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.
7763 Commits
32 Branches
484MB
Tree: 628bc5e862
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '628bc5e862'
${ noResults }
FFmpeg/libswscale/swscale.c

2862 lines
81KB
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 St, 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, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09

  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 didnt 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 (its 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. #include <inttypes.h>

  58. #include <string.h>

  59. #include <math.h>

  60. #include <stdio.h>

  61. #include <unistd.h>

  62. #include "config.h"

  63. #include <assert.h>

  64. #ifdef 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. #include "swscale.h"

  71. #include "swscale_internal.h"

  72. #include "x86_cpu.h"

  73. #include "bswap.h"

  74. #include "rgb2rgb.h"

  75. #ifdef USE_FASTMEMCPY

  76. #include "libvo/fastmemcpy.h"

  77. #endif

  78. 

  79. #undef MOVNTQ

  80. #undef PAVGB

  81. 

  82. //#undef HAVE_MMX2

  83. //#define HAVE_3DNOW

  84. //#undef HAVE_MMX

  85. //#undef ARCH_X86

  86. //#define WORDS_BIGENDIAN

  87. #define DITHER1XBPP

  88. 

  89. #define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit

  90. 

  91. #define RET 0xC3 //near return opcode for X86

  92. 

  93. #ifdef MP_DEBUG

  94. #define ASSERT(x) assert(x);

  95. #else

  96. #define ASSERT(x) ;

  97. #endif

  98. 

  99. #ifdef M_PI

  100. #define PI M_PI

  101. #else

  102. #define PI 3.14159265358979323846

  103. #endif

  104. 

  105. #define isSupportedIn(x)  ((x)==PIX_FMT_YUV420P || (x)==PIX_FMT_YUYV422 || (x)==PIX_FMT_UYVY422\

  106. 			|| (x)==PIX_FMT_RGB32|| (x)==PIX_FMT_BGR24|| (x)==PIX_FMT_BGR565|| (x)==PIX_FMT_BGR555\

  107. 			|| (x)==PIX_FMT_BGR32|| (x)==PIX_FMT_RGB24|| (x)==PIX_FMT_RGB565|| (x)==PIX_FMT_RGB555\

  108. 			|| (x)==PIX_FMT_GRAY8 || (x)==PIX_FMT_YUV410P\

  109. 			|| (x)==PIX_FMT_GRAY16BE || (x)==PIX_FMT_GRAY16LE\

  110. 			|| (x)==PIX_FMT_YUV444P || (x)==PIX_FMT_YUV422P || (x)==PIX_FMT_YUV411P)

  111. #define isSupportedOut(x) ((x)==PIX_FMT_YUV420P || (x)==PIX_FMT_YUYV422 || (x)==PIX_FMT_UYVY422\

  112. 			|| (x)==PIX_FMT_YUV444P || (x)==PIX_FMT_YUV422P || (x)==PIX_FMT_YUV411P\

  113. 			|| isRGB(x) || isBGR(x)\

  114. 			|| (x)==PIX_FMT_NV12 || (x)==PIX_FMT_NV21\

  115. 			|| (x)==PIX_FMT_GRAY16BE || (x)==PIX_FMT_GRAY16LE\

  116. 			|| (x)==PIX_FMT_GRAY8 || (x)==PIX_FMT_YUV410P)

  117. #define isPacked(x)    ((x)==PIX_FMT_YUYV422 || (x)==PIX_FMT_UYVY422 ||isRGB(x) || isBGR(x))

  118. 

  119. #define RGB2YUV_SHIFT 16

  120. #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))

  121. #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))

  122. #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))

  123. #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))

  124. #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))

  125. #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))

  126. #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))

  127. #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))

  128. #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))

  129. 

  130. extern const int32_t Inverse_Table_6_9[8][4];

  131. 

  132. /*

  133. NOTES

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

  135. 

  136. TODO

  137. more intelligent missalignment avoidance for the horizontal scaler

  138. write special vertical cubic upscale version

  139. Optimize C code (yv12 / minmax)

  140. add support for packed pixel yuv input & output

  141. add support for Y8 output

  142. optimize bgr24 & bgr32

  143. add BGR4 output support

  144. write special BGR->BGR scaler

  145. */

  146. 

  147. #if defined(ARCH_X86) && defined (CONFIG_GPL)

  148. static uint64_t attribute_used __attribute__((aligned(8))) bF8=       0xF8F8F8F8F8F8F8F8LL;

  149. static uint64_t attribute_used __attribute__((aligned(8))) bFC=       0xFCFCFCFCFCFCFCFCLL;

  150. static uint64_t __attribute__((aligned(8))) w10=       0x0010001000100010LL;

  151. static uint64_t attribute_used __attribute__((aligned(8))) w02=       0x0002000200020002LL;

  152. static uint64_t attribute_used __attribute__((aligned(8))) bm00001111=0x00000000FFFFFFFFLL;

  153. static uint64_t attribute_used __attribute__((aligned(8))) bm00000111=0x0000000000FFFFFFLL;

  154. static uint64_t attribute_used __attribute__((aligned(8))) bm11111000=0xFFFFFFFFFF000000LL;

  155. static uint64_t attribute_used __attribute__((aligned(8))) bm01010101=0x00FF00FF00FF00FFLL;

  156. 

  157. static volatile uint64_t attribute_used __attribute__((aligned(8))) b5Dither;

  158. static volatile uint64_t attribute_used __attribute__((aligned(8))) g5Dither;

  159. static volatile uint64_t attribute_used __attribute__((aligned(8))) g6Dither;

  160. static volatile uint64_t attribute_used __attribute__((aligned(8))) r5Dither;

  161. 

  162. static uint64_t __attribute__((aligned(8))) dither4[2]={

  163. 	0x0103010301030103LL,

  164. 	0x0200020002000200LL,};

  165. 

  166. static uint64_t __attribute__((aligned(8))) dither8[2]={

  167. 	0x0602060206020602LL,

  168. 	0x0004000400040004LL,};

  169. 

  170. static uint64_t __attribute__((aligned(8))) b16Mask=   0x001F001F001F001FLL;

  171. static uint64_t attribute_used __attribute__((aligned(8))) g16Mask=   0x07E007E007E007E0LL;

  172. static uint64_t attribute_used __attribute__((aligned(8))) r16Mask=   0xF800F800F800F800LL;

  173. static uint64_t __attribute__((aligned(8))) b15Mask=   0x001F001F001F001FLL;

  174. static uint64_t attribute_used __attribute__((aligned(8))) g15Mask=   0x03E003E003E003E0LL;

  175. static uint64_t attribute_used __attribute__((aligned(8))) r15Mask=   0x7C007C007C007C00LL;

  176. 

  177. static uint64_t attribute_used __attribute__((aligned(8))) M24A=   0x00FF0000FF0000FFLL;

  178. static uint64_t attribute_used __attribute__((aligned(8))) M24B=   0xFF0000FF0000FF00LL;

  179. static uint64_t attribute_used __attribute__((aligned(8))) M24C=   0x0000FF0000FF0000LL;

  180. 

  181. #ifdef FAST_BGR2YV12

  182. static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;

  183. static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;

  184. static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;

  185. #else

  186. static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;

  187. static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;

  188. static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;

  189. #endif /* FAST_BGR2YV12 */

  190. static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;

  191. static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL;

  192. static const uint64_t w1111       attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;

  193. #endif /* defined(ARCH_X86) */

  194. 

  195. // clipping helper table for C implementations:

  196. static unsigned char clip_table[768];

  197. 

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

  199. 		  

  200. extern const uint8_t dither_2x2_4[2][8];

  201. extern const uint8_t dither_2x2_8[2][8];

  202. extern const uint8_t dither_8x8_32[8][8];

  203. extern const uint8_t dither_8x8_73[8][8];

  204. extern const uint8_t dither_8x8_220[8][8];

  205. 

  206. char *sws_format_name(enum PixelFormat format)

  207. {

  208.     switch (format) {

  209.         case PIX_FMT_YUV420P:

  210.             return "yuv420p";

  211.         case PIX_FMT_YUYV422:

  212.             return "yuyv422";

  213.         case PIX_FMT_RGB24:

  214.             return "rgb24";

  215.         case PIX_FMT_BGR24:

  216.             return "bgr24";

  217.         case PIX_FMT_YUV422P:

  218.             return "yuv422p";

  219.         case PIX_FMT_YUV444P:

  220.             return "yuv444p";

  221.         case PIX_FMT_RGB32:

  222.             return "rgb32";

  223.         case PIX_FMT_YUV410P:

  224.             return "yuv410p";

  225.         case PIX_FMT_YUV411P:

  226.             return "yuv411p";

  227.         case PIX_FMT_RGB565:

  228.             return "rgb565";

  229.         case PIX_FMT_RGB555:

  230.             return "rgb555";

  231.         case PIX_FMT_GRAY16BE:

  232.             return "gray16be";

  233.         case PIX_FMT_GRAY16LE:

  234.             return "gray16le";

  235.         case PIX_FMT_GRAY8:

  236.             return "gray8";

  237.         case PIX_FMT_MONOWHITE:

  238.             return "mono white";

  239.         case PIX_FMT_MONOBLACK:

  240.             return "mono black";

  241.         case PIX_FMT_PAL8:

  242.             return "Palette";

  243.         case PIX_FMT_YUVJ420P:

  244.             return "yuvj420p";

  245.         case PIX_FMT_YUVJ422P:

  246.             return "yuvj422p";

  247.         case PIX_FMT_YUVJ444P:

  248.             return "yuvj444p";

  249.         case PIX_FMT_XVMC_MPEG2_MC:

  250.             return "xvmc_mpeg2_mc";

  251.         case PIX_FMT_XVMC_MPEG2_IDCT:

  252.             return "xvmc_mpeg2_idct";

  253.         case PIX_FMT_UYVY422:

  254.             return "uyvy422";

  255.         case PIX_FMT_UYYVYY411:

  256.             return "uyyvyy411";

  257.         case PIX_FMT_RGB32_1:

  258.             return "rgb32x";

  259.         case PIX_FMT_BGR32_1:

  260.             return "bgr32x";

  261.         case PIX_FMT_BGR32:

  262.             return "bgr32";

  263.         case PIX_FMT_BGR565:

  264.             return "bgr565";

  265.         case PIX_FMT_BGR555:

  266.             return "bgr555";

  267.         case PIX_FMT_BGR8:

  268.             return "bgr8";

  269.         case PIX_FMT_BGR4:

  270.             return "bgr4";

  271.         case PIX_FMT_BGR4_BYTE:

  272.             return "bgr4 byte";

  273.         case PIX_FMT_RGB8:

  274.             return "rgb8";

  275.         case PIX_FMT_RGB4:

  276.             return "rgb4";

  277.         case PIX_FMT_RGB4_BYTE:

  278.             return "rgb4 byte";

  279.         case PIX_FMT_NV12:

  280.             return "nv12";

  281.         case PIX_FMT_NV21:

  282.             return "nv21";

  283.         default:

  284.             return "Unknown format";

  285.     }

  286. }

  287. 

  288. #if defined(ARCH_X86) && defined (CONFIG_GPL)

  289. void in_asm_used_var_warning_killer()

  290. {

  291.  volatile int i= bF8+bFC+w10+

  292.  bm00001111+bm00000111+bm11111000+b16Mask+g16Mask+r16Mask+b15Mask+g15Mask+r15Mask+

  293.  M24A+M24B+M24C+w02 + b5Dither+g5Dither+r5Dither+g6Dither+dither4[0]+dither8[0]+bm01010101;

  294.  if(i) i=0;

  295. }

  296. #endif

  297. 

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

  299. 				    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,

  300. 				    uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW)

  301. {

  302. 	//FIXME Optimize (just quickly writen not opti..)

  303. 	int i;

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

  305. 	{

  306. 		int val=1<<18;

  307. 		int j;

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

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

  310. 

  311. 		dest[i]= FFMIN(FFMAX(val>>19, 0), 255);

  312. 	}

  313. 

  314. 	if(uDest != NULL)

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

  316. 		{

  317. 			int u=1<<18;

  318. 			int v=1<<18;

  319. 			int j;

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

  321. 			{

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

  323. 				v += chrSrc[j][i + 2048] * chrFilter[j];

  324. 			}

  325. 

  326. 			uDest[i]= FFMIN(FFMAX(u>>19, 0), 255);

  327. 			vDest[i]= FFMIN(FFMAX(v>>19, 0), 255);

  328. 		}

  329. }

  330. 

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

  332. 				int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,

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

  334. {

  335. 	//FIXME Optimize (just quickly writen not opti..)

  336. 	int i;

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

  338. 	{

  339. 		int val=1<<18;

  340. 		int j;

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

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

  343. 

  344. 		dest[i]= FFMIN(FFMAX(val>>19, 0), 255);

  345. 	}

  346. 

  347. 	if(uDest == NULL)

  348. 		return;

  349. 

  350. 	if(dstFormat == PIX_FMT_NV12)

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

  352. 		{

  353. 			int u=1<<18;

  354. 			int v=1<<18;

  355. 			int j;

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

  357. 			{

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

  359. 				v += chrSrc[j][i + 2048] * chrFilter[j];

  360. 			}

  361. 

  362. 			uDest[2*i]= FFMIN(FFMAX(u>>19, 0), 255);

  363. 			uDest[2*i+1]= FFMIN(FFMAX(v>>19, 0), 255);

  364. 		}

  365. 	else

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

  367. 		{

  368. 			int u=1<<18;

  369. 			int v=1<<18;

  370. 			int j;

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

  372. 			{

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

  374. 				v += chrSrc[j][i + 2048] * chrFilter[j];

  375. 			}

  376. 

  377. 			uDest[2*i]= FFMIN(FFMAX(v>>19, 0), 255);

  378. 			uDest[2*i+1]= FFMIN(FFMAX(u>>19, 0), 255);

  379. 		}

  380. }

  381. 

  382. #define YSCALE_YUV_2_PACKEDX_C(type) \

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

  384. 			int j;\

  385. 			int Y1=1<<18;\

  386. 			int Y2=1<<18;\

  387. 			int U=1<<18;\

  388. 			int V=1<<18;\

  389. 			type *r, *b, *g;\

  390. 			const int i2= 2*i;\

  391. 			\

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

  393. 			{\

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

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

  396. 			}\

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

  398. 			{\

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

  400. 				V += chrSrc[j][i+2048] * chrFilter[j];\

  401. 			}\

  402. 			Y1>>=19;\

  403. 			Y2>>=19;\

  404. 			U >>=19;\

  405. 			V >>=19;\

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

  407. 			{\

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

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

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

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

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

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

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

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

  416. 			}

  417.                         

  418. #define YSCALE_YUV_2_RGBX_C(type) \

  419. 			YSCALE_YUV_2_PACKEDX_C(type)\

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

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

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

  423. 

  424. #define YSCALE_YUV_2_PACKED2_C \

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

  426. 			const int i2= 2*i;\

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

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

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

  430. 			int V= (uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19;\

  431. 

  432. #define YSCALE_YUV_2_RGB2_C(type) \

  433. 			YSCALE_YUV_2_PACKED2_C\

  434. 			type *r, *b, *g;\

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

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

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

  438. 

  439. #define YSCALE_YUV_2_PACKED1_C \

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

  441. 			const int i2= 2*i;\

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

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

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

  445. 			int V= (uvbuf1[i+2048])>>7;\

  446. 

  447. #define YSCALE_YUV_2_RGB1_C(type) \

  448. 			YSCALE_YUV_2_PACKED1_C\

  449. 			type *r, *b, *g;\

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

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

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

  453. 

  454. #define YSCALE_YUV_2_PACKED1B_C \

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

  456. 			const int i2= 2*i;\

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

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

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

  460. 			int V= (uvbuf0[i+2048] + uvbuf1[i+2048])>>8;\

  461. 

  462. #define YSCALE_YUV_2_RGB1B_C(type) \

  463. 			YSCALE_YUV_2_PACKED1B_C\

  464. 			type *r, *b, *g;\

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

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

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

  468. 

  469. #define YSCALE_YUV_2_ANYRGB_C(func, func2)\

  470. 	switch(c->dstFormat)\

  471. 	{\

  472. 	case PIX_FMT_RGB32:\

  473. 	case PIX_FMT_BGR32:\

  474. 		func(uint32_t)\

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

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

  477. 		}		\

  478. 		break;\

  479. 	case PIX_FMT_RGB24:\

  480. 		func(uint8_t)\

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

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

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

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

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

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

  487. 			dest+=6;\

  488. 		}\

  489. 		break;\

  490. 	case PIX_FMT_BGR24:\

  491. 		func(uint8_t)\

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

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

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

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

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

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

  498. 			dest+=6;\

  499. 		}\

  500. 		break;\

  501. 	case PIX_FMT_RGB565:\

  502. 	case PIX_FMT_BGR565:\

  503. 		{\

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

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

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

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

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

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

  510. 			func(uint16_t)\

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

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

  513. 			}\

  514. 		}\

  515. 		break;\

  516. 	case PIX_FMT_RGB555:\

  517. 	case PIX_FMT_BGR555:\

  518. 		{\

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

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

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

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

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

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

  525. 			func(uint16_t)\

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

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

  528. 			}\

  529. 		}\

  530. 		break;\

  531. 	case PIX_FMT_RGB8:\

  532. 	case PIX_FMT_BGR8:\

  533. 		{\

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

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

  536. 			func(uint8_t)\

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

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

  539. 			}\

  540. 		}\

  541. 		break;\

  542. 	case PIX_FMT_RGB4:\

  543. 	case PIX_FMT_BGR4:\

  544. 		{\

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

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

  547. 			func(uint8_t)\

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

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

  550. 			}\

  551. 		}\

  552. 		break;\

  553. 	case PIX_FMT_RGB4_BYTE:\

  554. 	case PIX_FMT_BGR4_BYTE:\

  555. 		{\

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

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

  558. 			func(uint8_t)\

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

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

  561. 			}\

  562. 		}\

  563. 		break;\

  564. 	case PIX_FMT_MONOBLACK:\

  565. 		{\

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

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

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

  569. 				int acc;\

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

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

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

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

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

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

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

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

  578. 				((uint8_t*)dest)[0]= acc;\

  579. 				dest++;\

  580. 			}\

  581. \

  582. /*\

  583. ((uint8_t*)dest)-= dstW>>4;\

  584. {\

  585. 			int acc=0;\

  586. 			int left=0;\

  587. 			static int top[1024];\

  588. 			static int last_new[1024][1024];\

  589. 			static int last_in3[1024][1024];\

  590. 			static int drift[1024][1024];\

  591. 			int topLeft=0;\

  592. 			int shift=0;\

  593. 			int count=0;\

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

  595. 			int error_new=0;\

  596. 			int error_in3=0;\

  597. 			int f=0;\

  598. 			\

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

  600. 				int in= ((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19);\

  601. 				int in2 = (76309 * (in - 16) + 32768) >> 16;\

  602. 				int in3 = (in2 < 0) ? 0 : ((in2 > 255) ? 255 : in2);\

  603. 				int old= (left*7 + topLeft + top[i]*5 + top[i+1]*3)/20 + in3\

  604. 					+ (last_new[y][i] - in3)*f/256;\

  605. 				int new= old> 128 ? 255 : 0;\

  606. \

  607. 				error_new+= FFABS(last_new[y][i] - new);\

  608. 				error_in3+= FFABS(last_in3[y][i] - in3);\

  609. 				f= error_new - error_in3*4;\

  610. 				if(f<0) f=0;\

  611. 				if(f>256) f=256;\

  612. \

  613. 				topLeft= top[i];\

  614. 				left= top[i]= old - new;\

  615. 				last_new[y][i]= new;\

  616. 				last_in3[y][i]= in3;\

  617. \

  618. 				acc+= acc + (new&1);\

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

  620. 					((uint8_t*)dest)[0]= acc;\

  621. 					((uint8_t*)dest)++;\

  622. 				}\

  623. 			}\

  624. }\

  625. */\

  626. 		}\

  627. 		break;\

  628. 	case PIX_FMT_YUYV422:\

  629. 		func2\

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

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

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

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

  634. 		}		\

  635. 		break;\

  636. 	case PIX_FMT_UYVY422:\

  637. 		func2\

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

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

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

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

  642. 		}		\

  643. 		break;\

  644. 	}\

  645. 

  646. 

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

  648. 				    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,

  649. 				    uint8_t *dest, int dstW, int y)

  650. {

  651. 	int i;

  652. 	switch(c->dstFormat)

  653. 	{

  654. 	case PIX_FMT_BGR32:

  655. 	case PIX_FMT_RGB32:

  656. 		YSCALE_YUV_2_RGBX_C(uint32_t)

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

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

  659. 		}

  660. 		break;

  661. 	case PIX_FMT_RGB24:

  662. 		YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

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

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

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

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

  669. 			dest+=6;

  670. 		}

  671. 		break;

  672. 	case PIX_FMT_BGR24:

  673. 		YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

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

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

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

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

  680. 			dest+=6;

  681. 		}

  682. 		break;

  683. 	case PIX_FMT_RGB565:

  684. 	case PIX_FMT_BGR565:

  685. 		{

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

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

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

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

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

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

  692. 			YSCALE_YUV_2_RGBX_C(uint16_t)

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

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

  695. 			}

  696. 		}

  697. 		break;

  698. 	case PIX_FMT_RGB555:

  699. 	case PIX_FMT_BGR555:

  700. 		{

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

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

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

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

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

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

  707. 			YSCALE_YUV_2_RGBX_C(uint16_t)

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

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

  710. 			}

  711. 		}

  712. 		break;

  713. 	case PIX_FMT_RGB8:

  714. 	case PIX_FMT_BGR8:

  715. 		{

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

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

  718. 			YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  721. 			}

  722. 		}

  723. 		break;

  724. 	case PIX_FMT_RGB4:

  725. 	case PIX_FMT_BGR4:

  726. 		{

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

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

  729. 			YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  732. 			}

  733. 		}

  734. 		break;

  735. 	case PIX_FMT_RGB4_BYTE:

  736. 	case PIX_FMT_BGR4_BYTE:

  737. 		{

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

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

  740. 			YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  743. 			}

  744. 		}

  745. 		break;

  746. 	case PIX_FMT_MONOBLACK:

  747. 		{

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

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

  750. 			int acc=0;

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

  752. 				int j;

  753. 				int Y1=1<<18;

  754. 				int Y2=1<<18;

  755. 

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

  757. 				{

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

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

  760. 				}

  761. 				Y1>>=19;

  762. 				Y2>>=19;

  763. 				if((Y1|Y2)&256)

  764. 				{

  765. 					if(Y1>255)   Y1=255;

  766. 					else if(Y1<0)Y1=0;

  767. 					if(Y2>255)   Y2=255;

  768. 					else if(Y2<0)Y2=0;

  769. 				}

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

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

  772. 				if((i&7)==6){

  773. 					((uint8_t*)dest)[0]= acc;

  774. 					dest++;

  775. 				}

  776. 			}

  777. 		}

  778. 		break;

  779. 	case PIX_FMT_YUYV422:

  780. 		YSCALE_YUV_2_PACKEDX_C(void)

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

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

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

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

  785. 		}

  786.                 break;

  787. 	case PIX_FMT_UYVY422:

  788. 		YSCALE_YUV_2_PACKEDX_C(void)

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

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

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

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

  793. 		}

  794.                 break;

  795. 	}

  796. }

  797. 

  798. 

  799. //Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one

  800. //Plain C versions

  801. #if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) || !defined(CONFIG_GPL)

  802. #define COMPILE_C

  803. #endif

  804. 

  805. #ifdef ARCH_POWERPC

  806. #if (defined (HAVE_ALTIVEC) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)

  807. #define COMPILE_ALTIVEC

  808. #endif //HAVE_ALTIVEC

  809. #endif //ARCH_POWERPC

  810. 

  811. #if defined(ARCH_X86)

  812. 

  813. #if ((defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)

  814. #define COMPILE_MMX

  815. #endif

  816. 

  817. #if (defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)

  818. #define COMPILE_MMX2

  819. #endif

  820. 

  821. #if ((defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)

  822. #define COMPILE_3DNOW

  823. #endif

  824. #endif //ARCH_X86 || ARCH_X86_64

  825. 

  826. #undef HAVE_MMX

  827. #undef HAVE_MMX2

  828. #undef HAVE_3DNOW

  829. 

  830. #ifdef COMPILE_C

  831. #undef HAVE_MMX

  832. #undef HAVE_MMX2

  833. #undef HAVE_3DNOW

  834. #undef HAVE_ALTIVEC

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

  836. #include "swscale_template.c"

  837. #endif

  838. 

  839. #ifdef ARCH_POWERPC

  840. #ifdef COMPILE_ALTIVEC

  841. #undef RENAME

  842. #define HAVE_ALTIVEC

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

  844. #include "swscale_template.c"

  845. #endif

  846. #endif //ARCH_POWERPC

  847. 

  848. #if defined(ARCH_X86)

  849. 

  850. //X86 versions

  851. /*

  852. #undef RENAME

  853. #undef HAVE_MMX

  854. #undef HAVE_MMX2

  855. #undef HAVE_3DNOW

  856. #define ARCH_X86

  857. #define RENAME(a) a ## _X86

  858. #include "swscale_template.c"

  859. */

  860. //MMX versions

  861. #ifdef COMPILE_MMX

  862. #undef RENAME

  863. #define HAVE_MMX

  864. #undef HAVE_MMX2

  865. #undef HAVE_3DNOW

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

  867. #include "swscale_template.c"

  868. #endif

  869. 

  870. //MMX2 versions

  871. #ifdef COMPILE_MMX2

  872. #undef RENAME

  873. #define HAVE_MMX

  874. #define HAVE_MMX2

  875. #undef HAVE_3DNOW

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

  877. #include "swscale_template.c"

  878. #endif

  879. 

  880. //3DNOW versions

  881. #ifdef COMPILE_3DNOW

  882. #undef RENAME

  883. #define HAVE_MMX

  884. #undef HAVE_MMX2

  885. #define HAVE_3DNOW

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

  887. #include "swscale_template.c"

  888. #endif

  889. 

  890. #endif //ARCH_X86 || ARCH_X86_64

  891. 

  892. // minor note: the HAVE_xyz is messed up after that line so don't use it

  893. 

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

  895. {

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

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

  898. 	else		return getSplineCoeff(	0.0, 

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

  900. 						        c + 3.0*d,

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

  902. 						dist-1.0);

  903. }

  904. 

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

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

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

  908. {

  909. 	int i;

  910. 	int filterSize;

  911. 	int filter2Size;

  912. 	int minFilterSize;

  913. 	double *filter=NULL;

  914. 	double *filter2=NULL;

  915. #if defined(ARCH_X86)

  916. 	if(flags & SWS_CPU_CAPS_MMX)

  917. 		asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions)

  918. #endif

  919. 

  920. 	// Note the +1 is for the MMXscaler which reads over the end

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

  922. 

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

  924. 	{

  925. 		int i;

  926. 		filterSize= 1;

  927. 		filter= av_malloc(dstW*sizeof(double)*filterSize);

  928. 		for(i=0; i<dstW*filterSize; i++) filter[i]=0;

  929. 

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

  931. 		{

  932. 			filter[i*filterSize]=1;

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

  934. 		}

  935. 

  936. 	}

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

  938. 	{

  939. 		int i;

  940. 		int xDstInSrc;

  941. 		filterSize= 1;

  942. 		filter= av_malloc(dstW*sizeof(double)*filterSize);

  943. 		

  944. 		xDstInSrc= xInc/2 - 0x8000;

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

  946. 		{

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

  948. 

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

  950. 			filter[i]= 1.0;

  951. 			xDstInSrc+= xInc;

  952. 		}

  953. 	}

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

  955. 	{

  956. 		int i;

  957. 		int xDstInSrc;

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

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

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

  961. 		filter= av_malloc(dstW*sizeof(double)*filterSize);

  962. 

  963. 		xDstInSrc= xInc/2 - 0x8000;

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

  965. 		{

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

  967. 			int j;

  968. 

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

  970. 				//Bilinear upscale / linear interpolate / Area averaging

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

  972. 				{

  973. 					double d= FFABS((xx<<16) - xDstInSrc)/(double)(1<<16);

  974. 					double coeff= 1.0 - d;

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

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

  977. 					xx++;

  978. 				}

  979. 			xDstInSrc+= xInc;

  980. 		}

  981. 	}

  982. 	else

  983. 	{

  984. 		double xDstInSrc;

  985. 		double sizeFactor, filterSizeInSrc;

  986. 		const double xInc1= (double)xInc / (double)(1<<16);

  987. 

  988. 		if     (flags&SWS_BICUBIC)	sizeFactor= 4.0;

  989. 		else if(flags&SWS_X)		sizeFactor= 8.0;

  990. 		else if(flags&SWS_AREA)		sizeFactor= 1.0; //downscale only, for upscale it is bilinear

  991. 		else if(flags&SWS_GAUSS)	sizeFactor= 8.0;   // infinite ;)

  992. 		else if(flags&SWS_LANCZOS)	sizeFactor= param[0] != SWS_PARAM_DEFAULT ? 2.0*param[0] : 6.0;

  993. 		else if(flags&SWS_SINC)		sizeFactor= 20.0; // infinite ;)

  994. 		else if(flags&SWS_SPLINE)	sizeFactor= 20.0;  // infinite ;)

  995. 		else if(flags&SWS_BILINEAR)	sizeFactor= 2.0;

  996. 		else {

  997. 			sizeFactor= 0.0; //GCC warning killer

  998. 			ASSERT(0)

  999. 		}

  1000. 		

  1001. 		if(xInc1 <= 1.0)	filterSizeInSrc= sizeFactor; // upscale

  1002. 		else			filterSizeInSrc= sizeFactor*srcW / (double)dstW;

  1003. 

  1004. 		filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible

  1005. 		if(filterSize > srcW-2) filterSize=srcW-2;

  1006. 

  1007. 		filter= av_malloc(dstW*sizeof(double)*filterSize);

  1008. 

  1009. 		xDstInSrc= xInc1 / 2.0 - 0.5;

  1010. 		for(i=0; i<dstW; i++)

  1011. 		{

  1012. 			int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);

  1013. 			int j;

  1014. 			(*filterPos)[i]= xx;

  1015. 			for(j=0; j<filterSize; j++)

  1016. 			{

  1017. 				double d= FFABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;

  1018. 				double coeff;

  1019. 				if(flags & SWS_BICUBIC)

  1020. 				{

  1021. 					double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;

  1022. 					double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;

  1023. 

  1024. 					if(d<1.0) 

  1025. 						coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;

  1026. 					else if(d<2.0)

  1027. 						coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;

  1028. 					else

  1029. 						coeff=0.0;

  1030. 				}

  1031. /*				else if(flags & SWS_X)

  1032. 				{

  1033. 					double p= param ? param*0.01 : 0.3;

  1034. 					coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1035. 					coeff*= pow(2.0, - p*d*d);

  1036. 				}*/

  1037. 				else if(flags & SWS_X)

  1038. 				{

  1039. 					double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

  1040. 					

  1041. 					if(d<1.0)

  1042. 						coeff = cos(d*PI);

  1043. 					else

  1044. 						coeff=-1.0;

  1045. 					if(coeff<0.0) 	coeff= -pow(-coeff, A);

  1046. 					else		coeff=  pow( coeff, A);

  1047. 					coeff= coeff*0.5 + 0.5;

  1048. 				}

  1049. 				else if(flags & SWS_AREA)

  1050. 				{

  1051. 					double srcPixelSize= 1.0/xInc1;

  1052. 					if(d + srcPixelSize/2 < 0.5) coeff= 1.0;

  1053. 					else if(d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5;

  1054. 					else coeff=0.0;

  1055. 				}

  1056. 				else if(flags & SWS_GAUSS)

  1057. 				{

  1058. 					double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1059. 					coeff = pow(2.0, - p*d*d);

  1060. 				}

  1061. 				else if(flags & SWS_SINC)

  1062. 				{

  1063. 					coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1064. 				}

  1065. 				else if(flags & SWS_LANCZOS)

  1066. 				{

  1067. 					double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0; 

  1068. 					coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;

  1069. 					if(d>p) coeff=0;

  1070. 				}

  1071. 				else if(flags & SWS_BILINEAR)

  1072. 				{

  1073. 					coeff= 1.0 - d;

  1074. 					if(coeff<0) coeff=0;

  1075. 				}

  1076. 				else if(flags & SWS_SPLINE)

  1077. 				{

  1078. 					double p=-2.196152422706632;

  1079. 					coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);

  1080. 				}

  1081. 				else {

  1082. 					coeff= 0.0; //GCC warning killer

  1083. 					ASSERT(0)

  1084. 				}

  1085. 

  1086. 				filter[i*filterSize + j]= coeff;

  1087. 				xx++;

  1088. 			}

  1089. 			xDstInSrc+= xInc1;

  1090. 		}

  1091. 	}

  1092. 

  1093. 	/* apply src & dst Filter to filter -> filter2

  1094. 	   av_free(filter);

  1095. 	*/

  1096. 	ASSERT(filterSize>0)

  1097. 	filter2Size= filterSize;

  1098. 	if(srcFilter) filter2Size+= srcFilter->length - 1;

  1099. 	if(dstFilter) filter2Size+= dstFilter->length - 1;

  1100. 	ASSERT(filter2Size>0)

  1101. 	filter2= av_malloc(filter2Size*dstW*sizeof(double));

  1102. 

  1103. 	for(i=0; i<dstW; i++)

  1104. 	{

  1105. 		int j;

  1106. 		SwsVector scaleFilter;

  1107. 		SwsVector *outVec;

  1108. 

  1109. 		scaleFilter.coeff= filter + i*filterSize;

  1110. 		scaleFilter.length= filterSize;

  1111. 

  1112. 		if(srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);

  1113. 		else	      outVec= &scaleFilter;

  1114. 

  1115. 		ASSERT(outVec->length == filter2Size)

  1116. 		//FIXME dstFilter

  1117. 

  1118. 		for(j=0; j<outVec->length; j++)

  1119. 		{

  1120. 			filter2[i*filter2Size + j]= outVec->coeff[j];

  1121. 		}

  1122. 

  1123. 		(*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

  1124. 

  1125. 		if(outVec != &scaleFilter) sws_freeVec(outVec);

  1126. 	}

  1127. 	av_free(filter); filter=NULL;

  1128. 

  1129. 	/* try to reduce the filter-size (step1 find size and shift left) */

  1130. 	// Assume its near normalized (*0.5 or *2.0 is ok but * 0.001 is not)

  1131. 	minFilterSize= 0;

  1132. 	for(i=dstW-1; i>=0; i--)

  1133. 	{

  1134. 		int min= filter2Size;

  1135. 		int j;

  1136. 		double cutOff=0.0;

  1137. 

  1138. 		/* get rid off near zero elements on the left by shifting left */

  1139. 		for(j=0; j<filter2Size; j++)

  1140. 		{

  1141. 			int k;

  1142. 			cutOff += FFABS(filter2[i*filter2Size]);

  1143. 

  1144. 			if(cutOff > SWS_MAX_REDUCE_CUTOFF) break;

  1145. 

  1146. 			/* preserve Monotonicity because the core can't handle the filter otherwise */

  1147. 			if(i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

  1148. 

  1149. 			// Move filter coeffs left

  1150. 			for(k=1; k<filter2Size; k++)

  1151. 				filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];

  1152. 			filter2[i*filter2Size + k - 1]= 0.0;

  1153. 			(*filterPos)[i]++;

  1154. 		}

  1155. 

  1156. 		cutOff=0.0;

  1157. 		/* count near zeros on the right */

  1158. 		for(j=filter2Size-1; j>0; j--)

  1159. 		{

  1160. 			cutOff += FFABS(filter2[i*filter2Size + j]);

  1161. 

  1162. 			if(cutOff > SWS_MAX_REDUCE_CUTOFF) break;

  1163. 			min--;

  1164. 		}

  1165. 

  1166. 		if(min>minFilterSize) minFilterSize= min;

  1167. 	}

  1168. 

  1169.         if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1170.           // we can handle the special case 4,

  1171.           // so we don't want to go to the full 8

  1172.           if (minFilterSize < 5)

  1173.             filterAlign = 4;

  1174. 

  1175.           // we really don't want to waste our time

  1176.           // doing useless computation, so fall-back on

  1177.           // the scalar C code for very small filter.

  1178.           // vectorizing is worth it only if you have

  1179.           // decent-sized vector.

  1180.           if (minFilterSize < 3)

  1181.             filterAlign = 1;

  1182.         }

  1183. 

  1184.         if (flags & SWS_CPU_CAPS_MMX) {

  1185.                 // special case for unscaled vertical filtering

  1186.                 if(minFilterSize == 1 && filterAlign == 2)

  1187.                         filterAlign= 1;

  1188.         }

  1189. 

  1190. 	ASSERT(minFilterSize > 0)

  1191. 	filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));

  1192. 	ASSERT(filterSize > 0)

  1193. 	filter= av_malloc(filterSize*dstW*sizeof(double));

  1194.         if(filterSize >= MAX_FILTER_SIZE)

  1195.                 return -1;

  1196. 	*outFilterSize= filterSize;

  1197. 

  1198. 	if(flags&SWS_PRINT_INFO)

  1199. 		MSG_V("SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);

  1200. 	/* try to reduce the filter-size (step2 reduce it) */

  1201. 	for(i=0; i<dstW; i++)

  1202. 	{

  1203. 		int j;

  1204. 

  1205. 		for(j=0; j<filterSize; j++)

  1206. 		{

  1207. 			if(j>=filter2Size) filter[i*filterSize + j]= 0.0;

  1208. 			else		   filter[i*filterSize + j]= filter2[i*filter2Size + j];

  1209. 		}

  1210. 	}

  1211. 	av_free(filter2); filter2=NULL;

  1212. 	

  1213. 

  1214. 	//FIXME try to align filterpos if possible

  1215. 

  1216. 	//fix borders

  1217. 	for(i=0; i<dstW; i++)

  1218. 	{

  1219. 		int j;

  1220. 		if((*filterPos)[i] < 0)

  1221. 		{

  1222. 			// Move filter coeffs left to compensate for filterPos

  1223. 			for(j=1; j<filterSize; j++)

  1224. 			{

  1225. 				int left= FFMAX(j + (*filterPos)[i], 0);

  1226. 				filter[i*filterSize + left] += filter[i*filterSize + j];

  1227. 				filter[i*filterSize + j]=0;

  1228. 			}

  1229. 			(*filterPos)[i]= 0;

  1230. 		}

  1231. 

  1232. 		if((*filterPos)[i] + filterSize > srcW)

  1233. 		{

  1234. 			int shift= (*filterPos)[i] + filterSize - srcW;

  1235. 			// Move filter coeffs right to compensate for filterPos

  1236. 			for(j=filterSize-2; j>=0; j--)

  1237. 			{

  1238. 				int right= FFMIN(j + shift, filterSize-1);

  1239. 				filter[i*filterSize +right] += filter[i*filterSize +j];

  1240. 				filter[i*filterSize +j]=0;

  1241. 			}

  1242. 			(*filterPos)[i]= srcW - filterSize;

  1243. 		}

  1244. 	}

  1245. 

  1246. 	// Note the +1 is for the MMXscaler which reads over the end

  1247. 	/* align at 16 for AltiVec (needed by hScale_altivec_real) */

  1248. 	*outFilter= av_malloc(*outFilterSize*(dstW+1)*sizeof(int16_t));

  1249. 	memset(*outFilter, 0, *outFilterSize*(dstW+1)*sizeof(int16_t));

  1250. 

  1251. 	/* Normalize & Store in outFilter */

  1252. 	for(i=0; i<dstW; i++)

  1253. 	{

  1254. 		int j;

  1255. 		double error=0;

  1256. 		double sum=0;

  1257. 		double scale= one;

  1258. 

  1259. 		for(j=0; j<filterSize; j++)

  1260. 		{

  1261. 			sum+= filter[i*filterSize + j];

  1262. 		}

  1263. 		scale/= sum;

  1264. 		for(j=0; j<*outFilterSize; j++)

  1265. 		{

  1266. 			double v= filter[i*filterSize + j]*scale + error;

  1267. 			int intV= floor(v + 0.5);

  1268. 			(*outFilter)[i*(*outFilterSize) + j]= intV;

  1269. 			error = v - intV;

  1270. 		}

  1271. 	}

  1272. 	

  1273. 	(*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end

  1274. 	for(i=0; i<*outFilterSize; i++)

  1275. 	{

  1276. 		int j= dstW*(*outFilterSize);

  1277. 		(*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

  1278. 	}

  1279. 

  1280. 	av_free(filter);

  1281.         return 0;

  1282. }

  1283. 

  1284. #ifdef COMPILE_MMX2

  1285. static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)

  1286. {

  1287. 	uint8_t *fragmentA;

  1288. 	long imm8OfPShufW1A;

  1289. 	long imm8OfPShufW2A;

  1290. 	long fragmentLengthA;

  1291. 	uint8_t *fragmentB;

  1292. 	long imm8OfPShufW1B;

  1293. 	long imm8OfPShufW2B;

  1294. 	long fragmentLengthB;

  1295. 	int fragmentPos;

  1296. 

  1297. 	int xpos, i;

  1298. 

  1299. 	// create an optimized horizontal scaling routine

  1300. 

  1301. 	//code fragment

  1302. 

  1303. 	asm volatile(

  1304. 		"jmp 9f				\n\t"

  1305. 	// Begin

  1306. 		"0:				\n\t"

  1307. 		"movq (%%"REG_d", %%"REG_a"), %%mm3\n\t" 

  1308. 		"movd (%%"REG_c", %%"REG_S"), %%mm0\n\t" 

  1309. 		"movd 1(%%"REG_c", %%"REG_S"), %%mm1\n\t"

  1310. 		"punpcklbw %%mm7, %%mm1		\n\t"

  1311. 		"punpcklbw %%mm7, %%mm0		\n\t"

  1312. 		"pshufw $0xFF, %%mm1, %%mm1	\n\t"

  1313. 		"1:				\n\t"

  1314. 		"pshufw $0xFF, %%mm0, %%mm0	\n\t"

  1315. 		"2:				\n\t"

  1316. 		"psubw %%mm1, %%mm0		\n\t"

  1317. 		"movl 8(%%"REG_b", %%"REG_a"), %%esi\n\t"

  1318. 		"pmullw %%mm3, %%mm0		\n\t"

  1319. 		"psllw $7, %%mm1		\n\t"

  1320. 		"paddw %%mm1, %%mm0		\n\t"

  1321. 

  1322. 		"movq %%mm0, (%%"REG_D", %%"REG_a")\n\t"

  1323. 

  1324. 		"add $8, %%"REG_a"		\n\t"

  1325. 	// End

  1326. 		"9:				\n\t"

  1327. //		"int $3\n\t"

  1328. 		"lea 0b, %0			\n\t"

  1329. 		"lea 1b, %1			\n\t"

  1330. 		"lea 2b, %2			\n\t"

  1331. 		"dec %1				\n\t"

  1332. 		"dec %2				\n\t"

  1333. 		"sub %0, %1			\n\t"

  1334. 		"sub %0, %2			\n\t"

  1335. 		"lea 9b, %3			\n\t"

  1336. 		"sub %0, %3			\n\t"

  1337. 

  1338. 

  1339. 		:"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),

  1340. 		"=r" (fragmentLengthA)

  1341. 	);

  1342. 

  1343. 	asm volatile(

  1344. 		"jmp 9f				\n\t"

  1345. 	// Begin

  1346. 		"0:				\n\t"

  1347. 		"movq (%%"REG_d", %%"REG_a"), %%mm3\n\t" 

  1348. 		"movd (%%"REG_c", %%"REG_S"), %%mm0\n\t" 

  1349. 		"punpcklbw %%mm7, %%mm0		\n\t"

  1350. 		"pshufw $0xFF, %%mm0, %%mm1	\n\t"

  1351. 		"1:				\n\t"

  1352. 		"pshufw $0xFF, %%mm0, %%mm0	\n\t"

  1353. 		"2:				\n\t"

  1354. 		"psubw %%mm1, %%mm0		\n\t"

  1355. 		"movl 8(%%"REG_b", %%"REG_a"), %%esi\n\t"

  1356. 		"pmullw %%mm3, %%mm0		\n\t"

  1357. 		"psllw $7, %%mm1		\n\t"

  1358. 		"paddw %%mm1, %%mm0		\n\t"

  1359. 

  1360. 		"movq %%mm0, (%%"REG_D", %%"REG_a")\n\t"

  1361. 

  1362. 		"add $8, %%"REG_a"		\n\t"

  1363. 	// End

  1364. 		"9:				\n\t"

  1365. //		"int $3\n\t"

  1366. 		"lea 0b, %0			\n\t"

  1367. 		"lea 1b, %1			\n\t"

  1368. 		"lea 2b, %2			\n\t"

  1369. 		"dec %1				\n\t"

  1370. 		"dec %2				\n\t"

  1371. 		"sub %0, %1			\n\t"

  1372. 		"sub %0, %2			\n\t"

  1373. 		"lea 9b, %3			\n\t"

  1374. 		"sub %0, %3			\n\t"

  1375. 

  1376. 

  1377. 		:"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),

  1378. 		"=r" (fragmentLengthB)

  1379. 	);

  1380. 

  1381. 	xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers

  1382. 	fragmentPos=0;

  1383. 	

  1384. 	for(i=0; i<dstW/numSplits; i++)

  1385. 	{

  1386. 		int xx=xpos>>16;

  1387. 

  1388. 		if((i&3) == 0)

  1389. 		{

  1390. 			int a=0;

  1391. 			int b=((xpos+xInc)>>16) - xx;

  1392. 			int c=((xpos+xInc*2)>>16) - xx;

  1393. 			int d=((xpos+xInc*3)>>16) - xx;

  1394. 

  1395. 			filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;

  1396. 			filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;

  1397. 			filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;

  1398. 			filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;

  1399. 			filterPos[i/2]= xx;

  1400. 

  1401. 			if(d+1<4)

  1402. 			{

  1403. 				int maxShift= 3-(d+1);

  1404. 				int shift=0;

  1405. 

  1406. 				memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);

  1407. 

  1408. 				funnyCode[fragmentPos + imm8OfPShufW1B]=

  1409. 					(a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);

  1410. 				funnyCode[fragmentPos + imm8OfPShufW2B]=

  1411. 					a | (b<<2) | (c<<4) | (d<<6);

  1412. 

  1413. 				if(i+3>=dstW) shift=maxShift; //avoid overread

  1414. 				else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

  1415. 

  1416. 				if(shift && i>=shift)

  1417. 				{

  1418. 					funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;

  1419. 					funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;

  1420. 					filterPos[i/2]-=shift;

  1421. 				}

  1422. 

  1423. 				fragmentPos+= fragmentLengthB;

  1424. 			}

  1425. 			else

  1426. 			{

  1427. 				int maxShift= 3-d;

  1428. 				int shift=0;

  1429. 

  1430. 				memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);

  1431. 

  1432. 				funnyCode[fragmentPos + imm8OfPShufW1A]=

  1433. 				funnyCode[fragmentPos + imm8OfPShufW2A]=

  1434. 					a | (b<<2) | (c<<4) | (d<<6);

  1435. 

  1436. 				if(i+4>=dstW) shift=maxShift; //avoid overread

  1437. 				else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align

  1438. 

  1439. 				if(shift && i>=shift)

  1440. 				{

  1441. 					funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;

  1442. 					funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;

  1443. 					filterPos[i/2]-=shift;

  1444. 				}

  1445. 

  1446. 				fragmentPos+= fragmentLengthA;

  1447. 			}

  1448. 

  1449. 			funnyCode[fragmentPos]= RET;

  1450. 		}

  1451. 		xpos+=xInc;

  1452. 	}

  1453. 	filterPos[i/2]= xpos>>16; // needed to jump to the next part

  1454. }

  1455. #endif /* COMPILE_MMX2 */

  1456. 

  1457. static void globalInit(void){

  1458.     // generating tables:

  1459.     int i;

  1460.     for(i=0; i<768; i++){

  1461. 	int c= FFMIN(FFMAX(i-256, 0), 255);

  1462. 	clip_table[i]=c;

  1463.     }

  1464. }

  1465. 

  1466. static SwsFunc getSwsFunc(int flags){

  1467.     

  1468. #if defined(RUNTIME_CPUDETECT) && defined (CONFIG_GPL)

  1469. #if defined(ARCH_X86)

  1470. 	// ordered per speed fasterst first

  1471. 	if(flags & SWS_CPU_CAPS_MMX2)

  1472. 		return swScale_MMX2;

  1473. 	else if(flags & SWS_CPU_CAPS_3DNOW)

  1474. 		return swScale_3DNow;

  1475. 	else if(flags & SWS_CPU_CAPS_MMX)

  1476. 		return swScale_MMX;

  1477. 	else

  1478. 		return swScale_C;

  1479. 

  1480. #else

  1481. #ifdef ARCH_POWERPC

  1482. 	if(flags & SWS_CPU_CAPS_ALTIVEC)

  1483. 	  return swScale_altivec;

  1484. 	else

  1485. 	  return swScale_C;

  1486. #endif

  1487. 	return swScale_C;

  1488. #endif /* defined(ARCH_X86) */

  1489. #else //RUNTIME_CPUDETECT

  1490. #ifdef HAVE_MMX2

  1491. 	return swScale_MMX2;

  1492. #elif defined (HAVE_3DNOW)

  1493. 	return swScale_3DNow;

  1494. #elif defined (HAVE_MMX)

  1495. 	return swScale_MMX;

  1496. #elif defined (HAVE_ALTIVEC)

  1497. 	return swScale_altivec;

  1498. #else

  1499. 	return swScale_C;

  1500. #endif

  1501. #endif //!RUNTIME_CPUDETECT

  1502. }

  1503. 

  1504. static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1505.              int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1506. 	uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1507. 	/* Copy Y plane */

  1508. 	if(dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1509. 		memcpy(dst, src[0], srcSliceH*dstStride[0]);

  1510. 	else

  1511. 	{

  1512. 		int i;

  1513. 		uint8_t *srcPtr= src[0];

  1514. 		uint8_t *dstPtr= dst;

  1515. 		for(i=0; i<srcSliceH; i++)

  1516. 		{

  1517. 			memcpy(dstPtr, srcPtr, c->srcW);

  1518. 			srcPtr+= srcStride[0];

  1519. 			dstPtr+= dstStride[0];

  1520. 		}

  1521. 	}

  1522. 	dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1523. 	if (c->dstFormat == PIX_FMT_NV12)

  1524. 		interleaveBytes( src[1],src[2],dst,c->srcW/2,srcSliceH/2,srcStride[1],srcStride[2],dstStride[0] );

  1525. 	else

  1526. 		interleaveBytes( src[2],src[1],dst,c->srcW/2,srcSliceH/2,srcStride[2],srcStride[1],dstStride[0] );

  1527. 

  1528. 	return srcSliceH;

  1529. }

  1530. 

  1531. static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1532.              int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1533. 	uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1534. 

  1535. 	yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );

  1536. 

  1537. 	return srcSliceH;

  1538. }

  1539. 

  1540. static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1541.              int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1542. 	uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1543. 

  1544. 	yv12touyvy( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );

  1545. 

  1546. 	return srcSliceH;

  1547. }

  1548. 

  1549. /* {RGB,BGR}{15,16,24,32} -> {RGB,BGR}{15,16,24,32} */

  1550. static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1551. 			   int srcSliceH, uint8_t* dst[], int dstStride[]){

  1552. 	const int srcFormat= c->srcFormat;

  1553. 	const int dstFormat= c->dstFormat;

  1554. 	const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;

  1555. 	const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;

  1556. 	const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1557. 	const int dstId= fmt_depth(dstFormat) >> 2;

  1558. 	void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  1559. 

  1560. 	/* BGR -> BGR */

  1561. 	if(   (isBGR(srcFormat) && isBGR(dstFormat))

  1562. 	   || (isRGB(srcFormat) && isRGB(dstFormat))){

  1563. 		switch(srcId | (dstId<<4)){

  1564. 		case 0x34: conv= rgb16to15; break;

  1565. 		case 0x36: conv= rgb24to15; break;

  1566. 		case 0x38: conv= rgb32to15; break;

  1567. 		case 0x43: conv= rgb15to16; break;

  1568. 		case 0x46: conv= rgb24to16; break;

  1569. 		case 0x48: conv= rgb32to16; break;

  1570. 		case 0x63: conv= rgb15to24; break;

  1571. 		case 0x64: conv= rgb16to24; break;

  1572. 		case 0x68: conv= rgb32to24; break;

  1573. 		case 0x83: conv= rgb15to32; break;

  1574. 		case 0x84: conv= rgb16to32; break;

  1575. 		case 0x86: conv= rgb24to32; break;

  1576. 		default: MSG_ERR("swScaler: internal error %s -> %s converter\n", 

  1577. 				 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1578. 		}

  1579. 	}else if(   (isBGR(srcFormat) && isRGB(dstFormat))

  1580. 		 || (isRGB(srcFormat) && isBGR(dstFormat))){

  1581. 		switch(srcId | (dstId<<4)){

  1582. 		case 0x33: conv= rgb15tobgr15; break;

  1583. 		case 0x34: conv= rgb16tobgr15; break;

  1584. 		case 0x36: conv= rgb24tobgr15; break;

  1585. 		case 0x38: conv= rgb32tobgr15; break;

  1586. 		case 0x43: conv= rgb15tobgr16; break;

  1587. 		case 0x44: conv= rgb16tobgr16; break;

  1588. 		case 0x46: conv= rgb24tobgr16; break;

  1589. 		case 0x48: conv= rgb32tobgr16; break;

  1590. 		case 0x63: conv= rgb15tobgr24; break;

  1591. 		case 0x64: conv= rgb16tobgr24; break;

  1592. 		case 0x66: conv= rgb24tobgr24; break;

  1593. 		case 0x68: conv= rgb32tobgr24; break;

  1594. 		case 0x83: conv= rgb15tobgr32; break;

  1595. 		case 0x84: conv= rgb16tobgr32; break;

  1596. 		case 0x86: conv= rgb24tobgr32; break;

  1597. 		case 0x88: conv= rgb32tobgr32; break;

  1598. 		default: MSG_ERR("swScaler: internal error %s -> %s converter\n", 

  1599. 				 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1600. 		}

  1601. 	}else{

  1602. 		MSG_ERR("swScaler: internal error %s -> %s converter\n", 

  1603. 			 sws_format_name(srcFormat), sws_format_name(dstFormat));

  1604. 	}

  1605. 

  1606. 	if(dstStride[0]*srcBpp == srcStride[0]*dstBpp)

  1607. 		conv(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  1608. 	else

  1609. 	{

  1610. 		int i;

  1611. 		uint8_t *srcPtr= src[0];

  1612. 		uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1613. 

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

  1615. 		{

  1616. 			conv(srcPtr, dstPtr, c->srcW*srcBpp);

  1617. 			srcPtr+= srcStride[0];

  1618. 			dstPtr+= dstStride[0];

  1619. 		}

  1620. 	}     

  1621. 	return srcSliceH;

  1622. }

  1623. 

  1624. static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1625.              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1626. 

  1627. 	rgb24toyv12(

  1628. 		src[0], 

  1629. 		dst[0]+ srcSliceY    *dstStride[0], 

  1630. 		dst[1]+(srcSliceY>>1)*dstStride[1], 

  1631. 		dst[2]+(srcSliceY>>1)*dstStride[2],

  1632. 		c->srcW, srcSliceH, 

  1633. 		dstStride[0], dstStride[1], srcStride[0]);

  1634. 	return srcSliceH;

  1635. }

  1636. 

  1637. static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1638.              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1639. 	int i;

  1640. 

  1641. 	/* copy Y */

  1642. 	if(srcStride[0]==dstStride[0] && srcStride[0] > 0) 

  1643. 		memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);

  1644. 	else{

  1645. 		uint8_t *srcPtr= src[0];

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

  1647. 

  1648. 		for(i=0; i<srcSliceH; i++)

  1649. 		{

  1650. 			memcpy(dstPtr, srcPtr, c->srcW);

  1651. 			srcPtr+= srcStride[0];

  1652. 			dstPtr+= dstStride[0];

  1653. 		}

  1654. 	}

  1655. 

  1656. 	if(c->dstFormat==PIX_FMT_YUV420P){

  1657. 		planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);

  1658. 		planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);

  1659. 	}else{

  1660. 		planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);

  1661. 		planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);

  1662. 	}

  1663. 	return srcSliceH;

  1664. }

  1665. 

  1666. /* unscaled copy like stuff (assumes nearly identical formats) */

  1667. static int simpleCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1668.              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1669. 

  1670. 	if(isPacked(c->srcFormat))

  1671. 	{

  1672. 		if(dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1673. 			memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  1674. 		else

  1675. 		{

  1676. 			int i;

  1677. 			uint8_t *srcPtr= src[0];

  1678. 			uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1679. 			int length=0;

  1680. 

  1681. 			/* universal length finder */

  1682. 			while(length+c->srcW <= FFABS(dstStride[0]) 

  1683. 			   && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  1684. 			ASSERT(length!=0);

  1685. 

  1686. 			for(i=0; i<srcSliceH; i++)

  1687. 			{

  1688. 				memcpy(dstPtr, srcPtr, length);

  1689. 				srcPtr+= srcStride[0];

  1690. 				dstPtr+= dstStride[0];

  1691. 			}

  1692. 		}

  1693. 	}

  1694. 	else 

  1695. 	{ /* Planar YUV or gray */

  1696. 		int plane;

  1697. 		for(plane=0; plane<3; plane++)

  1698. 		{

  1699. 			int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  1700. 			int y=      plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  1701. 			int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  1702. 

  1703. 			if((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)

  1704. 			{

  1705. 				if(!isGray(c->dstFormat))

  1706. 					memset(dst[plane], 128, dstStride[plane]*height);

  1707. 			}

  1708. 			else

  1709. 			{

  1710. 				if(dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)

  1711. 					memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);

  1712. 				else

  1713. 				{

  1714. 					int i;

  1715. 					uint8_t *srcPtr= src[plane];

  1716. 					uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  1717. 					for(i=0; i<height; i++)

  1718. 					{

  1719. 						memcpy(dstPtr, srcPtr, length);

  1720. 						srcPtr+= srcStride[plane];

  1721. 						dstPtr+= dstStride[plane];

  1722. 					}

  1723. 				}

  1724. 			}

  1725. 		}

  1726. 	}

  1727. 	return srcSliceH;

  1728. }

  1729. 

  1730. static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1731.              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1732. 

  1733. 	int length= c->srcW;

  1734. 	int y=      srcSliceY;

  1735. 	int height= srcSliceH;

  1736. 	int i, j;

  1737. 	uint8_t *srcPtr= src[0];

  1738. 	uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  1739. 

  1740. 	if(!isGray(c->dstFormat)){

  1741. 		int height= -((-srcSliceH)>>c->chrDstVSubSample);

  1742. 		memset(dst[1], 128, dstStride[1]*height);

  1743. 		memset(dst[2], 128, dstStride[2]*height);

  1744. 	}

  1745. 	if(c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;

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

  1747. 	{

  1748. 		for(j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  1749. 		srcPtr+= srcStride[0];

  1750. 		dstPtr+= dstStride[0];

  1751. 	}

  1752. 	return srcSliceH;

  1753. }

  1754. 

  1755. static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1756.              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1757. 

  1758. 	int length= c->srcW;

  1759. 	int y=      srcSliceY;

  1760. 	int height= srcSliceH;

  1761. 	int i, j;

  1762. 	uint8_t *srcPtr= src[0];

  1763. 	uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  1764. 	for(i=0; i<height; i++)

  1765. 	{

  1766. 		for(j=0; j<length; j++)

  1767. 		{

  1768. 			dstPtr[j<<1] = srcPtr[j];

  1769. 			dstPtr[(j<<1)+1] = srcPtr[j];

  1770. 		}

  1771. 		srcPtr+= srcStride[0];

  1772. 		dstPtr+= dstStride[0];

  1773. 	}

  1774. 	return srcSliceH;

  1775. }

  1776. 

  1777. static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1778.              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1779. 

  1780. 	int length= c->srcW;

  1781. 	int y=      srcSliceY;

  1782. 	int height= srcSliceH;

  1783. 	int i, j;

  1784. 	uint16_t *srcPtr= src[0];

  1785. 	uint16_t *dstPtr= dst[0] + dstStride[0]*y/2;

  1786. 	for(i=0; i<height; i++)

  1787. 	{

  1788. 		for(j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);

  1789. 		srcPtr+= srcStride[0]/2;

  1790. 		dstPtr+= dstStride[0]/2;

  1791. 	}

  1792. 	return srcSliceH;

  1793. }

  1794. 

  1795. 

  1796. static void getSubSampleFactors(int *h, int *v, int format){

  1797. 	switch(format){

  1798. 	case PIX_FMT_UYVY422:

  1799. 	case PIX_FMT_YUYV422:

  1800. 		*h=1;

  1801. 		*v=0;

  1802. 		break;

  1803. 	case PIX_FMT_YUV420P:

  1804. 	case PIX_FMT_GRAY16BE:

  1805. 	case PIX_FMT_GRAY16LE:

  1806. 	case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented

  1807. 	case PIX_FMT_NV12:

  1808. 	case PIX_FMT_NV21:

  1809. 		*h=1;

  1810. 		*v=1;

  1811. 		break;

  1812. 	case PIX_FMT_YUV410P:

  1813. 		*h=2;

  1814. 		*v=2;

  1815. 		break;

  1816. 	case PIX_FMT_YUV444P:

  1817. 		*h=0;

  1818. 		*v=0;

  1819. 		break;

  1820. 	case PIX_FMT_YUV422P:

  1821. 		*h=1;

  1822. 		*v=0;

  1823. 		break;

  1824. 	case PIX_FMT_YUV411P:

  1825. 		*h=2;

  1826. 		*v=0;

  1827. 		break;

  1828. 	default:

  1829. 		*h=0;

  1830. 		*v=0;

  1831. 		break;

  1832. 	}

  1833. }

  1834. 

  1835. static uint16_t roundToInt16(int64_t f){

  1836. 	int r= (f + (1<<15))>>16;

  1837. 	     if(r<-0x7FFF) return 0x8000;

  1838. 	else if(r> 0x7FFF) return 0x7FFF;

  1839. 	else               return r;

  1840. }

  1841. 

  1842. /**

  1843.  * @param inv_table the yuv2rgb coeffs, normally Inverse_Table_6_9[x]

  1844.  * @param fullRange if 1 then the luma range is 0..255 if 0 its 16..235

  1845.  * @return -1 if not supported

  1846.  */

  1847. int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){

  1848. 	int64_t crv =  inv_table[0];

  1849. 	int64_t cbu =  inv_table[1];

  1850. 	int64_t cgu = -inv_table[2];

  1851. 	int64_t cgv = -inv_table[3];

  1852. 	int64_t cy  = 1<<16;

  1853. 	int64_t oy  = 0;

  1854. 

  1855. 	if(isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  1856. 	memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);

  1857. 	memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);

  1858. 

  1859. 	c->brightness= brightness;

  1860. 	c->contrast  = contrast;

  1861. 	c->saturation= saturation;

  1862. 	c->srcRange  = srcRange;

  1863. 	c->dstRange  = dstRange;

  1864. 

  1865. 	c->uOffset=   0x0400040004000400LL;

  1866. 	c->vOffset=   0x0400040004000400LL;

  1867. 

  1868. 	if(!srcRange){

  1869. 		cy= (cy*255) / 219;

  1870. 		oy= 16<<16;

  1871. 	}

  1872. 

  1873. 	cy = (cy *contrast             )>>16;

  1874. 	crv= (crv*contrast * saturation)>>32;

  1875. 	cbu= (cbu*contrast * saturation)>>32;

  1876. 	cgu= (cgu*contrast * saturation)>>32;

  1877. 	cgv= (cgv*contrast * saturation)>>32;

  1878. 

  1879. 	oy -= 256*brightness;

  1880. 

  1881. 	c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;

  1882. 	c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;

  1883. 	c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;

  1884. 	c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;

  1885. 	c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;

  1886. 	c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;

  1887. 

  1888. 	yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);

  1889. 	//FIXME factorize

  1890. 

  1891. #ifdef COMPILE_ALTIVEC

  1892. 	if (c->flags & SWS_CPU_CAPS_ALTIVEC)

  1893. 	    yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);

  1894. #endif	

  1895. 	return 0;

  1896. }

  1897. 

  1898. /**

  1899.  * @return -1 if not supported

  1900.  */

  1901. int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){

  1902. 	if(isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  1903. 

  1904. 	*inv_table = c->srcColorspaceTable;

  1905. 	*table     = c->dstColorspaceTable;

  1906. 	*srcRange  = c->srcRange;

  1907. 	*dstRange  = c->dstRange;

  1908. 	*brightness= c->brightness;

  1909. 	*contrast  = c->contrast;

  1910. 	*saturation= c->saturation;

  1911. 	

  1912. 	return 0;	

  1913. }

  1914. 

  1915. static int handle_jpeg(int *format)

  1916. {

  1917. 	switch (*format) {

  1918. 		case PIX_FMT_YUVJ420P:

  1919. 			*format = PIX_FMT_YUV420P;

  1920. 			return 1;

  1921. 		case PIX_FMT_YUVJ422P:

  1922. 			*format = PIX_FMT_YUV422P;

  1923. 			return 1;

  1924. 		case PIX_FMT_YUVJ444P:

  1925. 			*format = PIX_FMT_YUV444P;

  1926. 			return 1;

  1927. 		default:

  1928. 			return 0;

  1929. 	}

  1930. }

  1931. 

  1932. SwsContext *sws_getContext(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat, int flags,

  1933.                          SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){

  1934. 

  1935. 	SwsContext *c;

  1936. 	int i;

  1937. 	int usesVFilter, usesHFilter;

  1938. 	int unscaled, needsDither;

  1939. 	int srcRange, dstRange;

  1940. 	SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};

  1941. #if defined(ARCH_X86)

  1942. 	if(flags & SWS_CPU_CAPS_MMX)

  1943. 		asm volatile("emms\n\t"::: "memory");

  1944. #endif

  1945. 

  1946. #if !defined(RUNTIME_CPUDETECT) || !defined (CONFIG_GPL) //ensure that the flags match the compiled variant if cpudetect is off

  1947. 	flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC);

  1948. #ifdef HAVE_MMX2

  1949. 	flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  1950. #elif defined (HAVE_3DNOW)

  1951. 	flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  1952. #elif defined (HAVE_MMX)

  1953. 	flags |= SWS_CPU_CAPS_MMX;

  1954. #elif defined (HAVE_ALTIVEC)

  1955. 	flags |= SWS_CPU_CAPS_ALTIVEC;

  1956. #endif

  1957. #endif /* RUNTIME_CPUDETECT */

  1958. 	if(clip_table[512] != 255) globalInit();

  1959. 	if(rgb15to16 == NULL) sws_rgb2rgb_init(flags);

  1960. 

  1961. 	unscaled = (srcW == dstW && srcH == dstH);

  1962. 	needsDither= (isBGR(dstFormat) || isRGB(dstFormat)) 

  1963. 		     && (fmt_depth(dstFormat))<24

  1964. 		     && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

  1965. 

  1966. 	srcRange = handle_jpeg(&srcFormat);

  1967. 	dstRange = handle_jpeg(&dstFormat);

  1968. 

  1969. 	if(!isSupportedIn(srcFormat)) 

  1970. 	{

  1971. 		MSG_ERR("swScaler: %s is not supported as input format\n", sws_format_name(srcFormat));

  1972. 		return NULL;

  1973. 	}

  1974. 	if(!isSupportedOut(dstFormat))

  1975. 	{

  1976. 		MSG_ERR("swScaler: %s is not supported as output format\n", sws_format_name(dstFormat));

  1977. 		return NULL;

  1978. 	}

  1979. 

  1980. 	/* sanity check */

  1981. 	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

  1982. 	{

  1983. 		 MSG_ERR("swScaler: %dx%d -> %dx%d is invalid scaling dimension\n", 

  1984. 			srcW, srcH, dstW, dstH);

  1985. 		return NULL;

  1986. 	}

  1987. 

  1988. 	if(!dstFilter) dstFilter= &dummyFilter;

  1989. 	if(!srcFilter) srcFilter= &dummyFilter;

  1990. 

  1991. 	c= av_malloc(sizeof(SwsContext));

  1992. 	memset(c, 0, sizeof(SwsContext));

  1993. 

  1994. 	c->srcW= srcW;

  1995. 	c->srcH= srcH;

  1996. 	c->dstW= dstW;

  1997. 	c->dstH= dstH;

  1998. 	c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;

  1999. 	c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;

  2000. 	c->flags= flags;

  2001. 	c->dstFormat= dstFormat;

  2002. 	c->srcFormat= srcFormat;

  2003.         c->vRounder= 4* 0x0001000100010001ULL;

  2004. 

  2005. 	usesHFilter= usesVFilter= 0;

  2006. 	if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesVFilter=1;

  2007. 	if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesHFilter=1;

  2008. 	if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesVFilter=1;

  2009. 	if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesHFilter=1;

  2010. 	if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesVFilter=1;

  2011. 	if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesHFilter=1;

  2012. 	if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesVFilter=1;

  2013. 	if(srcFilter->chrH!=NULL && srcFilter->chrH->length>1) usesHFilter=1;

  2014. 

  2015. 	getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);

  2016. 	getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

  2017. 

  2018. 	// reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation

  2019. 	if((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

  2020. 

  2021. 	// drop some chroma lines if the user wants it

  2022. 	c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;

  2023. 	c->chrSrcVSubSample+= c->vChrDrop;

  2024. 

  2025. 	// drop every 2. pixel for chroma calculation unless user wants full chroma

  2026. 	if((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)) 

  2027. 		c->chrSrcHSubSample=1;

  2028. 

  2029. 	if(param){

  2030. 		c->param[0] = param[0];

  2031. 		c->param[1] = param[1];

  2032. 	}else{

  2033. 		c->param[0] =

  2034. 		c->param[1] = SWS_PARAM_DEFAULT;

  2035. 	}

  2036. 

  2037. 	c->chrIntHSubSample= c->chrDstHSubSample;

  2038. 	c->chrIntVSubSample= c->chrSrcVSubSample;

  2039. 

  2040. 	// note the -((-x)>>y) is so that we allways round toward +inf

  2041. 	c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);

  2042. 	c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);

  2043. 	c->chrDstW= -((-dstW) >> c->chrDstHSubSample);

  2044. 	c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

  2045. 

  2046. 	sws_setColorspaceDetails(c, Inverse_Table_6_9[SWS_CS_DEFAULT], srcRange, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16); 

  2047. 

  2048. 	/* unscaled special Cases */

  2049. 	if(unscaled && !usesHFilter && !usesVFilter)

  2050. 	{

  2051. 		/* yv12_to_nv12 */

  2052. 		if(srcFormat == PIX_FMT_YUV420P && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))

  2053. 		{

  2054. 			c->swScale= PlanarToNV12Wrapper;

  2055. 		}

  2056. #ifdef CONFIG_GPL

  2057. 		/* yuv2bgr */

  2058. 		if((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat)))

  2059. 		{

  2060. 			c->swScale= yuv2rgb_get_func_ptr(c);

  2061. 		}

  2062. #endif

  2063. 		

  2064. 		if( srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P )

  2065. 		{

  2066. 			c->swScale= yvu9toyv12Wrapper;

  2067. 		}

  2068. 

  2069. 		/* bgr24toYV12 */

  2070. 		if(srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P)

  2071. 			c->swScale= bgr24toyv12Wrapper;

  2072. 		

  2073. 		/* rgb/bgr -> rgb/bgr (no dither needed forms) */

  2074. 		if(   (isBGR(srcFormat) || isRGB(srcFormat))

  2075. 		   && (isBGR(dstFormat) || isRGB(dstFormat)) 

  2076. 		   && !needsDither)

  2077. 			c->swScale= rgb2rgbWrapper;

  2078. 

  2079. 		/* LQ converters if -sws 0 or -sws 4*/

  2080. 		if(c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){

  2081. 			/* rgb/bgr -> rgb/bgr (dither needed forms) */

  2082. 			if(  (isBGR(srcFormat) || isRGB(srcFormat))

  2083. 			  && (isBGR(dstFormat) || isRGB(dstFormat)) 

  2084. 			  && needsDither)

  2085. 				c->swScale= rgb2rgbWrapper;

  2086. 

  2087. 			/* yv12_to_yuy2 */

  2088. 			if(srcFormat == PIX_FMT_YUV420P && 

  2089. 			    (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422))

  2090. 			{

  2091. 				if (dstFormat == PIX_FMT_YUYV422)

  2092. 				    c->swScale= PlanarToYuy2Wrapper;

  2093. 				else

  2094. 				    c->swScale= PlanarToUyvyWrapper;

  2095. 			}

  2096. 		}

  2097. 

  2098. #ifdef COMPILE_ALTIVEC

  2099. 		if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  2100. 		    ((srcFormat == PIX_FMT_YUV420P && 

  2101. 		      (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422)))) {

  2102. 		  // unscaled YV12 -> packed YUV, we want speed

  2103. 		  if (dstFormat == PIX_FMT_YUYV422)

  2104. 		    c->swScale= yv12toyuy2_unscaled_altivec;

  2105. 		  else

  2106. 		    c->swScale= yv12touyvy_unscaled_altivec;

  2107. 		}

  2108. #endif

  2109. 

  2110. 		/* simple copy */

  2111. 		if(   srcFormat == dstFormat

  2112. 		   || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  2113. 		   || (isPlanarYUV(dstFormat) && isGray(srcFormat))

  2114. 		  )

  2115. 		{

  2116. 			c->swScale= simpleCopy;

  2117. 		}

  2118. 

  2119. 		/* gray16{le,be} conversions */

  2120. 		if(isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))

  2121. 		{

  2122. 			c->swScale= gray16togray;

  2123. 		}

  2124. 		if((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))

  2125. 		{

  2126. 			c->swScale= graytogray16;

  2127. 		}

  2128. 		if(srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))

  2129. 		{

  2130. 			c->swScale= gray16swap;

  2131. 		}		

  2132. 

  2133. 		if(c->swScale){

  2134. 			if(flags&SWS_PRINT_INFO)

  2135. 				MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", 

  2136. 					sws_format_name(srcFormat), sws_format_name(dstFormat));

  2137. 			return c;

  2138. 		}

  2139. 	}

  2140. 

  2141. 	if(flags & SWS_CPU_CAPS_MMX2)

  2142. 	{

  2143. 		c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;

  2144. 		if(!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))

  2145. 		{

  2146. 			if(flags&SWS_PRINT_INFO)

  2147. 				MSG_INFO("SwScaler: output Width is not a multiple of 32 -> no MMX2 scaler\n");

  2148. 		}

  2149. 		if(usesHFilter) c->canMMX2BeUsed=0;

  2150. 	}

  2151. 	else

  2152. 		c->canMMX2BeUsed=0;

  2153. 

  2154. 	c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;

  2155. 	c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

  2156. 

  2157. 	// match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst

  2158. 	// but only for the FAST_BILINEAR mode otherwise do correct scaling

  2159. 	// n-2 is the last chrominance sample available

  2160. 	// this is not perfect, but noone shuld notice the difference, the more correct variant

  2161. 	// would be like the vertical one, but that would require some special code for the

  2162. 	// first and last pixel

  2163. 	if(flags&SWS_FAST_BILINEAR)

  2164. 	{

  2165. 		if(c->canMMX2BeUsed)

  2166. 		{

  2167. 			c->lumXInc+= 20;

  2168. 			c->chrXInc+= 20;

  2169. 		}

  2170. 		//we don't use the x86asm scaler if mmx is available

  2171. 		else if(flags & SWS_CPU_CAPS_MMX)

  2172. 		{

  2173. 			c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;

  2174. 			c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;

  2175. 		}

  2176. 	}

  2177. 

  2178. 	/* precalculate horizontal scaler filter coefficients */

  2179. 	{

  2180. 		const int filterAlign=

  2181. 		  (flags & SWS_CPU_CAPS_MMX) ? 4 :

  2182. 		  (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2183. 		  1;

  2184. 

  2185. 		initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,

  2186. 				 srcW      ,       dstW, filterAlign, 1<<14,

  2187. 				 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2188. 				 srcFilter->lumH, dstFilter->lumH, c->param);

  2189. 		initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,

  2190. 				 c->chrSrcW, c->chrDstW, filterAlign, 1<<14,

  2191. 				 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2192. 				 srcFilter->chrH, dstFilter->chrH, c->param);

  2193. 

  2194. #define MAX_FUNNY_CODE_SIZE 10000

  2195. #if defined(COMPILE_MMX2)

  2196. // can't downscale !!!

  2197. 		if(c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))

  2198. 		{

  2199. #ifdef MAP_ANONYMOUS

  2200. 			c->funnyYCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2201. 			c->funnyUVCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2202. #else

  2203. 			c->funnyYCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2204. 			c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2205. #endif

  2206. 

  2207. 			c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));

  2208. 			c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));

  2209. 			c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));

  2210. 			c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));

  2211. 

  2212. 			initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);

  2213. 			initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);

  2214. 		}

  2215. #endif /* defined(COMPILE_MMX2) */

  2216. 	} // Init Horizontal stuff

  2217. 

  2218. 

  2219. 

  2220. 	/* precalculate vertical scaler filter coefficients */

  2221. 	{

  2222. 		const int filterAlign=

  2223. 		  (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :

  2224. 		  (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2225. 		  1;

  2226. 

  2227. 		initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,

  2228. 				srcH      ,        dstH, filterAlign, (1<<12)-4,

  2229. 				(flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2230. 				srcFilter->lumV, dstFilter->lumV, c->param);

  2231. 		initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,

  2232. 				c->chrSrcH, c->chrDstH, filterAlign, (1<<12)-4,

  2233. 				(flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2234. 				srcFilter->chrV, dstFilter->chrV, c->param);

  2235. 

  2236. #ifdef HAVE_ALTIVEC

  2237. 		c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);

  2238. 		c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);

  2239. 

  2240. 		for (i=0;i<c->vLumFilterSize*c->dstH;i++) {

  2241.                   int j;

  2242. 		  short *p = (short *)&c->vYCoeffsBank[i];

  2243. 		  for (j=0;j<8;j++)

  2244. 		    p[j] = c->vLumFilter[i];

  2245. 		}

  2246. 

  2247. 		for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {

  2248.                   int j;

  2249. 		  short *p = (short *)&c->vCCoeffsBank[i];

  2250. 		  for (j=0;j<8;j++)

  2251. 		    p[j] = c->vChrFilter[i];

  2252. 		}

  2253. #endif

  2254. 	}

  2255. 

  2256. 	// Calculate Buffer Sizes so that they won't run out while handling these damn slices

  2257. 	c->vLumBufSize= c->vLumFilterSize;

  2258. 	c->vChrBufSize= c->vChrFilterSize;

  2259. 	for(i=0; i<dstH; i++)

  2260. 	{

  2261. 		int chrI= i*c->chrDstH / dstH;

  2262. 		int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,

  2263. 				 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

  2264. 

  2265. 		nextSlice>>= c->chrSrcVSubSample;

  2266. 		nextSlice<<= c->chrSrcVSubSample;

  2267. 		if(c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)

  2268. 			c->vLumBufSize= nextSlice - c->vLumFilterPos[i   ];

  2269. 		if(c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))

  2270. 			c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];

  2271. 	}

  2272. 

  2273. 	// allocate pixbufs (we use dynamic allocation because otherwise we would need to

  2274. 	c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));

  2275. 	c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));

  2276. 	//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)

  2277. 	/* align at 16 bytes for AltiVec */

  2278. 	for(i=0; i<c->vLumBufSize; i++)

  2279. 		c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_malloc(4000);

  2280. 	for(i=0; i<c->vChrBufSize; i++)

  2281. 		c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc(8000);

  2282. 

  2283. 	//try to avoid drawing green stuff between the right end and the stride end

  2284. 	for(i=0; i<c->vLumBufSize; i++) memset(c->lumPixBuf[i], 0, 4000);

  2285. 	for(i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, 8000);

  2286. 

  2287. 	ASSERT(c->chrDstH <= dstH)

  2288. 

  2289. 	if(flags&SWS_PRINT_INFO)

  2290. 	{

  2291. #ifdef DITHER1XBPP

  2292. 		char *dither= " dithered";

  2293. #else

  2294. 		char *dither= "";

  2295. #endif

  2296. 		if(flags&SWS_FAST_BILINEAR)

  2297. 			MSG_INFO("SwScaler: FAST_BILINEAR scaler, ");

  2298. 		else if(flags&SWS_BILINEAR)

  2299. 			MSG_INFO("SwScaler: BILINEAR scaler, ");

  2300. 		else if(flags&SWS_BICUBIC)

  2301. 			MSG_INFO("SwScaler: BICUBIC scaler, ");

  2302. 		else if(flags&SWS_X)

  2303. 			MSG_INFO("SwScaler: Experimental scaler, ");

  2304. 		else if(flags&SWS_POINT)

  2305. 			MSG_INFO("SwScaler: Nearest Neighbor / POINT scaler, ");

  2306. 		else if(flags&SWS_AREA)

  2307. 			MSG_INFO("SwScaler: Area Averageing scaler, ");

  2308. 		else if(flags&SWS_BICUBLIN)

  2309. 			MSG_INFO("SwScaler: luma BICUBIC / chroma BILINEAR scaler, ");

  2310. 		else if(flags&SWS_GAUSS)

  2311. 			MSG_INFO("SwScaler: Gaussian scaler, ");

  2312. 		else if(flags&SWS_SINC)

  2313. 			MSG_INFO("SwScaler: Sinc scaler, ");

  2314. 		else if(flags&SWS_LANCZOS)

  2315. 			MSG_INFO("SwScaler: Lanczos scaler, ");

  2316. 		else if(flags&SWS_SPLINE)

  2317. 			MSG_INFO("SwScaler: Bicubic spline scaler, ");

  2318. 		else

  2319. 			MSG_INFO("SwScaler: ehh flags invalid?! ");

  2320. 

  2321. 		if(dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)

  2322. 			MSG_INFO("from %s to%s %s ", 

  2323. 				sws_format_name(srcFormat), dither, sws_format_name(dstFormat));

  2324. 		else

  2325. 			MSG_INFO("from %s to %s ", 

  2326. 				sws_format_name(srcFormat), sws_format_name(dstFormat));

  2327. 

  2328. 		if(flags & SWS_CPU_CAPS_MMX2)

  2329. 			MSG_INFO("using MMX2\n");

  2330. 		else if(flags & SWS_CPU_CAPS_3DNOW)

  2331. 			MSG_INFO("using 3DNOW\n");

  2332. 		else if(flags & SWS_CPU_CAPS_MMX)

  2333. 			MSG_INFO("using MMX\n");

  2334. 		else if(flags & SWS_CPU_CAPS_ALTIVEC)

  2335. 			MSG_INFO("using AltiVec\n");

  2336. 		else 

  2337. 			MSG_INFO("using C\n");

  2338. 	}

  2339. 

  2340. 	if(flags & SWS_PRINT_INFO)

  2341. 	{

  2342. 		if(flags & SWS_CPU_CAPS_MMX)

  2343. 		{

  2344. 			if(c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))

  2345. 				MSG_V("SwScaler: using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");

  2346. 			else

  2347. 			{

  2348. 				if(c->hLumFilterSize==4)

  2349. 					MSG_V("SwScaler: using 4-tap MMX scaler for horizontal luminance scaling\n");

  2350. 				else if(c->hLumFilterSize==8)

  2351. 					MSG_V("SwScaler: using 8-tap MMX scaler for horizontal luminance scaling\n");

  2352. 				else

  2353. 					MSG_V("SwScaler: using n-tap MMX scaler for horizontal luminance scaling\n");

  2354. 

  2355. 				if(c->hChrFilterSize==4)

  2356. 					MSG_V("SwScaler: using 4-tap MMX scaler for horizontal chrominance scaling\n");

  2357. 				else if(c->hChrFilterSize==8)

  2358. 					MSG_V("SwScaler: using 8-tap MMX scaler for horizontal chrominance scaling\n");

  2359. 				else

  2360. 					MSG_V("SwScaler: using n-tap MMX scaler for horizontal chrominance scaling\n");

  2361. 			}

  2362. 		}

  2363. 		else

  2364. 		{

  2365. #if defined(ARCH_X86)

  2366. 			MSG_V("SwScaler: using X86-Asm scaler for horizontal scaling\n");

  2367. #else

  2368. 			if(flags & SWS_FAST_BILINEAR)

  2369. 				MSG_V("SwScaler: using FAST_BILINEAR C scaler for horizontal scaling\n");

  2370. 			else

  2371. 				MSG_V("SwScaler: using C scaler for horizontal scaling\n");

  2372. #endif

  2373. 		}

  2374. 		if(isPlanarYUV(dstFormat))

  2375. 		{

  2376. 			if(c->vLumFilterSize==1)

  2377. 				MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2378. 			else

  2379. 				MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2380. 		}

  2381. 		else

  2382. 		{

  2383. 			if(c->vLumFilterSize==1 && c->vChrFilterSize==2)

  2384. 				MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"

  2385. 				       "SwScaler:       2-tap scaler for vertical chrominance scaling (BGR)\n",(flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2386. 			else if(c->vLumFilterSize==2 && c->vChrFilterSize==2)

  2387. 				MSG_V("SwScaler: using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2388. 			else

  2389. 				MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2390. 		}

  2391. 

  2392. 		if(dstFormat==PIX_FMT_BGR24)

  2393. 			MSG_V("SwScaler: using %s YV12->BGR24 Converter\n",

  2394. 				(flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));

  2395. 		else if(dstFormat==PIX_FMT_RGB32)

  2396. 			MSG_V("SwScaler: using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2397. 		else if(dstFormat==PIX_FMT_BGR565)

  2398. 			MSG_V("SwScaler: using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2399. 		else if(dstFormat==PIX_FMT_BGR555)

  2400. 			MSG_V("SwScaler: using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2401. 

  2402. 		MSG_V("SwScaler: %dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);

  2403. 	}

  2404. 	if(flags & SWS_PRINT_INFO)

  2405. 	{

  2406. 		MSG_DBG2("SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2407. 			c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);

  2408. 		MSG_DBG2("SwScaler:Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2409. 			c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);

  2410. 	}

  2411. 

  2412. 	c->swScale= getSwsFunc(flags);

  2413. 	return c;

  2414. }

  2415. 

  2416. /**

  2417.  * swscale warper, so we don't need to export the SwsContext.

  2418.  * assumes planar YUV to be in YUV order instead of YVU

  2419.  */

  2420. int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2421.                            int srcSliceH, uint8_t* dst[], int dstStride[]){

  2422. 	if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  2423. 	    MSG_ERR("swScaler: slices start in the middle!\n");

  2424. 	    return 0;

  2425. 	}

  2426. 	if (c->sliceDir == 0) {

  2427. 	    if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  2428. 	}

  2429. 

  2430. 	// copy strides, so they can safely be modified

  2431. 	if (c->sliceDir == 1) {

  2432. 	    // slices go from top to bottom

  2433. 	    int srcStride2[3]= {srcStride[0], srcStride[1], srcStride[2]};

  2434. 	    int dstStride2[3]= {dstStride[0], dstStride[1], dstStride[2]};

  2435. 	    return c->swScale(c, src, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);

  2436. 	} else {

  2437. 	    // slices go from bottom to top => we flip the image internally

  2438. 	    uint8_t* src2[3]= {src[0] + (srcSliceH-1)*srcStride[0],

  2439. 			       src[1] + ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1],

  2440. 			       src[2] + ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2]

  2441. 	    };

  2442. 	    uint8_t* dst2[3]= {dst[0] + (c->dstH-1)*dstStride[0],

  2443. 			       dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],

  2444. 			       dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2]};

  2445. 	    int srcStride2[3]= {-srcStride[0], -srcStride[1], -srcStride[2]};

  2446. 	    int dstStride2[3]= {-dstStride[0], -dstStride[1], -dstStride[2]};

  2447. 	    

  2448. 	    return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  2449. 	}

  2450. }

  2451. 

  2452. /**

  2453.  * swscale warper, so we don't need to export the SwsContext

  2454.  */

  2455. int sws_scale(SwsContext *c, uint8_t* srcParam[], int srcStride[], int srcSliceY,

  2456.                            int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  2457. 	uint8_t *src[3];

  2458. 	uint8_t *dst[3];

  2459. 	src[0] = srcParam[0]; src[1] = srcParam[1]; src[2] = srcParam[2];

  2460. 	dst[0] = dstParam[0]; dst[1] = dstParam[1]; dst[2] = dstParam[2];

  2461. //printf("sws: slice %d %d\n", srcSliceY, srcSliceH);

  2462. 

  2463. 	return c->swScale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);

  2464. }

  2465. 

  2466. SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur, 

  2467. 				float lumaSharpen, float chromaSharpen,

  2468. 				float chromaHShift, float chromaVShift,

  2469. 				int verbose)

  2470. {

  2471. 	SwsFilter *filter= av_malloc(sizeof(SwsFilter));

  2472. 

  2473. 	if(lumaGBlur!=0.0){

  2474. 		filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);

  2475. 		filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);

  2476. 	}else{

  2477. 		filter->lumH= sws_getIdentityVec();

  2478. 		filter->lumV= sws_getIdentityVec();

  2479. 	}

  2480. 

  2481. 	if(chromaGBlur!=0.0){

  2482. 		filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);

  2483. 		filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);

  2484. 	}else{

  2485. 		filter->chrH= sws_getIdentityVec();

  2486. 		filter->chrV= sws_getIdentityVec();

  2487. 	}

  2488. 

  2489. 	if(chromaSharpen!=0.0){

  2490. 		SwsVector *id= sws_getIdentityVec();

  2491.                 sws_scaleVec(filter->chrH, -chromaSharpen);

  2492.                 sws_scaleVec(filter->chrV, -chromaSharpen);

  2493. 		sws_addVec(filter->chrH, id);

  2494. 		sws_addVec(filter->chrV, id);

  2495. 		sws_freeVec(id);

  2496. 	}

  2497. 

  2498. 	if(lumaSharpen!=0.0){

  2499. 		SwsVector *id= sws_getIdentityVec();

  2500.                 sws_scaleVec(filter->lumH, -lumaSharpen);

  2501.                 sws_scaleVec(filter->lumV, -lumaSharpen);

  2502. 		sws_addVec(filter->lumH, id);

  2503. 		sws_addVec(filter->lumV, id);

  2504. 		sws_freeVec(id);

  2505. 	}

  2506. 

  2507. 	if(chromaHShift != 0.0)

  2508. 		sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));

  2509. 

  2510. 	if(chromaVShift != 0.0)

  2511. 		sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));

  2512. 

  2513. 	sws_normalizeVec(filter->chrH, 1.0);

  2514. 	sws_normalizeVec(filter->chrV, 1.0);

  2515. 	sws_normalizeVec(filter->lumH, 1.0);

  2516. 	sws_normalizeVec(filter->lumV, 1.0);

  2517. 

  2518. 	if(verbose) sws_printVec(filter->chrH);

  2519. 	if(verbose) sws_printVec(filter->lumH);

  2520. 

  2521.         return filter;

  2522. }

  2523. 

  2524. /**

  2525.  * returns a normalized gaussian curve used to filter stuff

  2526.  * quality=3 is high quality, lowwer is lowwer quality

  2527.  */

  2528. SwsVector *sws_getGaussianVec(double variance, double quality){

  2529. 	const int length= (int)(variance*quality + 0.5) | 1;

  2530. 	int i;

  2531. 	double *coeff= av_malloc(length*sizeof(double));

  2532. 	double middle= (length-1)*0.5;

  2533. 	SwsVector *vec= av_malloc(sizeof(SwsVector));

  2534. 

  2535. 	vec->coeff= coeff;

  2536. 	vec->length= length;

  2537. 

  2538. 	for(i=0; i<length; i++)

  2539. 	{

  2540. 		double dist= i-middle;

  2541. 		coeff[i]= exp( -dist*dist/(2*variance*variance) ) / sqrt(2*variance*PI);

  2542. 	}

  2543. 

  2544. 	sws_normalizeVec(vec, 1.0);

  2545. 

  2546. 	return vec;

  2547. }

  2548. 

  2549. SwsVector *sws_getConstVec(double c, int length){

  2550. 	int i;

  2551. 	double *coeff= av_malloc(length*sizeof(double));

  2552. 	SwsVector *vec= av_malloc(sizeof(SwsVector));

  2553. 

  2554. 	vec->coeff= coeff;

  2555. 	vec->length= length;

  2556. 

  2557. 	for(i=0; i<length; i++)

  2558. 		coeff[i]= c;

  2559. 

  2560. 	return vec;

  2561. }

  2562. 

  2563. 

  2564. SwsVector *sws_getIdentityVec(void){

  2565.         return sws_getConstVec(1.0, 1);

  2566. }

  2567. 

  2568. double sws_dcVec(SwsVector *a){

  2569. 	int i;

  2570.         double sum=0;

  2571. 

  2572. 	for(i=0; i<a->length; i++)

  2573. 		sum+= a->coeff[i];

  2574. 

  2575.         return sum;

  2576. }

  2577. 

  2578. void sws_scaleVec(SwsVector *a, double scalar){

  2579. 	int i;

  2580. 

  2581. 	for(i=0; i<a->length; i++)

  2582. 		a->coeff[i]*= scalar;

  2583. }

  2584. 

  2585. void sws_normalizeVec(SwsVector *a, double height){

  2586.         sws_scaleVec(a, height/sws_dcVec(a));

  2587. }

  2588. 

  2589. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){

  2590. 	int length= a->length + b->length - 1;

  2591. 	double *coeff= av_malloc(length*sizeof(double));

  2592. 	int i, j;

  2593. 	SwsVector *vec= av_malloc(sizeof(SwsVector));

  2594. 

  2595. 	vec->coeff= coeff;

  2596. 	vec->length= length;

  2597. 

  2598. 	for(i=0; i<length; i++) coeff[i]= 0.0;

  2599. 

  2600. 	for(i=0; i<a->length; i++)

  2601. 	{

  2602. 		for(j=0; j<b->length; j++)

  2603. 		{

  2604. 			coeff[i+j]+= a->coeff[i]*b->coeff[j];

  2605. 		}

  2606. 	}

  2607. 

  2608. 	return vec;

  2609. }

  2610. 

  2611. static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){

  2612. 	int length= FFMAX(a->length, b->length);

  2613. 	double *coeff= av_malloc(length*sizeof(double));

  2614. 	int i;

  2615. 	SwsVector *vec= av_malloc(sizeof(SwsVector));

  2616. 

  2617. 	vec->coeff= coeff;

  2618. 	vec->length= length;

  2619. 

  2620. 	for(i=0; i<length; i++) coeff[i]= 0.0;

  2621. 

  2622. 	for(i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  2623. 	for(i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];

  2624. 

  2625. 	return vec;

  2626. }

  2627. 

  2628. static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){

  2629. 	int length= FFMAX(a->length, b->length);

  2630. 	double *coeff= av_malloc(length*sizeof(double));

  2631. 	int i;

  2632. 	SwsVector *vec= av_malloc(sizeof(SwsVector));

  2633. 

  2634. 	vec->coeff= coeff;

  2635. 	vec->length= length;

  2636. 

  2637. 	for(i=0; i<length; i++) coeff[i]= 0.0;

  2638. 

  2639. 	for(i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  2640. 	for(i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];

  2641. 

  2642. 	return vec;

  2643. }

  2644. 

  2645. /* shift left / or right if "shift" is negative */

  2646. static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){

  2647. 	int length= a->length + FFABS(shift)*2;

  2648. 	double *coeff= av_malloc(length*sizeof(double));

  2649. 	int i;

  2650. 	SwsVector *vec= av_malloc(sizeof(SwsVector));

  2651. 

  2652. 	vec->coeff= coeff;

  2653. 	vec->length= length;

  2654. 

  2655. 	for(i=0; i<length; i++) coeff[i]= 0.0;

  2656. 

  2657. 	for(i=0; i<a->length; i++)

  2658. 	{

  2659. 		coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];

  2660. 	}

  2661. 

  2662. 	return vec;

  2663. }

  2664. 

  2665. void sws_shiftVec(SwsVector *a, int shift){

  2666. 	SwsVector *shifted= sws_getShiftedVec(a, shift);

  2667. 	av_free(a->coeff);

  2668. 	a->coeff= shifted->coeff;

  2669. 	a->length= shifted->length;

  2670. 	av_free(shifted);

  2671. }

  2672. 

  2673. void sws_addVec(SwsVector *a, SwsVector *b){

  2674. 	SwsVector *sum= sws_sumVec(a, b);

  2675. 	av_free(a->coeff);

  2676. 	a->coeff= sum->coeff;

  2677. 	a->length= sum->length;

  2678. 	av_free(sum);

  2679. }

  2680. 

  2681. void sws_subVec(SwsVector *a, SwsVector *b){

  2682. 	SwsVector *diff= sws_diffVec(a, b);

  2683. 	av_free(a->coeff);

  2684. 	a->coeff= diff->coeff;

  2685. 	a->length= diff->length;

  2686. 	av_free(diff);

  2687. }

  2688. 

  2689. void sws_convVec(SwsVector *a, SwsVector *b){

  2690. 	SwsVector *conv= sws_getConvVec(a, b);

  2691. 	av_free(a->coeff);  

  2692. 	a->coeff= conv->coeff;

  2693. 	a->length= conv->length;

  2694. 	av_free(conv);

  2695. }

  2696. 

  2697. SwsVector *sws_cloneVec(SwsVector *a){

  2698. 	double *coeff= av_malloc(a->length*sizeof(double));

  2699. 	int i;

  2700. 	SwsVector *vec= av_malloc(sizeof(SwsVector));

  2701. 

  2702. 	vec->coeff= coeff;

  2703. 	vec->length= a->length;

  2704. 

  2705. 	for(i=0; i<a->length; i++) coeff[i]= a->coeff[i];

  2706. 

  2707. 	return vec;

  2708. }

  2709. 

  2710. void sws_printVec(SwsVector *a){

  2711. 	int i;

  2712. 	double max=0;

  2713. 	double min=0;

  2714. 	double range;

  2715. 

  2716. 	for(i=0; i<a->length; i++)

  2717. 		if(a->coeff[i]>max) max= a->coeff[i];

  2718. 

  2719. 	for(i=0; i<a->length; i++)

  2720. 		if(a->coeff[i]<min) min= a->coeff[i];

  2721. 

  2722. 	range= max - min;

  2723. 

  2724. 	for(i=0; i<a->length; i++)

  2725. 	{

  2726. 		int x= (int)((a->coeff[i]-min)*60.0/range +0.5);

  2727. 		MSG_DBG2("%1.3f ", a->coeff[i]);

  2728. 		for(;x>0; x--) MSG_DBG2(" ");

  2729. 		MSG_DBG2("|\n");

  2730. 	}

  2731. }

  2732. 

  2733. void sws_freeVec(SwsVector *a){

  2734. 	if(!a) return;

  2735. 	av_free(a->coeff);

  2736. 	a->coeff=NULL;

  2737. 	a->length=0;

  2738. 	av_free(a);

  2739. }

  2740. 

  2741. void sws_freeFilter(SwsFilter *filter){

  2742. 	if(!filter) return;

  2743. 

  2744. 	if(filter->lumH) sws_freeVec(filter->lumH);

  2745. 	if(filter->lumV) sws_freeVec(filter->lumV);

  2746. 	if(filter->chrH) sws_freeVec(filter->chrH);

  2747. 	if(filter->chrV) sws_freeVec(filter->chrV);

  2748. 	av_free(filter);

  2749. }

  2750. 

  2751. 

  2752. void sws_freeContext(SwsContext *c){

  2753. 	int i;

  2754. 	if(!c) return;

  2755. 

  2756. 	if(c->lumPixBuf)

  2757. 	{

  2758. 		for(i=0; i<c->vLumBufSize; i++)

  2759. 		{

  2760. 			av_free(c->lumPixBuf[i]);

  2761. 			c->lumPixBuf[i]=NULL;

  2762. 		}

  2763. 		av_free(c->lumPixBuf);

  2764. 		c->lumPixBuf=NULL;

  2765. 	}

  2766. 

  2767. 	if(c->chrPixBuf)

  2768. 	{

  2769. 		for(i=0; i<c->vChrBufSize; i++)

  2770. 		{

  2771. 			av_free(c->chrPixBuf[i]);

  2772. 			c->chrPixBuf[i]=NULL;

  2773. 		}

  2774. 		av_free(c->chrPixBuf);

  2775. 		c->chrPixBuf=NULL;

  2776. 	}

  2777. 

  2778. 	av_free(c->vLumFilter);

  2779. 	c->vLumFilter = NULL;

  2780. 	av_free(c->vChrFilter);

  2781. 	c->vChrFilter = NULL;

  2782. 	av_free(c->hLumFilter);

  2783. 	c->hLumFilter = NULL;

  2784. 	av_free(c->hChrFilter);

  2785. 	c->hChrFilter = NULL;

  2786. #ifdef HAVE_ALTIVEC

  2787. 	av_free(c->vYCoeffsBank);

  2788. 	c->vYCoeffsBank = NULL;

  2789. 	av_free(c->vCCoeffsBank);

  2790. 	c->vCCoeffsBank = NULL;

  2791. #endif

  2792. 

  2793. 	av_free(c->vLumFilterPos);

  2794. 	c->vLumFilterPos = NULL;

  2795. 	av_free(c->vChrFilterPos);

  2796. 	c->vChrFilterPos = NULL;

  2797. 	av_free(c->hLumFilterPos);

  2798. 	c->hLumFilterPos = NULL;

  2799. 	av_free(c->hChrFilterPos);

  2800. 	c->hChrFilterPos = NULL;

  2801. 

  2802. #if defined(ARCH_X86) && defined(CONFIG_GPL)

  2803. #ifdef MAP_ANONYMOUS

  2804. 	if(c->funnyYCode) munmap(c->funnyYCode, MAX_FUNNY_CODE_SIZE);

  2805. 	if(c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);

  2806. #else

  2807. 	av_free(c->funnyYCode);

  2808. 	av_free(c->funnyUVCode);

  2809. #endif

  2810. 	c->funnyYCode=NULL;

  2811. 	c->funnyUVCode=NULL;

  2812. #endif /* defined(ARCH_X86) */

  2813. 

  2814. 	av_free(c->lumMmx2Filter);

  2815. 	c->lumMmx2Filter=NULL;

  2816. 	av_free(c->chrMmx2Filter);

  2817. 	c->chrMmx2Filter=NULL;

  2818. 	av_free(c->lumMmx2FilterPos);

  2819. 	c->lumMmx2FilterPos=NULL;

  2820. 	av_free(c->chrMmx2FilterPos);

  2821. 	c->chrMmx2FilterPos=NULL;

  2822. 	av_free(c->yuvTable);

  2823. 	c->yuvTable=NULL;

  2824. 

  2825. 	av_free(c);

  2826. }

  2827. 

  2828. /**

  2829.  * Checks if context is valid or reallocs a new one instead.

  2830.  * If context is NULL, just calls sws_getContext() to get a new one.

  2831.  * Otherwise, checks if the parameters are the same already saved in context.

  2832.  * If that is the case, returns the current context.

  2833.  * Otherwise, frees context and gets a new one.

  2834.  *

  2835.  * Be warned that srcFilter, dstFilter are not checked, they are

  2836.  * asumed to remain valid.

  2837.  */

  2838. struct SwsContext *sws_getCachedContext(struct SwsContext *context,

  2839.                         int srcW, int srcH, int srcFormat,

  2840.                         int dstW, int dstH, int dstFormat, int flags,

  2841.                         SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)

  2842. {

  2843.     if (context != NULL) {

  2844.         if ((context->srcW != srcW) || (context->srcH != srcH) ||

  2845.             (context->srcFormat != srcFormat) ||

  2846.             (context->dstW != dstW) || (context->dstH != dstH) ||

  2847.             (context->dstFormat != dstFormat) || (context->flags != flags) ||

  2848.             (context->param != param))

  2849.         {

  2850.             sws_freeContext(context);

  2851.             context = NULL;

  2852.         }

  2853.     }

  2854.     if (context == NULL) {

  2855.         return sws_getContext(srcW, srcH, srcFormat,

  2856.                         dstW, dstH, dstFormat, flags,

  2857.                         srcFilter, dstFilter, param);

  2858.     }

  2859.     return context;

  2860. }