2603.         if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  2604.             c->swScale= UYVY2YUV420Wrapper;

  2605.         if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)

  2606.             c->swScale= YUYV2YUV422Wrapper;

  2607.         if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)

  2608.             c->swScale= UYVY2YUV422Wrapper;

  2609. 

  2610. #ifdef COMPILE_ALTIVEC

  2611.         if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  2612.             !(c->flags & SWS_BITEXACT) &&

  2613.             srcFormat == PIX_FMT_YUV420P) {

  2614.           // unscaled YV12 -> packed YUV, we want speed

  2615.           if (dstFormat == PIX_FMT_YUYV422)

  2616.               c->swScale= yv12toyuy2_unscaled_altivec;

  2617.           else if (dstFormat == PIX_FMT_UYVY422)

  2618.               c->swScale= yv12touyvy_unscaled_altivec;

  2619.         }

  2620. #endif

  2621. 

  2622.         /* simple copy */

  2623.         if (  srcFormat == dstFormat

  2624.             || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)

  2625.             || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)

  2626.             || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  2627.             || (isPlanarYUV(dstFormat) && isGray(srcFormat))

  2628.             || (isGray(dstFormat) && isGray(srcFormat))

  2629.             || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)

  2630.                 && c->chrDstHSubSample == c->chrSrcHSubSample

  2631.                 && c->chrDstVSubSample == c->chrSrcVSubSample

  2632.                 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21

  2633.                 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))

  2634.         {

  2635.             if (isPacked(c->srcFormat))

  2636.                 c->swScale= packedCopy;

  2637.             else /* Planar YUV or gray */

  2638.                 c->swScale= planarCopy;

  2639.         }

  2640. #if ARCH_BFIN

  2641.         if (flags & SWS_CPU_CAPS_BFIN)

  2642.             ff_bfin_get_unscaled_swscale (c);

  2643. #endif

  2644. 

  2645.         if (c->swScale) {

  2646.             if (flags&SWS_PRINT_INFO)

  2647.                 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",

  2648.                        sws_format_name(srcFormat), sws_format_name(dstFormat));

  2649.             return c;

  2650.         }

  2651.     }

  2652. 

  2653.     if (flags & SWS_CPU_CAPS_MMX2) {

  2654.         c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;

  2655.         if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {

  2656.             if (flags&SWS_PRINT_INFO)

  2657.                 av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");

  2658.         }

  2659.         if (usesHFilter) c->canMMX2BeUsed=0;

  2660.     }

  2661.     else

  2662.         c->canMMX2BeUsed=0;

  2663. 

  2664.     c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;

  2665.     c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

  2666. 

  2667.     // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst

  2668.     // but only for the FAST_BILINEAR mode otherwise do correct scaling

  2669.     // n-2 is the last chrominance sample available

  2670.     // this is not perfect, but no one should notice the difference, the more correct variant

  2671.     // would be like the vertical one, but that would require some special code for the

  2672.     // first and last pixel

  2673.     if (flags&SWS_FAST_BILINEAR) {

  2674.         if (c->canMMX2BeUsed) {

  2675.             c->lumXInc+= 20;

  2676.             c->chrXInc+= 20;

  2677.         }

  2678.         //we don't use the x86 asm scaler if MMX is available

  2679.         else if (flags & SWS_CPU_CAPS_MMX) {

  2680.             c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;

  2681.             c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;

  2682.         }

  2683.     }

  2684. 

  2685.     /* precalculate horizontal scaler filter coefficients */

  2686.     {

  2687.         const int filterAlign=

  2688.             (flags & SWS_CPU_CAPS_MMX) ? 4 :

  2689.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2690.             1;

  2691. 

  2692.         if (initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,

  2693.                        srcW      ,       dstW, filterAlign, 1<<14,

  2694.                        (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2695.                        srcFilter->lumH, dstFilter->lumH, c->param) < 0)

  2696.             goto fail;

  2697.         if (initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,

  2698.                        c->chrSrcW, c->chrDstW, filterAlign, 1<<14,

  2699.                        (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2700.                        srcFilter->chrH, dstFilter->chrH, c->param) < 0)

  2701.             goto fail;

  2702. 

  2703. #if defined(COMPILE_MMX2)

  2704. // can't downscale !!!

  2705.         if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {

  2706.             c->lumMmx2FilterCodeSize = initMMX2HScaler(      dstW, c->lumXInc, NULL, NULL, NULL, 8);

  2707.             c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);

  2708. 

  2709. #ifdef MAP_ANONYMOUS

  2710.             c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2711.             c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2712. #elif HAVE_VIRTUALALLOC

  2713.             c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);

  2714.             c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);

  2715. #else

  2716.             c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize);

  2717.             c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);

  2718. #endif

  2719. 

  2720.             FF_ALLOCZ_OR_GOTO(c, c->lumMmx2Filter   , (dstW        /8+8)*sizeof(int16_t), fail);

  2721.             FF_ALLOCZ_OR_GOTO(c, c->chrMmx2Filter   , (c->chrDstW  /4+8)*sizeof(int16_t), fail);

  2722.             FF_ALLOCZ_OR_GOTO(c, c->lumMmx2FilterPos, (dstW      /2/8+8)*sizeof(int32_t), fail);

  2723.             FF_ALLOCZ_OR_GOTO(c, c->chrMmx2FilterPos, (c->chrDstW/2/4+8)*sizeof(int32_t), fail);

  2724. 

  2725.             initMMX2HScaler(      dstW, c->lumXInc, c->lumMmx2FilterCode, c->lumMmx2Filter, c->lumMmx2FilterPos, 8);

  2726.             initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);

  2727. 

  2728. #ifdef MAP_ANONYMOUS

  2729.             mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);

  2730.             mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);

  2731. #endif

  2732.         }

  2733. #endif /* defined(COMPILE_MMX2) */

  2734.     } // initialize horizontal stuff

  2735. 

  2736. 

  2737. 

  2738.     /* precalculate vertical scaler filter coefficients */

  2739.     {

  2740.         const int filterAlign=

  2741.             (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :

  2742.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2743.             1;

  2744. 

  2745.         if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,

  2746.                        srcH      ,        dstH, filterAlign, (1<<12),

  2747.                        (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2748.                        srcFilter->lumV, dstFilter->lumV, c->param) < 0)

  2749.             goto fail;

  2750.         if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,

  2751.                        c->chrSrcH, c->chrDstH, filterAlign, (1<<12),

  2752.                        (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2753.                        srcFilter->chrV, dstFilter->chrV, c->param) < 0)

  2754.             goto fail;

  2755. 

  2756. #ifdef COMPILE_ALTIVEC

  2757.         FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof (vector signed short)*c->vLumFilterSize*c->dstH, fail);

  2758.         FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH, fail);

  2759. 

  2760.         for (i=0;i<c->vLumFilterSize*c->dstH;i++) {

  2761.             int j;

  2762.             short *p = (short *)&c->vYCoeffsBank[i];

  2763.             for (j=0;j<8;j++)

  2764.                 p[j] = c->vLumFilter[i];

  2765.         }

  2766. 

  2767.         for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {

  2768.             int j;

  2769.             short *p = (short *)&c->vCCoeffsBank[i];

  2770.             for (j=0;j<8;j++)

  2771.                 p[j] = c->vChrFilter[i];

  2772.         }

  2773. #endif

  2774.     }

  2775. 

  2776.     // calculate buffer sizes so that they won't run out while handling these damn slices

  2777.     c->vLumBufSize= c->vLumFilterSize;

  2778.     c->vChrBufSize= c->vChrFilterSize;

  2779.     for (i=0; i<dstH; i++) {

  2780.         int chrI= i*c->chrDstH / dstH;

  2781.         int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,

  2782.                            ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

  2783. 

  2784.         nextSlice>>= c->chrSrcVSubSample;

  2785.         nextSlice<<= c->chrSrcVSubSample;

  2786.         if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)

  2787.             c->vLumBufSize= nextSlice - c->vLumFilterPos[i];

  2788.         if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))

  2789.             c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];

  2790.     }

  2791. 

  2792.     // allocate pixbufs (we use dynamic allocation because otherwise we would need to

  2793.     // allocate several megabytes to handle all possible cases)

  2794.     FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);

  2795.     FF_ALLOC_OR_GOTO(c, c->chrPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail);

  2796.     if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))

  2797.         FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);

  2798.     //Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna replace the 4000/8000)

  2799.     /* align at 16 bytes for AltiVec */

  2800.     for (i=0; i<c->vLumBufSize; i++) {

  2801.         FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i+c->vLumBufSize], VOF+1, fail);

  2802.         c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize];

  2803.     }

  2804.     for (i=0; i<c->vChrBufSize; i++) {

  2805.         FF_ALLOC_OR_GOTO(c, c->chrPixBuf[i+c->vChrBufSize], (VOF+1)*2, fail);

  2806.         c->chrPixBuf[i] = c->chrPixBuf[i+c->vChrBufSize];

  2807.     }

  2808.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)

  2809.         for (i=0; i<c->vLumBufSize; i++) {

  2810.             FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i+c->vLumBufSize], VOF+1, fail);

  2811.             c->alpPixBuf[i] = c->alpPixBuf[i+c->vLumBufSize];

  2812.         }

  2813. 

  2814.     //try to avoid drawing green stuff between the right end and the stride end

  2815.     for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);

  2816. 

  2817.     assert(2*VOFW == VOF);

  2818. 

  2819.     assert(c->chrDstH <= dstH);

  2820. 

  2821.     if (flags&SWS_PRINT_INFO) {

  2822. #ifdef DITHER1XBPP

  2823.         const char *dither= " dithered";

  2824. #else

  2825.         const char *dither= "";

  2826. #endif

  2827.         if (flags&SWS_FAST_BILINEAR)

  2828.             av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");

  2829.         else if (flags&SWS_BILINEAR)

  2830.             av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");

  2831.         else if (flags&SWS_BICUBIC)

  2832.             av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");

  2833.         else if (flags&SWS_X)

  2834.             av_log(c, AV_LOG_INFO, "Experimental scaler, ");

  2835.         else if (flags&SWS_POINT)

  2836.             av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");

  2837.         else if (flags&SWS_AREA)

  2838.             av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");

  2839.         else if (flags&SWS_BICUBLIN)

  2840.             av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");

  2841.         else if (flags&SWS_GAUSS)

  2842.             av_log(c, AV_LOG_INFO, "Gaussian scaler, ");

  2843.         else if (flags&SWS_SINC)

  2844.             av_log(c, AV_LOG_INFO, "Sinc scaler, ");

  2845.         else if (flags&SWS_LANCZOS)

  2846.             av_log(c, AV_LOG_INFO, "Lanczos scaler, ");

  2847.         else if (flags&SWS_SPLINE)

  2848.             av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");

  2849.         else

  2850.             av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");

  2851. 

  2852.         if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)

  2853.             av_log(c, AV_LOG_INFO, "from %s to%s %s ",

  2854.                    sws_format_name(srcFormat), dither, sws_format_name(dstFormat));

  2855.         else

  2856.             av_log(c, AV_LOG_INFO, "from %s to %s ",

  2857.                    sws_format_name(srcFormat), sws_format_name(dstFormat));

  2858. 

  2859.         if (flags & SWS_CPU_CAPS_MMX2)

  2860.             av_log(c, AV_LOG_INFO, "using MMX2\n");

  2861.         else if (flags & SWS_CPU_CAPS_3DNOW)

  2862.             av_log(c, AV_LOG_INFO, "using 3DNOW\n");

  2863.         else if (flags & SWS_CPU_CAPS_MMX)

  2864.             av_log(c, AV_LOG_INFO, "using MMX\n");

  2865.         else if (flags & SWS_CPU_CAPS_ALTIVEC)

  2866.             av_log(c, AV_LOG_INFO, "using AltiVec\n");

  2867.         else

  2868.             av_log(c, AV_LOG_INFO, "using C\n");

  2869.     }

  2870. 

  2871.     if (flags & SWS_PRINT_INFO) {

  2872.         if (flags & SWS_CPU_CAPS_MMX) {

  2873.             if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))

  2874.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");

  2875.             else {

  2876.                 if (c->hLumFilterSize==4)

  2877.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");

  2878.                 else if (c->hLumFilterSize==8)

  2879.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");

  2880.                 else

  2881.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");

  2882. 

  2883.                 if (c->hChrFilterSize==4)

  2884.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");

  2885.                 else if (c->hChrFilterSize==8)

  2886.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");

  2887.                 else

  2888.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");

  2889.             }

  2890.         } else {

  2891. #if ARCH_X86

  2892.             av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");

  2893. #else

  2894.             if (flags & SWS_FAST_BILINEAR)

  2895.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");

  2896.             else

  2897.                 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");

  2898. #endif

  2899.         }

  2900.         if (isPlanarYUV(dstFormat)) {

  2901.             if (c->vLumFilterSize==1)

  2902.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2903.             else

  2904.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2905.         } else {

  2906.             if (c->vLumFilterSize==1 && c->vChrFilterSize==2)

  2907.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"

  2908.                        "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2909.             else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)

  2910.                 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2911.             else

  2912.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2913.         }

  2914. 

  2915.         if (dstFormat==PIX_FMT_BGR24)

  2916.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",

  2917.                    (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));

  2918.         else if (dstFormat==PIX_FMT_RGB32)

  2919.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2920.         else if (dstFormat==PIX_FMT_BGR565)

  2921.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2922.         else if (dstFormat==PIX_FMT_BGR555)

  2923.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2924. 

  2925.         av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);

  2926.     }

  2927.     if (flags & SWS_PRINT_INFO) {

  2928.         av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2929.                c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);

  2930.         av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2931.                c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);

  2932.     }

  2933. 

  2934.     c->swScale= getSwsFunc(c);

  2935.     return c;

  2936. 

  2937. fail:

  2938.     sws_freeContext(c);

  2939.     return NULL;

  2940. }

  2941. 

  2942. static void reset_ptr(uint8_t* src[], int format)

  2943. {

  2944.     if(!isALPHA(format))

  2945.         src[3]=NULL;

  2946.     if(!isPlanarYUV(format)) {

  2947.         src[3]=src[2]=NULL;

  2948.         if(   format != PIX_FMT_PAL8

  2949.            && format != PIX_FMT_RGB8

  2950.            && format != PIX_FMT_BGR8

  2951.            && format != PIX_FMT_RGB4_BYTE

  2952.            && format != PIX_FMT_BGR4_BYTE

  2953.           )

  2954.             src[1]= NULL;

  2955.     }

  2956. }

  2957. 

  2958. /**

  2959.  * swscale wrapper, so we don't need to export the SwsContext.

  2960.  * Assumes planar YUV to be in YUV order instead of YVU.

  2961.  */

  2962. int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2963.               int srcSliceH, uint8_t* dst[], int dstStride[])

  2964. {

  2965.     int i;

  2966.     uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};

  2967.     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};

  2968. 

  2969.     // do not mess up sliceDir if we have a "trailing" 0-size slice

  2970.     if (srcSliceH == 0)

  2971.         return 0;

  2972. 

  2973.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  2974.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  2975.         return 0;

  2976.     }

  2977.     if (c->sliceDir == 0) {

  2978.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  2979.     }

  2980. 

  2981.     if (usePal(c->srcFormat)) {

  2982.         for (i=0; i<256; i++) {

  2983.             int p, r, g, b,y,u,v;

  2984.             if(c->srcFormat == PIX_FMT_PAL8) {

  2985.                 p=((uint32_t*)(src[1]))[i];

  2986.                 r= (p>>16)&0xFF;

  2987.                 g= (p>> 8)&0xFF;

  2988.                 b=  p     &0xFF;

  2989.             } else if(c->srcFormat == PIX_FMT_RGB8) {

  2990.                 r= (i>>5    )*36;

  2991.                 g= ((i>>2)&7)*36;

  2992.                 b= (i&3     )*85;

  2993.             } else if(c->srcFormat == PIX_FMT_BGR8) {

  2994.                 b= (i>>6    )*85;

  2995.                 g= ((i>>3)&7)*36;

  2996.                 r= (i&7     )*36;

  2997.             } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {

  2998.                 r= (i>>3    )*255;

  2999.                 g= ((i>>1)&3)*85;

  3000.                 b= (i&1     )*255;

  3001.             } else {

  3002.                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);

  3003.                 b= (i>>3    )*255;

  3004.                 g= ((i>>1)&3)*85;

  3005.                 r= (i&1     )*255;

  3006.             }

  3007.             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3008.             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3009.             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3010.             c->pal_yuv[i]= y + (u<<8) + (v<<16);

  3011. 

  3012. 

  3013.             switch(c->dstFormat) {

  3014.             case PIX_FMT_BGR32:

  3015. #if !HAVE_BIGENDIAN

  3016.             case PIX_FMT_RGB24:

  3017. #endif

  3018.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);

  3019.                 break;

  3020.             case PIX_FMT_BGR32_1:

  3021. #if HAVE_BIGENDIAN

  3022.             case PIX_FMT_BGR24:

  3023. #endif

  3024.                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;

  3025.                 break;

  3026.             case PIX_FMT_RGB32_1:

  3027. #if HAVE_BIGENDIAN

  3028.             case PIX_FMT_RGB24:

  3029. #endif

  3030.                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;

  3031.                 break;

  3032.             case PIX_FMT_RGB32:

  3033. #if !HAVE_BIGENDIAN

  3034.             case PIX_FMT_BGR24:

  3035. #endif

  3036.             default:

  3037.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);

  3038.             }

  3039.         }

  3040.     }

  3041. 

  3042.     // copy strides, so they can safely be modified

  3043.     if (c->sliceDir == 1) {

  3044.         // slices go from top to bottom

  3045.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};

  3046.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};

  3047. 

  3048.         reset_ptr(src2, c->srcFormat);

  3049.         reset_ptr(dst2, c->dstFormat);

  3050. 

  3051.         /* reset slice direction at end of frame */

  3052.         if (srcSliceY + srcSliceH == c->srcH)

  3053.             c->sliceDir = 0;

  3054. 

  3055.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);

  3056.     } else {

  3057.         // slices go from bottom to top => we flip the image internally

  3058.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};

  3059.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};

  3060. 

  3061.         src2[0] += (srcSliceH-1)*srcStride[0];

  3062.         if (!usePal(c->srcFormat))

  3063.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  3064.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  3065.         src2[3] += (srcSliceH-1)*srcStride[3];

  3066.         dst2[0] += ( c->dstH                      -1)*dstStride[0];

  3067.         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];

  3068.         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];

  3069.         dst2[3] += ( c->dstH                      -1)*dstStride[3];

  3070. 

  3071.         reset_ptr(src2, c->srcFormat);

  3072.         reset_ptr(dst2, c->dstFormat);

  3073. 

  3074.         /* reset slice direction at end of frame */

  3075.         if (!srcSliceY)

  3076.             c->sliceDir = 0;

  3077. 

  3078.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  3079.     }

  3080. }

  3081. 

  3082. #if LIBSWSCALE_VERSION_MAJOR < 1

  3083. int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  3084.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  3085. {

  3086.     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);

  3087. }

  3088. #endif

  3089. 

  3090. SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,

  3091.                                 float lumaSharpen, float chromaSharpen,

  3092.                                 float chromaHShift, float chromaVShift,

  3093.                                 int verbose)

  3094. {

  3095.     SwsFilter *filter= av_malloc(sizeof(SwsFilter));

  3096.     if (!filter)

  3097.         return NULL;

  3098. 

  3099.     if (lumaGBlur!=0.0) {

  3100.         filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);

  3101.         filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);

  3102.     } else {

  3103.         filter->lumH= sws_getIdentityVec();

  3104.         filter->lumV= sws_getIdentityVec();

  3105.     }

  3106. 

  3107.     if (chromaGBlur!=0.0) {

  3108.         filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);

  3109.         filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);

  3110.     } else {

  3111.         filter->chrH= sws_getIdentityVec();

  3112.         filter->chrV= sws_getIdentityVec();

  3113.     }

  3114. 

  3115.     if (chromaSharpen!=0.0) {

  3116.         SwsVector *id= sws_getIdentityVec();

  3117.         sws_scaleVec(filter->chrH, -chromaSharpen);

  3118.         sws_scaleVec(filter->chrV, -chromaSharpen);

  3119.         sws_addVec(filter->chrH, id);

  3120.         sws_addVec(filter->chrV, id);

  3121.         sws_freeVec(id);

  3122.     }

  3123. 

  3124.     if (lumaSharpen!=0.0) {

  3125.         SwsVector *id= sws_getIdentityVec();

  3126.         sws_scaleVec(filter->lumH, -lumaSharpen);

  3127.         sws_scaleVec(filter->lumV, -lumaSharpen);

  3128.         sws_addVec(filter->lumH, id);

  3129.         sws_addVec(filter->lumV, id);

  3130.         sws_freeVec(id);

  3131.     }

  3132. 

  3133.     if (chromaHShift != 0.0)

  3134.         sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));

  3135. 

  3136.     if (chromaVShift != 0.0)

  3137.         sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));

  3138. 

  3139.     sws_normalizeVec(filter->chrH, 1.0);

  3140.     sws_normalizeVec(filter->chrV, 1.0);

  3141.     sws_normalizeVec(filter->lumH, 1.0);

  3142.     sws_normalizeVec(filter->lumV, 1.0);

  3143. 

  3144.     if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);

  3145.     if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);

  3146. 

  3147.     return filter;

  3148. }

  3149. 

  3150. SwsVector *sws_allocVec(int length)

  3151. {

  3152.     SwsVector *vec = av_malloc(sizeof(SwsVector));

  3153.     if (!vec)

  3154.         return NULL;

  3155.     vec->length = length;

  3156.     vec->coeff  = av_malloc(sizeof(double) * length);

  3157.     if (!vec->coeff)

  3158.         av_freep(&vec);

  3159.     return vec;

  3160. }

  3161. 

  3162. SwsVector *sws_getGaussianVec(double variance, double quality)

  3163. {

  3164.     const int length= (int)(variance*quality + 0.5) | 1;

  3165.     int i;

  3166.     double middle= (length-1)*0.5;

  3167.     SwsVector *vec= sws_allocVec(length);

  3168. 

  3169.     if (!vec)

  3170.         return NULL;

  3171. 

  3172.     for (i=0; i<length; i++) {

  3173.         double dist= i-middle;

  3174.         vec->coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);

  3175.     }

  3176. 

  3177.     sws_normalizeVec(vec, 1.0);

  3178. 

  3179.     return vec;

  3180. }

  3181. 

  3182. SwsVector *sws_getConstVec(double c, int length)

  3183. {

  3184.     int i;

  3185.     SwsVector *vec= sws_allocVec(length);

  3186. 

  3187.     if (!vec)

  3188.         return NULL;

  3189. 

  3190.     for (i=0; i<length; i++)

  3191.         vec->coeff[i]= c;

  3192. 

  3193.     return vec;

  3194. }

  3195. 

  3196. 

  3197. SwsVector *sws_getIdentityVec(void)

  3198. {

  3199.     return sws_getConstVec(1.0, 1);

  3200. }

  3201. 

  3202. double sws_dcVec(SwsVector *a)

  3203. {

  3204.     int i;

  3205.     double sum=0;

  3206. 

  3207.     for (i=0; i<a->length; i++)

  3208.         sum+= a->coeff[i];

  3209. 

  3210.     return sum;

  3211. }

  3212. 

  3213. void sws_scaleVec(SwsVector *a, double scalar)

  3214. {

  3215.     int i;

  3216. 

  3217.     for (i=0; i<a->length; i++)

  3218.         a->coeff[i]*= scalar;

  3219. }

  3220. 

  3221. void sws_normalizeVec(SwsVector *a, double height)

  3222. {

  3223.     sws_scaleVec(a, height/sws_dcVec(a));

  3224. }

  3225. 

  3226. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)

  3227. {

  3228.     int length= a->length + b->length - 1;

  3229.     int i, j;

  3230.     SwsVector *vec= sws_getConstVec(0.0, length);

  3231. 

  3232.     if (!vec)

  3233.         return NULL;

  3234. 

  3235.     for (i=0; i<a->length; i++) {

  3236.         for (j=0; j<b->length; j++) {

  3237.             vec->coeff[i+j]+= a->coeff[i]*b->coeff[j];

  3238.         }

  3239.     }

  3240. 

  3241.     return vec;

  3242. }

  3243. 

  3244. static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)

  3245. {

  3246.     int length= FFMAX(a->length, b->length);

  3247.     int i;

  3248.     SwsVector *vec= sws_getConstVec(0.0, length);

  3249. 

  3250.     if (!vec)

  3251.         return NULL;

  3252. 

  3253.     for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  3254.     for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];

  3255. 

  3256.     return vec;

  3257. }

  3258. 

  3259. static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)

  3260. {

  3261.     int length= FFMAX(a->length, b->length);

  3262.     int i;

  3263.     SwsVector *vec= sws_getConstVec(0.0, length);

  3264. 

  3265.     if (!vec)

  3266.         return NULL;

  3267. 

  3268.     for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  3269.     for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];

  3270. 

  3271.     return vec;

  3272. }

  3273. 

  3274. /* shift left / or right if "shift" is negative */

  3275. static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)

  3276. {

  3277.     int length= a->length + FFABS(shift)*2;

  3278.     int i;

  3279.     SwsVector *vec= sws_getConstVec(0.0, length);

  3280. 

  3281.     if (!vec)

  3282.         return NULL;

  3283. 

  3284.     for (i=0; i<a->length; i++) {

  3285.         vec->coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];

  3286.     }

  3287. 

  3288.     return vec;

  3289. }

  3290. 

  3291. void sws_shiftVec(SwsVector *a, int shift)

  3292. {

  3293.     SwsVector *shifted= sws_getShiftedVec(a, shift);

  3294.     av_free(a->coeff);

  3295.     a->coeff= shifted->coeff;

  3296.     a->length= shifted->length;

  3297.     av_free(shifted);

  3298. }

  3299. 

  3300. void sws_addVec(SwsVector *a, SwsVector *b)

  3301. {

  3302.     SwsVector *sum= sws_sumVec(a, b);

  3303.     av_free(a->coeff);

  3304.     a->coeff= sum->coeff;

  3305.     a->length= sum->length;

  3306.     av_free(sum);

  3307. }

  3308. 

  3309. void sws_subVec(SwsVector *a, SwsVector *b)

  3310. {

  3311.     SwsVector *diff= sws_diffVec(a, b);

  3312.     av_free(a->coeff);

  3313.     a->coeff= diff->coeff;

  3314.     a->length= diff->length;

  3315.     av_free(diff);

  3316. }

  3317. 

  3318. void sws_convVec(SwsVector *a, SwsVector *b)

  3319. {

  3320.     SwsVector *conv= sws_getConvVec(a, b);

  3321.     av_free(a->coeff);

  3322.     a->coeff= conv->coeff;

  3323.     a->length= conv->length;

  3324.     av_free(conv);

  3325. }

  3326. 

  3327. SwsVector *sws_cloneVec(SwsVector *a)

  3328. {

  3329.     int i;

  3330.     SwsVector *vec= sws_allocVec(a->length);

  3331. 

  3332.     if (!vec)

  3333.         return NULL;

  3334. 

  3335.     for (i=0; i<a->length; i++) vec->coeff[i]= a->coeff[i];

  3336. 

  3337.     return vec;

  3338. }

  3339. 

  3340. void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)

  3341. {

  3342.     int i;

  3343.     double max=0;

  3344.     double min=0;

  3345.     double range;

  3346. 

  3347.     for (i=0; i<a->length; i++)

  3348.         if (a->coeff[i]>max) max= a->coeff[i];

  3349. 

  3350.     for (i=0; i<a->length; i++)

  3351.         if (a->coeff[i]<min) min= a->coeff[i];

  3352. 

  3353.     range= max - min;

  3354. 

  3355.     for (i=0; i<a->length; i++) {

  3356.         int x= (int)((a->coeff[i]-min)*60.0/range +0.5);

  3357.         av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);

  3358.         for (;x>0; x--) av_log(log_ctx, log_level, " ");

  3359.         av_log(log_ctx, log_level, "|\n");

  3360.     }

  3361. }

  3362. 

  3363. #if LIBSWSCALE_VERSION_MAJOR < 1

  3364. void sws_printVec(SwsVector *a)

  3365. {

  3366.     sws_printVec2(a, NULL, AV_LOG_DEBUG);

  3367. }

  3368. #endif

  3369. 

  3370. void sws_freeVec(SwsVector *a)

  3371. {

  3372.     if (!a) return;

  3373.     av_freep(&a->coeff);

  3374.     a->length=0;

  3375.     av_free(a);

  3376. }

  3377. 

  3378. void sws_freeFilter(SwsFilter *filter)

  3379. {

  3380.     if (!filter) return;

  3381. 

  3382.     if (filter->lumH) sws_freeVec(filter->lumH);

  3383.     if (filter->lumV) sws_freeVec(filter->lumV);

  3384.     if (filter->chrH) sws_freeVec(filter->chrH);

  3385.     if (filter->chrV) sws_freeVec(filter->chrV);

  3386.     av_free(filter);

  3387. }

  3388. 

  3389. 

  3390. void sws_freeContext(SwsContext *c)

  3391. {

  3392.     int i;

  3393.     if (!c) return;

  3394. 

  3395.     if (c->lumPixBuf) {

  3396.         for (i=0; i<c->vLumBufSize; i++)

  3397.             av_freep(&c->lumPixBuf[i]);

  3398.         av_freep(&c->lumPixBuf);

  3399.     }

  3400. 

  3401.     if (c->chrPixBuf) {

  3402.         for (i=0; i<c->vChrBufSize; i++)

  3403.             av_freep(&c->chrPixBuf[i]);

  3404.         av_freep(&c->chrPixBuf);

  3405.     }

  3406. 

  3407.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  3408.         for (i=0; i<c->vLumBufSize; i++)

  3409.             av_freep(&c->alpPixBuf[i]);

  3410.         av_freep(&c->alpPixBuf);

  3411.     }

  3412. 

  3413.     av_freep(&c->vLumFilter);

  3414.     av_freep(&c->vChrFilter);

  3415.     av_freep(&c->hLumFilter);

  3416.     av_freep(&c->hChrFilter);

  3417. #ifdef COMPILE_ALTIVEC

  3418.     av_freep(&c->vYCoeffsBank);

  3419.     av_freep(&c->vCCoeffsBank);

  3420. #endif

  3421. 

  3422.     av_freep(&c->vLumFilterPos);

  3423.     av_freep(&c->vChrFilterPos);

  3424.     av_freep(&c->hLumFilterPos);

  3425.     av_freep(&c->hChrFilterPos);

  3426. 

  3427. #if ARCH_X86 && CONFIG_GPL

  3428. #ifdef MAP_ANONYMOUS

  3429.     if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);

  3430.     if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);

  3431. #elif HAVE_VIRTUALALLOC

  3432.     if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, MEM_RELEASE);

  3433.     if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, MEM_RELEASE);

  3434. #else

  3435.     av_free(c->lumMmx2FilterCode);

  3436.     av_free(c->chrMmx2FilterCode);

  3437. #endif

  3438.     c->lumMmx2FilterCode=NULL;

  3439.     c->chrMmx2FilterCode=NULL;

  3440. #endif /* ARCH_X86 && CONFIG_GPL */

  3441. 

  3442.     av_freep(&c->lumMmx2Filter);

  3443.     av_freep(&c->chrMmx2Filter);

  3444.     av_freep(&c->lumMmx2FilterPos);

  3445.     av_freep(&c->chrMmx2FilterPos);

  3446.     av_freep(&c->yuvTable);

  3447. 

  3448.     av_free(c);

  3449. }

  3450. 

  3451. struct SwsContext *sws_getCachedContext(struct SwsContext *context,

  3452.                                         int srcW, int srcH, enum PixelFormat srcFormat,

  3453.                                         int dstW, int dstH, enum PixelFormat dstFormat, int flags,

  3454.                                         SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)

  3455. {

  3456.     static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};

  3457. 

  3458.     if (!param)

  3459.         param = default_param;

  3460. 

  3461.     if (context) {

  3462.         if (context->srcW != srcW || context->srcH != srcH ||

  3463.             context->srcFormat != srcFormat ||

  3464.             context->dstW != dstW || context->dstH != dstH ||

  3465.             context->dstFormat != dstFormat || context->flags != flags ||

  3466.             context->param[0] != param[0] || context->param[1] != param[1])

  3467.         {

  3468.             sws_freeContext(context);

  3469.             context = NULL;

  3470.         }

  3471.     }

  3472.     if (!context) {

  3473.         return sws_getContext(srcW, srcH, srcFormat,

  3474.                               dstW, dstH, dstFormat, flags,

  3475.                               srcFilter, dstFilter, param);

  3476.     }

  3477.     return context;

  3478. }