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.
14622 Commits
32 Branches
748MB
Tree: bc5a04440b
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'bc5a04440b'
${ noResults }
FFmpeg/libswscale/swscale.c

2983 lines
101KB
Raw Blame History 

  1. /*

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

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

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

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

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

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

  10.  *

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

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

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

  14.  * GNU General Public License for more details.

  15.  *

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

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

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

  19.  *

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

  21.  * under the LGPL license too

  22.  */

  23. 

  24. /*

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

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

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

  28. 

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

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

  31.   x -> x

  32.   YUV9 -> YV12

  33.   YUV9/YV12 -> Y800

  34.   Y800 -> YUV9/YV12

  35.   BGR24 -> BGR32 & RGB24 -> RGB32

  36.   BGR32 -> BGR24 & RGB32 -> RGB24

  37.   BGR15 -> BGR16

  38. */

  39. 

  40. /*

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

  42.  YV12 -> BGR16

  43.  YV12 -> YV12

  44.  BGR15 -> BGR16

  45.  BGR16 -> BGR16

  46.  YVU9 -> YV12

  47. 

  48. untested special converters

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

  50.   YV12/I420 -> YV12/I420

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

  52.   BGR24 -> BGR32 & RGB24 -> RGB32

  53.   BGR32 -> BGR24 & RGB32 -> RGB24

  54.   BGR24 -> YV12

  55. */

  56. 

  57. #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 "rgb2rgb.h"

  73. #include "libavutil/x86_cpu.h"

  74. #include "libavutil/bswap.h"

  75. #include "libavcodec/opt.h"

  76. 

  77. #undef MOVNTQ

  78. #undef PAVGB

  79. 

  80. //#undef HAVE_MMX2

  81. //#define HAVE_3DNOW

  82. //#undef HAVE_MMX

  83. //#undef ARCH_X86

  84. //#define WORDS_BIGENDIAN

  85. #define DITHER1XBPP

  86. 

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

  88. 

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

  90. 

  91. #ifdef MP_DEBUG

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

  93. #else

  94. #define ASSERT(x) ;

  95. #endif

  96. 

  97. #ifdef M_PI

  98. #define PI M_PI

  99. #else

  100. #define PI 3.14159265358979323846

  101. #endif

  102. 

  103. #define isSupportedIn(x)    (       \

  104.            (x)==PIX_FMT_YUV420P     \

  105.         || (x)==PIX_FMT_YUVA420P    \

  106.         || (x)==PIX_FMT_YUYV422     \

  107.         || (x)==PIX_FMT_UYVY422     \

  108.         || (x)==PIX_FMT_RGB32       \

  109.         || (x)==PIX_FMT_BGR24       \

  110.         || (x)==PIX_FMT_BGR565      \

  111.         || (x)==PIX_FMT_BGR555      \

  112.         || (x)==PIX_FMT_BGR32       \

  113.         || (x)==PIX_FMT_RGB24       \

  114.         || (x)==PIX_FMT_RGB565      \

  115.         || (x)==PIX_FMT_RGB555      \

  116.         || (x)==PIX_FMT_GRAY8       \

  117.         || (x)==PIX_FMT_YUV410P     \

  118.         || (x)==PIX_FMT_GRAY16BE    \

  119.         || (x)==PIX_FMT_GRAY16LE    \

  120.         || (x)==PIX_FMT_YUV444P     \

  121.         || (x)==PIX_FMT_YUV422P     \

  122.         || (x)==PIX_FMT_YUV411P     \

  123.         || (x)==PIX_FMT_PAL8        \

  124.         || (x)==PIX_FMT_BGR8        \

  125.         || (x)==PIX_FMT_RGB8        \

  126.         || (x)==PIX_FMT_BGR4_BYTE   \

  127.         || (x)==PIX_FMT_RGB4_BYTE   \

  128.         || (x)==PIX_FMT_YUV440P     \

  129.     )

  130. #define isSupportedOut(x)   (       \

  131.            (x)==PIX_FMT_YUV420P     \

  132.         || (x)==PIX_FMT_YUYV422     \

  133.         || (x)==PIX_FMT_UYVY422     \

  134.         || (x)==PIX_FMT_YUV444P     \

  135.         || (x)==PIX_FMT_YUV422P     \

  136.         || (x)==PIX_FMT_YUV411P     \

  137.         || isRGB(x)                 \

  138.         || isBGR(x)                 \

  139.         || (x)==PIX_FMT_NV12        \

  140.         || (x)==PIX_FMT_NV21        \

  141.         || (x)==PIX_FMT_GRAY16BE    \

  142.         || (x)==PIX_FMT_GRAY16LE    \

  143.         || (x)==PIX_FMT_GRAY8       \

  144.         || (x)==PIX_FMT_YUV410P     \

  145.     )

  146. #define isPacked(x)         (       \

  147.            (x)==PIX_FMT_PAL8        \

  148.         || (x)==PIX_FMT_YUYV422     \

  149.         || (x)==PIX_FMT_UYVY422     \

  150.         || isRGB(x)                 \

  151.         || isBGR(x)                 \

  152.     )

  153. 

  154. #define RGB2YUV_SHIFT 16

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

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

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

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

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

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

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

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

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

  164. 

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

  166. 

  167. /*

  168. NOTES

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

  170. 

  171. TODO

  172. more intelligent misalignment avoidance for the horizontal scaler

  173. write special vertical cubic upscale version

  174. Optimize C code (yv12 / minmax)

  175. add support for packed pixel yuv input & output

  176. add support for Y8 output

  177. optimize bgr24 & bgr32

  178. add BGR4 output support

  179. write special BGR->BGR scaler

  180. */

  181. 

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

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

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

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

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

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

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

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

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

  191. 

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

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

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

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

  196. 

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

  198.         0x0103010301030103LL,

  199.         0x0200020002000200LL,};

  200. 

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

  202.         0x0602060206020602LL,

  203.         0x0004000400040004LL,};

  204. 

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

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

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

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

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

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

  211. 

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

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

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

  215. 

  216. #ifdef FAST_BGR2YV12

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

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

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

  220. #else

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

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

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

  224. #endif /* FAST_BGR2YV12 */

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

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

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

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

  229. 

  230. // clipping helper table for C implementations:

  231. static unsigned char clip_table[768];

  232. 

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

  234. 

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

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

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

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

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

  240. 

  241. static const char * sws_context_to_name(void * ptr) {

  242.     return "swscaler";

  243. }

  244. 

  245. #define OFFSET(x) offsetof(SwsContext, x)

  246. #define DEFAULT 0

  247. #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM

  248. 

  249. static const AVOption options[] = {

  250.     { "sws_flags", "scaler/cpu flags", OFFSET(flags), FF_OPT_TYPE_FLAGS, DEFAULT, 0, UINT_MAX, VE, "sws_flags" },

  251.     { "fast_bilinear", "fast bilinear", 0, FF_OPT_TYPE_CONST, SWS_FAST_BILINEAR, INT_MIN, INT_MAX, VE, "sws_flags" },

  252.     { "bilinear", "bilinear", 0, FF_OPT_TYPE_CONST, SWS_BILINEAR, INT_MIN, INT_MAX, VE, "sws_flags" },

  253.     { "bicubic", "bicubic", 0, FF_OPT_TYPE_CONST, SWS_BICUBIC, INT_MIN, INT_MAX, VE, "sws_flags" },

  254.     { "experimental", "experimental", 0, FF_OPT_TYPE_CONST, SWS_X, INT_MIN, INT_MAX, VE, "sws_flags" },

  255.     { "neighbor", "nearest neighbor", 0, FF_OPT_TYPE_CONST, SWS_POINT, INT_MIN, INT_MAX, VE, "sws_flags" },

  256.     { "area", "averaging area", 0, FF_OPT_TYPE_CONST, SWS_AREA, INT_MIN, INT_MAX, VE, "sws_flags" },

  257.     { "bicublin", "luma bicubic, chroma bilinear", 0, FF_OPT_TYPE_CONST, SWS_BICUBLIN, INT_MIN, INT_MAX, VE, "sws_flags" },

  258.     { "gauss", "gaussian", 0, FF_OPT_TYPE_CONST, SWS_GAUSS, INT_MIN, INT_MAX, VE, "sws_flags" },

  259.     { "sinc", "sinc", 0, FF_OPT_TYPE_CONST, SWS_SINC, INT_MIN, INT_MAX, VE, "sws_flags" },

  260.     { "lanczos", "lanczos", 0, FF_OPT_TYPE_CONST, SWS_LANCZOS, INT_MIN, INT_MAX, VE, "sws_flags" },

  261.     { "spline", "natural bicubic spline", 0, FF_OPT_TYPE_CONST, SWS_SPLINE, INT_MIN, INT_MAX, VE, "sws_flags" },

  262.     { "print_info", "print info", 0, FF_OPT_TYPE_CONST, SWS_PRINT_INFO, INT_MIN, INT_MAX, VE, "sws_flags" },

  263.     { "accurate_rnd", "accurate rounding", 0, FF_OPT_TYPE_CONST, SWS_ACCURATE_RND, INT_MIN, INT_MAX, VE, "sws_flags" },

  264.     { "mmx", "MMX SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_MMX, INT_MIN, INT_MAX, VE, "sws_flags" },

  265.     { "mmx2", "MMX2 SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_MMX2, INT_MIN, INT_MAX, VE, "sws_flags" },

  266.     { "3dnow", "3DNOW SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_3DNOW, INT_MIN, INT_MAX, VE, "sws_flags" },

  267.     { "altivec", "AltiVec SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_ALTIVEC, INT_MIN, INT_MAX, VE, "sws_flags" },

  268.     { "bfin", "Blackfin SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_BFIN, INT_MIN, INT_MAX, VE, "sws_flags" },

  269.     { "full_chroma_int", "full chroma interpolation", 0 , FF_OPT_TYPE_CONST, SWS_FULL_CHR_H_INT, INT_MIN, INT_MAX, VE, "sws_flags" },

  270.     { "full_chroma_inp", "full chroma input", 0 , FF_OPT_TYPE_CONST, SWS_FULL_CHR_H_INP, INT_MIN, INT_MAX, VE, "sws_flags" },

  271.     { NULL }

  272. };

  273. 

  274. #undef VE

  275. #undef DEFAULT

  276. 

  277. static const AVClass sws_context_class = { "SWScaler", sws_context_to_name, options };

  278. 

  279. const char *sws_format_name(enum PixelFormat format)

  280. {

  281.     switch (format) {

  282.         case PIX_FMT_YUV420P:

  283.             return "yuv420p";

  284.         case PIX_FMT_YUVA420P:

  285.             return "yuva420p";

  286.         case PIX_FMT_YUYV422:

  287.             return "yuyv422";

  288.         case PIX_FMT_RGB24:

  289.             return "rgb24";

  290.         case PIX_FMT_BGR24:

  291.             return "bgr24";

  292.         case PIX_FMT_YUV422P:

  293.             return "yuv422p";

  294.         case PIX_FMT_YUV444P:

  295.             return "yuv444p";

  296.         case PIX_FMT_RGB32:

  297.             return "rgb32";

  298.         case PIX_FMT_YUV410P:

  299.             return "yuv410p";

  300.         case PIX_FMT_YUV411P:

  301.             return "yuv411p";

  302.         case PIX_FMT_RGB565:

  303.             return "rgb565";

  304.         case PIX_FMT_RGB555:

  305.             return "rgb555";

  306.         case PIX_FMT_GRAY16BE:

  307.             return "gray16be";

  308.         case PIX_FMT_GRAY16LE:

  309.             return "gray16le";

  310.         case PIX_FMT_GRAY8:

  311.             return "gray8";

  312.         case PIX_FMT_MONOWHITE:

  313.             return "mono white";

  314.         case PIX_FMT_MONOBLACK:

  315.             return "mono black";

  316.         case PIX_FMT_PAL8:

  317.             return "Palette";

  318.         case PIX_FMT_YUVJ420P:

  319.             return "yuvj420p";

  320.         case PIX_FMT_YUVJ422P:

  321.             return "yuvj422p";

  322.         case PIX_FMT_YUVJ444P:

  323.             return "yuvj444p";

  324.         case PIX_FMT_XVMC_MPEG2_MC:

  325.             return "xvmc_mpeg2_mc";

  326.         case PIX_FMT_XVMC_MPEG2_IDCT:

  327.             return "xvmc_mpeg2_idct";

  328.         case PIX_FMT_UYVY422:

  329.             return "uyvy422";

  330.         case PIX_FMT_UYYVYY411:

  331.             return "uyyvyy411";

  332.         case PIX_FMT_RGB32_1:

  333.             return "rgb32x";

  334.         case PIX_FMT_BGR32_1:

  335.             return "bgr32x";

  336.         case PIX_FMT_BGR32:

  337.             return "bgr32";

  338.         case PIX_FMT_BGR565:

  339.             return "bgr565";

  340.         case PIX_FMT_BGR555:

  341.             return "bgr555";

  342.         case PIX_FMT_BGR8:

  343.             return "bgr8";

  344.         case PIX_FMT_BGR4:

  345.             return "bgr4";

  346.         case PIX_FMT_BGR4_BYTE:

  347.             return "bgr4 byte";

  348.         case PIX_FMT_RGB8:

  349.             return "rgb8";

  350.         case PIX_FMT_RGB4:

  351.             return "rgb4";

  352.         case PIX_FMT_RGB4_BYTE:

  353.             return "rgb4 byte";

  354.         case PIX_FMT_NV12:

  355.             return "nv12";

  356.         case PIX_FMT_NV21:

  357.             return "nv21";

  358.         case PIX_FMT_YUV440P:

  359.             return "yuv440p";

  360.         default:

  361.             return "Unknown format";

  362.     }

  363. }

  364. 

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

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

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

  368. {

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

  370.     int i;

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

  372.     {

  373.         int val=1<<18;

  374.         int j;

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

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

  377. 

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

  379.     }

  380. 

  381.     if (uDest)

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

  383.         {

  384.             int u=1<<18;

  385.             int v=1<<18;

  386.             int j;

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

  388.             {

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

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

  391.             }

  392. 

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

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

  395.         }

  396. }

  397. 

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

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

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

  401. {

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

  403.     int i;

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

  405.     {

  406.         int val=1<<18;

  407.         int j;

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

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

  410. 

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

  412.     }

  413. 

  414.     if (!uDest)

  415.         return;

  416. 

  417.     if (dstFormat == PIX_FMT_NV12)

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

  419.         {

  420.             int u=1<<18;

  421.             int v=1<<18;

  422.             int j;

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

  424.             {

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

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

  427.             }

  428. 

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

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

  431.         }

  432.     else

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

  434.         {

  435.             int u=1<<18;

  436.             int v=1<<18;

  437.             int j;

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

  439.             {

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

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

  442.             }

  443. 

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

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

  446.         }

  447. }

  448. 

  449. #define YSCALE_YUV_2_PACKEDX_C(type) \

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

  451.         int j;\

  452.         int Y1 = 1<<18;\

  453.         int Y2 = 1<<18;\

  454.         int U  = 1<<18;\

  455.         int V  = 1<<18;\

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

  457.         const int i2= 2*i;\

  458.         \

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

  460.         {\

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

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

  463.         }\

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

  465.         {\

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

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

  468.         }\

  469.         Y1>>=19;\

  470.         Y2>>=19;\

  471.         U >>=19;\

  472.         V >>=19;\

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

  474.         {\

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

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

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

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

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

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

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

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

  483.         }

  484. 

  485. #define YSCALE_YUV_2_RGBX_C(type) \

  486.     YSCALE_YUV_2_PACKEDX_C(type)  \

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

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

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

  490. 

  491. #define YSCALE_YUV_2_PACKED2_C   \

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

  493.         const int i2= 2*i;       \

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

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

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

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

  498. 

  499. #define YSCALE_YUV_2_RGB2_C(type) \

  500.     YSCALE_YUV_2_PACKED2_C\

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

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

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

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

  505. 

  506. #define YSCALE_YUV_2_PACKED1_C \

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

  508.         const int i2= 2*i;\

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

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

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

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

  513. 

  514. #define YSCALE_YUV_2_RGB1_C(type) \

  515.     YSCALE_YUV_2_PACKED1_C\

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

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

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

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

  520. 

  521. #define YSCALE_YUV_2_PACKED1B_C \

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

  523.         const int i2= 2*i;\

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

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

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

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

  528. 

  529. #define YSCALE_YUV_2_RGB1B_C(type) \

  530.     YSCALE_YUV_2_PACKED1B_C\

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

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

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

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

  535. 

  536. #define YSCALE_YUV_2_ANYRGB_C(func, func2)\

  537.     switch(c->dstFormat)\

  538.     {\

  539.     case PIX_FMT_RGB32:\

  540.     case PIX_FMT_BGR32:\

  541.         func(uint32_t)\

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

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

  544.         }                \

  545.         break;\

  546.     case PIX_FMT_RGB24:\

  547.         func(uint8_t)\

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

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

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

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

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

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

  554.             dest+=6;\

  555.         }\

  556.         break;\

  557.     case PIX_FMT_BGR24:\

  558.         func(uint8_t)\

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

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

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

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

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

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

  565.             dest+=6;\

  566.         }\

  567.         break;\

  568.     case PIX_FMT_RGB565:\

  569.     case PIX_FMT_BGR565:\

  570.         {\

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

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

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

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

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

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

  577.             func(uint16_t)\

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

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

  580.             }\

  581.         }\

  582.         break;\

  583.     case PIX_FMT_RGB555:\

  584.     case PIX_FMT_BGR555:\

  585.         {\

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

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

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

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

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

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

  592.             func(uint16_t)\

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

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

  595.             }\

  596.         }\

  597.         break;\

  598.     case PIX_FMT_RGB8:\

  599.     case PIX_FMT_BGR8:\

  600.         {\

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

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

  603.             func(uint8_t)\

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

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

  606.             }\

  607.         }\

  608.         break;\

  609.     case PIX_FMT_RGB4:\

  610.     case PIX_FMT_BGR4:\

  611.         {\

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

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

  614.             func(uint8_t)\

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

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

  617.             }\

  618.         }\

  619.         break;\

  620.     case PIX_FMT_RGB4_BYTE:\

  621.     case PIX_FMT_BGR4_BYTE:\

  622.         {\

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

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

  625.             func(uint8_t)\

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

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

  628.             }\

  629.         }\

  630.         break;\

  631.     case PIX_FMT_MONOBLACK:\

  632.         {\

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

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

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

  636.                 int acc;\

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

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

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

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

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

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

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

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

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

  646.                 dest++;\

  647.             }\

  648. \

  649. /*\

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

  651. {\

  652.             int acc=0;\

  653.             int left=0;\

  654.             static int top[1024];\

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

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

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

  658.             int topLeft=0;\

  659.             int shift=0;\

  660.             int count=0;\

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

  662.             int error_new=0;\

  663.             int error_in3=0;\

  664.             int f=0;\

  665.             \

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

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

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

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

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

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

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

  673. \

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

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

  676.                 f= error_new - error_in3*4;\

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

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

  679. \

  680.                 topLeft= top[i];\

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

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

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

  684. \

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

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

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

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

  689.                 }\

  690.             }\

  691. }\

  692. */\

  693.         }\

  694.         break;\

  695.     case PIX_FMT_YUYV422:\

  696.         func2\

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

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

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

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

  701.         }                \

  702.         break;\

  703.     case PIX_FMT_UYVY422:\

  704.         func2\

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

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

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

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

  709.         }                \

  710.         break;\

  711.     }\

  712. 

  713. 

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

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

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

  717. {

  718.     int i;

  719.     switch(c->dstFormat)

  720.     {

  721.     case PIX_FMT_BGR32:

  722.     case PIX_FMT_RGB32:

  723.         YSCALE_YUV_2_RGBX_C(uint32_t)

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

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

  726.         }

  727.         break;

  728.     case PIX_FMT_RGB24:

  729.         YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

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

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

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

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

  736.             dest+=6;

  737.         }

  738.         break;

  739.     case PIX_FMT_BGR24:

  740.         YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

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

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

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

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

  747.             dest+=6;

  748.         }

  749.         break;

  750.     case PIX_FMT_RGB565:

  751.     case PIX_FMT_BGR565:

  752.         {

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

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

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

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

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

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

  759.             YSCALE_YUV_2_RGBX_C(uint16_t)

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

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

  762.             }

  763.         }

  764.         break;

  765.     case PIX_FMT_RGB555:

  766.     case PIX_FMT_BGR555:

  767.         {

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

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

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

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

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

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

  774.             YSCALE_YUV_2_RGBX_C(uint16_t)

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

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

  777.             }

  778.         }

  779.         break;

  780.     case PIX_FMT_RGB8:

  781.     case PIX_FMT_BGR8:

  782.         {

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

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

  785.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  788.             }

  789.         }

  790.         break;

  791.     case PIX_FMT_RGB4:

  792.     case PIX_FMT_BGR4:

  793.         {

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

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

  796.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  799.             }

  800.         }

  801.         break;

  802.     case PIX_FMT_RGB4_BYTE:

  803.     case PIX_FMT_BGR4_BYTE:

  804.         {

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

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

  807.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  810.             }

  811.         }

  812.         break;

  813.     case PIX_FMT_MONOBLACK:

  814.         {

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

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

  817.             int acc=0;

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

  819.                 int j;

  820.                 int Y1=1<<18;

  821.                 int Y2=1<<18;

  822. 

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

  824.                 {

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

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

  827.                 }

  828.                 Y1>>=19;

  829.                 Y2>>=19;

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

  831.                 {

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

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

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

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

  836.                 }

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

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

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

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

  841.                     dest++;

  842.                 }

  843.             }

  844.         }

  845.         break;

  846.     case PIX_FMT_YUYV422:

  847.         YSCALE_YUV_2_PACKEDX_C(void)

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

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

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

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

  852.         }

  853.         break;

  854.     case PIX_FMT_UYVY422:

  855.         YSCALE_YUV_2_PACKEDX_C(void)

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

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

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

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

  860.         }

  861.         break;

  862.     }

  863. }

  864. 

  865. 

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

  867. //Plain C versions

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

  869. #define COMPILE_C

  870. #endif

  871. 

  872. #ifdef ARCH_POWERPC

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

  874. #define COMPILE_ALTIVEC

  875. #endif //HAVE_ALTIVEC

  876. #endif //ARCH_POWERPC

  877. 

  878. #if defined(ARCH_X86)

  879. 

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

  881. #define COMPILE_MMX

  882. #endif

  883. 

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

  885. #define COMPILE_MMX2

  886. #endif

  887. 

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

  889. #define COMPILE_3DNOW

  890. #endif

  891. #endif //ARCH_X86 || ARCH_X86_64

  892. 

  893. #undef HAVE_MMX

  894. #undef HAVE_MMX2

  895. #undef HAVE_3DNOW

  896. 

  897. #ifdef COMPILE_C

  898. #undef HAVE_MMX

  899. #undef HAVE_MMX2

  900. #undef HAVE_3DNOW

  901. #undef HAVE_ALTIVEC

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

  903. #include "swscale_template.c"

  904. #endif

  905. 

  906. #ifdef COMPILE_ALTIVEC

  907. #undef RENAME

  908. #define HAVE_ALTIVEC

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

  910. #include "swscale_template.c"

  911. #endif

  912. 

  913. #if defined(ARCH_X86)

  914. 

  915. //X86 versions

  916. /*

  917. #undef RENAME

  918. #undef HAVE_MMX

  919. #undef HAVE_MMX2

  920. #undef HAVE_3DNOW

  921. #define ARCH_X86

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

  923. #include "swscale_template.c"

  924. */

  925. //MMX versions

  926. #ifdef COMPILE_MMX

  927. #undef RENAME

  928. #define HAVE_MMX

  929. #undef HAVE_MMX2

  930. #undef HAVE_3DNOW

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

  932. #include "swscale_template.c"

  933. #endif

  934. 

  935. //MMX2 versions

  936. #ifdef COMPILE_MMX2

  937. #undef RENAME

  938. #define HAVE_MMX

  939. #define HAVE_MMX2

  940. #undef HAVE_3DNOW

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

  942. #include "swscale_template.c"

  943. #endif

  944. 

  945. //3DNOW versions

  946. #ifdef COMPILE_3DNOW

  947. #undef RENAME

  948. #define HAVE_MMX

  949. #undef HAVE_MMX2

  950. #define HAVE_3DNOW

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

  952. #include "swscale_template.c"

  953. #endif

  954. 

  955. #endif //ARCH_X86 || ARCH_X86_64

  956. 

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

  958. 

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

  960. {

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

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

  963.     else                return getSplineCoeff(        0.0,

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

  965.                                                     c + 3.0*d,

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

  967.                                             dist-1.0);

  968. }

  969. 

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

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

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

  973. {

  974.     int i;

  975.     int filterSize;

  976.     int filter2Size;

  977.     int minFilterSize;

  978.     double *filter=NULL;

  979.     double *filter2=NULL;

  980. #if defined(ARCH_X86)

  981.     if (flags & SWS_CPU_CAPS_MMX)

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

  983. #endif

  984. 

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

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

  987. 

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

  989.     {

  990.         int i;

  991.         filterSize= 1;

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

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

  994. 

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

  996.         {

  997.             filter[i*filterSize]=1;

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

  999.         }

  1000. 

  1001.     }

  1002.     else if (flags&SWS_POINT) // lame looking point sampling mode

  1003.     {

  1004.         int i;

  1005.         int xDstInSrc;

  1006.         filterSize= 1;

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

  1008. 

  1009.         xDstInSrc= xInc/2 - 0x8000;

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

  1011.         {

  1012.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1013. 

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

  1015.             filter[i]= 1.0;

  1016.             xDstInSrc+= xInc;

  1017.         }

  1018.     }

  1019.     else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale

  1020.     {

  1021.         int i;

  1022.         int xDstInSrc;

  1023.         if      (flags&SWS_BICUBIC) filterSize= 4;

  1024.         else if (flags&SWS_X      ) filterSize= 4;

  1025.         else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA

  1026.         filter= av_malloc(dstW*sizeof(double)*filterSize);

  1027. 

  1028.         xDstInSrc= xInc/2 - 0x8000;

  1029.         for (i=0; i<dstW; i++)

  1030.         {

  1031.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1032.             int j;

  1033. 

  1034.             (*filterPos)[i]= xx;

  1035.                 //Bilinear upscale / linear interpolate / Area averaging

  1036.                 for (j=0; j<filterSize; j++)

  1037.                 {

  1038.                     double d= FFABS((xx<<16) - xDstInSrc)/(double)(1<<16);

  1039.                     double coeff= 1.0 - d;

  1040.                     if (coeff<0) coeff=0;

  1041.                     filter[i*filterSize + j]= coeff;

  1042.                     xx++;

  1043.                 }

  1044.             xDstInSrc+= xInc;

  1045.         }

  1046.     }

  1047.     else

  1048.     {

  1049.         double xDstInSrc;

  1050.         double sizeFactor, filterSizeInSrc;

  1051.         const double xInc1= (double)xInc / (double)(1<<16);

  1052. 

  1053.         if      (flags&SWS_BICUBIC)      sizeFactor=  4.0;

  1054.         else if (flags&SWS_X)            sizeFactor=  8.0;

  1055.         else if (flags&SWS_AREA)         sizeFactor=  1.0; //downscale only, for upscale it is bilinear

  1056.         else if (flags&SWS_GAUSS)        sizeFactor=  8.0;   // infinite ;)

  1057.         else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? 2.0*param[0] : 6.0;

  1058.         else if (flags&SWS_SINC)         sizeFactor= 20.0; // infinite ;)

  1059.         else if (flags&SWS_SPLINE)       sizeFactor= 20.0;  // infinite ;)

  1060.         else if (flags&SWS_BILINEAR)     sizeFactor=  2.0;

  1061.         else {

  1062.             sizeFactor= 0.0; //GCC warning killer

  1063.             ASSERT(0)

  1064.         }

  1065. 

  1066.         if (xInc1 <= 1.0)       filterSizeInSrc= sizeFactor; // upscale

  1067.         else                    filterSizeInSrc= sizeFactor*srcW / (double)dstW;

  1068. 

  1069.         filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible

  1070.         if (filterSize > srcW-2) filterSize=srcW-2;

  1071. 

  1072.         filter= av_malloc(dstW*sizeof(double)*filterSize);

  1073. 

  1074.         xDstInSrc= xInc1 / 2.0 - 0.5;

  1075.         for (i=0; i<dstW; i++)

  1076.         {

  1077.             int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);

  1078.             int j;

  1079.             (*filterPos)[i]= xx;

  1080.             for (j=0; j<filterSize; j++)

  1081.             {

  1082.                 double d= FFABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;

  1083.                 double coeff;

  1084.                 if (flags & SWS_BICUBIC)

  1085.                 {

  1086.                     double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;

  1087.                     double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;

  1088. 

  1089.                     if (d<1.0)

  1090.                         coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;

  1091.                     else if (d<2.0)

  1092.                         coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;

  1093.                     else

  1094.                         coeff=0.0;

  1095.                 }

  1096. /*                else if (flags & SWS_X)

  1097.                 {

  1098.                     double p= param ? param*0.01 : 0.3;

  1099.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1100.                     coeff*= pow(2.0, - p*d*d);

  1101.                 }*/

  1102.                 else if (flags & SWS_X)

  1103.                 {

  1104.                     double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

  1105. 

  1106.                     if (d<1.0)

  1107.                         coeff = cos(d*PI);

  1108.                     else

  1109.                         coeff=-1.0;

  1110.                     if (coeff<0.0)      coeff= -pow(-coeff, A);

  1111.                     else                coeff=  pow( coeff, A);

  1112.                     coeff= coeff*0.5 + 0.5;

  1113.                 }

  1114.                 else if (flags & SWS_AREA)

  1115.                 {

  1116.                     double srcPixelSize= 1.0/xInc1;

  1117.                     if      (d + srcPixelSize/2 < 0.5) coeff= 1.0;

  1118.                     else if (d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5;

  1119.                     else coeff=0.0;

  1120.                 }

  1121.                 else if (flags & SWS_GAUSS)

  1122.                 {

  1123.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1124.                     coeff = pow(2.0, - p*d*d);

  1125.                 }

  1126.                 else if (flags & SWS_SINC)

  1127.                 {

  1128.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1129.                 }

  1130.                 else if (flags & SWS_LANCZOS)

  1131.                 {

  1132.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1133.                     coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;

  1134.                     if (d>p) coeff=0;

  1135.                 }

  1136.                 else if (flags & SWS_BILINEAR)

  1137.                 {

  1138.                     coeff= 1.0 - d;

  1139.                     if (coeff<0) coeff=0;

  1140.                 }

  1141.                 else if (flags & SWS_SPLINE)

  1142.                 {

  1143.                     double p=-2.196152422706632;

  1144.                     coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);

  1145.                 }

  1146.                 else {

  1147.                     coeff= 0.0; //GCC warning killer

  1148.                     ASSERT(0)

  1149.                 }

  1150. 

  1151.                 filter[i*filterSize + j]= coeff;

  1152.                 xx++;

  1153.             }

  1154.             xDstInSrc+= xInc1;

  1155.         }

  1156.     }

  1157. 

  1158.     /* apply src & dst Filter to filter -> filter2

  1159.        av_free(filter);

  1160.     */

  1161.     ASSERT(filterSize>0)

  1162.     filter2Size= filterSize;

  1163.     if (srcFilter) filter2Size+= srcFilter->length - 1;

  1164.     if (dstFilter) filter2Size+= dstFilter->length - 1;

  1165.     ASSERT(filter2Size>0)

  1166.     filter2= av_malloc(filter2Size*dstW*sizeof(double));

  1167. 

  1168.     for (i=0; i<dstW; i++)

  1169.     {

  1170.         int j;

  1171.         SwsVector scaleFilter;

  1172.         SwsVector *outVec;

  1173. 

  1174.         scaleFilter.coeff= filter + i*filterSize;

  1175.         scaleFilter.length= filterSize;

  1176. 

  1177.         if (srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);

  1178.         else           outVec= &scaleFilter;

  1179. 

  1180.         ASSERT(outVec->length == filter2Size)

  1181.         //FIXME dstFilter

  1182. 

  1183.         for (j=0; j<outVec->length; j++)

  1184.         {

  1185.             filter2[i*filter2Size + j]= outVec->coeff[j];

  1186.         }

  1187. 

  1188.         (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

  1189. 

  1190.         if (outVec != &scaleFilter) sws_freeVec(outVec);

  1191.     }

  1192.     av_free(filter); filter=NULL;

  1193. 

  1194.     /* try to reduce the filter-size (step1 find size and shift left) */

  1195.     // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).

  1196.     minFilterSize= 0;

  1197.     for (i=dstW-1; i>=0; i--)

  1198.     {

  1199.         int min= filter2Size;

  1200.         int j;

  1201.         double cutOff=0.0;

  1202. 

  1203.         /* get rid off near zero elements on the left by shifting left */

  1204.         for (j=0; j<filter2Size; j++)

  1205.         {

  1206.             int k;

  1207.             cutOff += FFABS(filter2[i*filter2Size]);

  1208. 

  1209.             if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;

  1210. 

  1211.             /* preserve monotonicity because the core can't handle the filter otherwise */

  1212.             if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

  1213. 

  1214.             // Move filter coeffs left

  1215.             for (k=1; k<filter2Size; k++)

  1216.                 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];

  1217.             filter2[i*filter2Size + k - 1]= 0.0;

  1218.             (*filterPos)[i]++;

  1219.         }

  1220. 

  1221.         cutOff=0.0;

  1222.         /* count near zeros on the right */

  1223.         for (j=filter2Size-1; j>0; j--)

  1224.         {

  1225.             cutOff += FFABS(filter2[i*filter2Size + j]);

  1226. 

  1227.             if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;

  1228.             min--;

  1229.         }

  1230. 

  1231.         if (min>minFilterSize) minFilterSize= min;

  1232.     }

  1233. 

  1234.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1235.         // we can handle the special case 4,

  1236.         // so we don't want to go to the full 8

  1237.         if (minFilterSize < 5)

  1238.             filterAlign = 4;

  1239. 

  1240.         // we really don't want to waste our time

  1241.         // doing useless computation, so fall-back on

  1242.         // the scalar C code for very small filter.

  1243.         // vectorizing is worth it only if you have

  1244.         // decent-sized vector.

  1245.         if (minFilterSize < 3)

  1246.             filterAlign = 1;

  1247.     }

  1248. 

  1249.     if (flags & SWS_CPU_CAPS_MMX) {

  1250.         // special case for unscaled vertical filtering

  1251.         if (minFilterSize == 1 && filterAlign == 2)

  1252.             filterAlign= 1;

  1253.     }

  1254. 

  1255.     ASSERT(minFilterSize > 0)

  1256.     filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));

  1257.     ASSERT(filterSize > 0)

  1258.     filter= av_malloc(filterSize*dstW*sizeof(double));

  1259.     if (filterSize >= MAX_FILTER_SIZE)

  1260.         return -1;

  1261.     *outFilterSize= filterSize;

  1262. 

  1263.     if (flags&SWS_PRINT_INFO)

  1264.         av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);

  1265.     /* try to reduce the filter-size (step2 reduce it) */

  1266.     for (i=0; i<dstW; i++)

  1267.     {

  1268.         int j;

  1269. 

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

  1271.         {

  1272.             if (j>=filter2Size) filter[i*filterSize + j]= 0.0;

  1273.             else               filter[i*filterSize + j]= filter2[i*filter2Size + j];

  1274.         }

  1275.     }

  1276.     av_free(filter2); filter2=NULL;

  1277. 

  1278. 

  1279.     //FIXME try to align filterpos if possible

  1280. 

  1281.     //fix borders

  1282.     for (i=0; i<dstW; i++)

  1283.     {

  1284.         int j;

  1285.         if ((*filterPos)[i] < 0)

  1286.         {

  1287.             // Move filter coeffs left to compensate for filterPos

  1288.             for (j=1; j<filterSize; j++)

  1289.             {

  1290.                 int left= FFMAX(j + (*filterPos)[i], 0);

  1291.                 filter[i*filterSize + left] += filter[i*filterSize + j];

  1292.                 filter[i*filterSize + j]=0;

  1293.             }

  1294.             (*filterPos)[i]= 0;

  1295.         }

  1296. 

  1297.         if ((*filterPos)[i] + filterSize > srcW)

  1298.         {

  1299.             int shift= (*filterPos)[i] + filterSize - srcW;

  1300.             // Move filter coeffs right to compensate for filterPos

  1301.             for (j=filterSize-2; j>=0; j--)

  1302.             {

  1303.                 int right= FFMIN(j + shift, filterSize-1);

  1304.                 filter[i*filterSize +right] += filter[i*filterSize +j];

  1305.                 filter[i*filterSize +j]=0;

  1306.             }

  1307.             (*filterPos)[i]= srcW - filterSize;

  1308.         }

  1309.     }

  1310. 

  1311.     // Note the +1 is for the MMXscaler which reads over the end

  1312.     /* align at 16 for AltiVec (needed by hScale_altivec_real) */

  1313.     *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));

  1314. 

  1315.     /* Normalize & Store in outFilter */

  1316.     for (i=0; i<dstW; i++)

  1317.     {

  1318.         int j;

  1319.         double error=0;

  1320.         double sum=0;

  1321.         double scale= one;

  1322. 

  1323.         for (j=0; j<filterSize; j++)

  1324.         {

  1325.             sum+= filter[i*filterSize + j];

  1326.         }

  1327.         scale/= sum;

  1328.         for (j=0; j<*outFilterSize; j++)

  1329.         {

  1330.             double v= filter[i*filterSize + j]*scale + error;

  1331.             int intV= floor(v + 0.5);

  1332.             (*outFilter)[i*(*outFilterSize) + j]= intV;

  1333.             error = v - intV;

  1334.         }

  1335.     }

  1336. 

  1337.     (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end

  1338.     for (i=0; i<*outFilterSize; i++)

  1339.     {

  1340.         int j= dstW*(*outFilterSize);

  1341.         (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

  1342.     }

  1343. 

  1344.     av_free(filter);

  1345.     return 0;

  1346. }

  1347. 

  1348. #ifdef COMPILE_MMX2

  1349. static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)

  1350. {

  1351.     uint8_t *fragmentA;

  1352.     long imm8OfPShufW1A;

  1353.     long imm8OfPShufW2A;

  1354.     long fragmentLengthA;

  1355.     uint8_t *fragmentB;

  1356.     long imm8OfPShufW1B;

  1357.     long imm8OfPShufW2B;

  1358.     long fragmentLengthB;

  1359.     int fragmentPos;

  1360. 

  1361.     int xpos, i;

  1362. 

  1363.     // create an optimized horizontal scaling routine

  1364. 

  1365.     //code fragment

  1366. 

  1367.     asm volatile(

  1368.         "jmp                         9f                 \n\t"

  1369.     // Begin

  1370.         "0:                                             \n\t"

  1371.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1372.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1373.         "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"

  1374.         "punpcklbw                %%mm7, %%mm1          \n\t"

  1375.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1376.         "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"

  1377.         "1:                                             \n\t"

  1378.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1379.         "2:                                             \n\t"

  1380.         "psubw                    %%mm1, %%mm0          \n\t"

  1381.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1382.         "pmullw                   %%mm3, %%mm0          \n\t"

  1383.         "psllw                       $7, %%mm1          \n\t"

  1384.         "paddw                    %%mm1, %%mm0          \n\t"

  1385. 

  1386.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1387. 

  1388.         "add                         $8, %%"REG_a"      \n\t"

  1389.     // End

  1390.         "9:                                             \n\t"

  1391. //        "int $3                                         \n\t"

  1392.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1393.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1394.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1395.         "dec                         %1                 \n\t"

  1396.         "dec                         %2                 \n\t"

  1397.         "sub                         %0, %1             \n\t"

  1398.         "sub                         %0, %2             \n\t"

  1399.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1400.         "sub                         %0, %3             \n\t"

  1401. 

  1402. 

  1403.         :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),

  1404.         "=r" (fragmentLengthA)

  1405.     );

  1406. 

  1407.     asm volatile(

  1408.         "jmp                         9f                 \n\t"

  1409.     // Begin

  1410.         "0:                                             \n\t"

  1411.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1412.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1413.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1414.         "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"

  1415.         "1:                                             \n\t"

  1416.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1417.         "2:                                             \n\t"

  1418.         "psubw                    %%mm1, %%mm0          \n\t"

  1419.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1420.         "pmullw                   %%mm3, %%mm0          \n\t"

  1421.         "psllw                       $7, %%mm1          \n\t"

  1422.         "paddw                    %%mm1, %%mm0          \n\t"

  1423. 

  1424.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1425. 

  1426.         "add                         $8, %%"REG_a"      \n\t"

  1427.     // End

  1428.         "9:                                             \n\t"

  1429. //        "int                       $3                   \n\t"

  1430.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1431.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1432.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1433.         "dec                         %1                 \n\t"

  1434.         "dec                         %2                 \n\t"

  1435.         "sub                         %0, %1             \n\t"

  1436.         "sub                         %0, %2             \n\t"

  1437.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1438.         "sub                         %0, %3             \n\t"

  1439. 

  1440. 

  1441.         :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),

  1442.         "=r" (fragmentLengthB)

  1443.     );

  1444. 

  1445.     xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers

  1446.     fragmentPos=0;

  1447. 

  1448.     for (i=0; i<dstW/numSplits; i++)

  1449.     {

  1450.         int xx=xpos>>16;

  1451. 

  1452.         if ((i&3) == 0)

  1453.         {

  1454.             int a=0;

  1455.             int b=((xpos+xInc)>>16) - xx;

  1456.             int c=((xpos+xInc*2)>>16) - xx;

  1457.             int d=((xpos+xInc*3)>>16) - xx;

  1458. 

  1459.             filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;

  1460.             filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;

  1461.             filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;

  1462.             filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;

  1463.             filterPos[i/2]= xx;

  1464. 

  1465.             if (d+1<4)

  1466.             {

  1467.                 int maxShift= 3-(d+1);

  1468.                 int shift=0;

  1469. 

  1470.                 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);

  1471. 

  1472.                 funnyCode[fragmentPos + imm8OfPShufW1B]=

  1473.                     (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);

  1474.                 funnyCode[fragmentPos + imm8OfPShufW2B]=

  1475.                     a | (b<<2) | (c<<4) | (d<<6);

  1476. 

  1477.                 if (i+3>=dstW) shift=maxShift; //avoid overread

  1478.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

  1479. 

  1480.                 if (shift && i>=shift)

  1481.                 {

  1482.                     funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;

  1483.                     funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;

  1484.                     filterPos[i/2]-=shift;

  1485.                 }

  1486. 

  1487.                 fragmentPos+= fragmentLengthB;

  1488.             }

  1489.             else

  1490.             {

  1491.                 int maxShift= 3-d;

  1492.                 int shift=0;

  1493. 

  1494.                 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);

  1495. 

  1496.                 funnyCode[fragmentPos + imm8OfPShufW1A]=

  1497.                 funnyCode[fragmentPos + imm8OfPShufW2A]=

  1498.                     a | (b<<2) | (c<<4) | (d<<6);

  1499. 

  1500.                 if (i+4>=dstW) shift=maxShift; //avoid overread

  1501.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align

  1502. 

  1503.                 if (shift && i>=shift)

  1504.                 {

  1505.                     funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;

  1506.                     funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;

  1507.                     filterPos[i/2]-=shift;

  1508.                 }

  1509. 

  1510.                 fragmentPos+= fragmentLengthA;

  1511.             }

  1512. 

  1513.             funnyCode[fragmentPos]= RET;

  1514.         }

  1515.         xpos+=xInc;

  1516.     }

  1517.     filterPos[i/2]= xpos>>16; // needed to jump to the next part

  1518. }

  1519. #endif /* COMPILE_MMX2 */

  1520. 

  1521. static void globalInit(void){

  1522.     // generating tables:

  1523.     int i;

  1524.     for (i=0; i<768; i++){

  1525.         int c= av_clip_uint8(i-256);

  1526.         clip_table[i]=c;

  1527.     }

  1528. }

  1529. 

  1530. static SwsFunc getSwsFunc(int flags){

  1531. 

  1532. #if defined(RUNTIME_CPUDETECT) && defined (CONFIG_GPL)

  1533. #if defined(ARCH_X86)

  1534.     // ordered per speed fastest first

  1535.     if (flags & SWS_CPU_CAPS_MMX2)

  1536.         return swScale_MMX2;

  1537.     else if (flags & SWS_CPU_CAPS_3DNOW)

  1538.         return swScale_3DNow;

  1539.     else if (flags & SWS_CPU_CAPS_MMX)

  1540.         return swScale_MMX;

  1541.     else

  1542.         return swScale_C;

  1543. 

  1544. #else

  1545. #ifdef ARCH_POWERPC

  1546.     if (flags & SWS_CPU_CAPS_ALTIVEC)

  1547.         return swScale_altivec;

  1548.     else

  1549.         return swScale_C;

  1550. #endif

  1551.     return swScale_C;

  1552. #endif /* defined(ARCH_X86) */

  1553. #else //RUNTIME_CPUDETECT

  1554. #ifdef HAVE_MMX2

  1555.     return swScale_MMX2;

  1556. #elif defined (HAVE_3DNOW)

  1557.     return swScale_3DNow;

  1558. #elif defined (HAVE_MMX)

  1559.     return swScale_MMX;

  1560. #elif defined (HAVE_ALTIVEC)

  1561.     return swScale_altivec;

  1562. #else

  1563.     return swScale_C;

  1564. #endif

  1565. #endif //!RUNTIME_CPUDETECT

  1566. }

  1567. 

  1568. static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1569.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1570.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1571.     /* Copy Y plane */

  1572.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1573.         memcpy(dst, src[0], srcSliceH*dstStride[0]);

  1574.     else

  1575.     {

  1576.         int i;

  1577.         uint8_t *srcPtr= src[0];

  1578.         uint8_t *dstPtr= dst;

  1579.         for (i=0; i<srcSliceH; i++)

  1580.         {

  1581.             memcpy(dstPtr, srcPtr, c->srcW);

  1582.             srcPtr+= srcStride[0];

  1583.             dstPtr+= dstStride[0];

  1584.         }

  1585.     }

  1586.     dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1587.     if (c->dstFormat == PIX_FMT_NV12)

  1588.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  1589.     else

  1590.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  1591. 

  1592.     return srcSliceH;

  1593. }

  1594. 

  1595. static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1596.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1597.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1598. 

  1599.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1600. 

  1601.     return srcSliceH;

  1602. }

  1603. 

  1604. static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1605.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1606.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1607. 

  1608.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1609. 

  1610.     return srcSliceH;

  1611. }

  1612. 

  1613. /* {RGB,BGR}{15,16,24,32} -> {RGB,BGR}{15,16,24,32} */

  1614. static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1615.                           int srcSliceH, uint8_t* dst[], int dstStride[]){

  1616.     const int srcFormat= c->srcFormat;

  1617.     const int dstFormat= c->dstFormat;

  1618.     const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;

  1619.     const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;

  1620.     const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1621.     const int dstId= fmt_depth(dstFormat) >> 2;

  1622.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  1623. 

  1624.     /* BGR -> BGR */

  1625.     if (  (isBGR(srcFormat) && isBGR(dstFormat))

  1626.        || (isRGB(srcFormat) && isRGB(dstFormat))){

  1627.         switch(srcId | (dstId<<4)){

  1628.         case 0x34: conv= rgb16to15; break;

  1629.         case 0x36: conv= rgb24to15; break;

  1630.         case 0x38: conv= rgb32to15; break;

  1631.         case 0x43: conv= rgb15to16; break;

  1632.         case 0x46: conv= rgb24to16; break;

  1633.         case 0x48: conv= rgb32to16; break;

  1634.         case 0x63: conv= rgb15to24; break;

  1635.         case 0x64: conv= rgb16to24; break;

  1636.         case 0x68: conv= rgb32to24; break;

  1637.         case 0x83: conv= rgb15to32; break;

  1638.         case 0x84: conv= rgb16to32; break;

  1639.         case 0x86: conv= rgb24to32; break;

  1640.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1641.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1642.         }

  1643.     }else if (  (isBGR(srcFormat) && isRGB(dstFormat))

  1644.              || (isRGB(srcFormat) && isBGR(dstFormat))){

  1645.         switch(srcId | (dstId<<4)){

  1646.         case 0x33: conv= rgb15tobgr15; break;

  1647.         case 0x34: conv= rgb16tobgr15; break;

  1648.         case 0x36: conv= rgb24tobgr15; break;

  1649.         case 0x38: conv= rgb32tobgr15; break;

  1650.         case 0x43: conv= rgb15tobgr16; break;

  1651.         case 0x44: conv= rgb16tobgr16; break;

  1652.         case 0x46: conv= rgb24tobgr16; break;

  1653.         case 0x48: conv= rgb32tobgr16; break;

  1654.         case 0x63: conv= rgb15tobgr24; break;

  1655.         case 0x64: conv= rgb16tobgr24; break;

  1656.         case 0x66: conv= rgb24tobgr24; break;

  1657.         case 0x68: conv= rgb32tobgr24; break;

  1658.         case 0x83: conv= rgb15tobgr32; break;

  1659.         case 0x84: conv= rgb16tobgr32; break;

  1660.         case 0x86: conv= rgb24tobgr32; break;

  1661.         case 0x88: conv= rgb32tobgr32; break;

  1662.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1663.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1664.         }

  1665.     }else{

  1666.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1667.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1668.     }

  1669. 

  1670.     if(conv)

  1671.     {

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

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

  1674.         else

  1675.         {

  1676.             int i;

  1677.             uint8_t *srcPtr= src[0];

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

  1679. 

  1680.             for (i=0; i<srcSliceH; i++)

  1681.             {

  1682.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  1683.                 srcPtr+= srcStride[0];

  1684.                 dstPtr+= dstStride[0];

  1685.             }

  1686.         }

  1687.     }

  1688.     return srcSliceH;

  1689. }

  1690. 

  1691. static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1692.                               int srcSliceH, uint8_t* dst[], int dstStride[]){

  1693. 

  1694.     rgb24toyv12(

  1695.         src[0],

  1696.         dst[0]+ srcSliceY    *dstStride[0],

  1697.         dst[1]+(srcSliceY>>1)*dstStride[1],

  1698.         dst[2]+(srcSliceY>>1)*dstStride[2],

  1699.         c->srcW, srcSliceH,

  1700.         dstStride[0], dstStride[1], srcStride[0]);

  1701.     return srcSliceH;

  1702. }

  1703. 

  1704. static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1705.                              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1706.     int i;

  1707. 

  1708.     /* copy Y */

  1709.     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)

  1710.         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);

  1711.     else{

  1712.         uint8_t *srcPtr= src[0];

  1713.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1714. 

  1715.         for (i=0; i<srcSliceH; i++)

  1716.         {

  1717.             memcpy(dstPtr, srcPtr, c->srcW);

  1718.             srcPtr+= srcStride[0];

  1719.             dstPtr+= dstStride[0];

  1720.         }

  1721.     }

  1722. 

  1723.     if (c->dstFormat==PIX_FMT_YUV420P){

  1724.         planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);

  1725.         planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);

  1726.     }else{

  1727.         planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);

  1728.         planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);

  1729.     }

  1730.     return srcSliceH;

  1731. }

  1732. 

  1733. /* unscaled copy like stuff (assumes nearly identical formats) */

  1734. static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1735.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  1736. {

  1737.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1738.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  1739.     else

  1740.     {

  1741.         int i;

  1742.         uint8_t *srcPtr= src[0];

  1743.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1744.         int length=0;

  1745. 

  1746.         /* universal length finder */

  1747.         while(length+c->srcW <= FFABS(dstStride[0])

  1748.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  1749.         ASSERT(length!=0);

  1750. 

  1751.         for (i=0; i<srcSliceH; i++)

  1752.         {

  1753.             memcpy(dstPtr, srcPtr, length);

  1754.             srcPtr+= srcStride[0];

  1755.             dstPtr+= dstStride[0];

  1756.         }

  1757.     }

  1758.     return srcSliceH;

  1759. }

  1760. 

  1761. static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1762.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  1763. {

  1764.     int plane;

  1765.     for (plane=0; plane<3; plane++)

  1766.     {

  1767.         int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  1768.         int y=      plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  1769.         int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  1770. 

  1771.         if ((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)

  1772.         {

  1773.             if (!isGray(c->dstFormat))

  1774.                 memset(dst[plane], 128, dstStride[plane]*height);

  1775.         }

  1776.         else

  1777.         {

  1778.             if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)

  1779.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);

  1780.             else

  1781.             {

  1782.                 int i;

  1783.                 uint8_t *srcPtr= src[plane];

  1784.                 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  1785.                 for (i=0; i<height; i++)

  1786.                 {

  1787.                     memcpy(dstPtr, srcPtr, length);

  1788.                     srcPtr+= srcStride[plane];

  1789.                     dstPtr+= dstStride[plane];

  1790.                 }

  1791.             }

  1792.         }

  1793.     }

  1794.     return srcSliceH;

  1795. }

  1796. 

  1797. static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1798.                         int srcSliceH, uint8_t* dst[], int dstStride[]){

  1799. 

  1800.     int length= c->srcW;

  1801.     int y=      srcSliceY;

  1802.     int height= srcSliceH;

  1803.     int i, j;

  1804.     uint8_t *srcPtr= src[0];

  1805.     uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  1806. 

  1807.     if (!isGray(c->dstFormat)){

  1808.         int height= -((-srcSliceH)>>c->chrDstVSubSample);

  1809.         memset(dst[1], 128, dstStride[1]*height);

  1810.         memset(dst[2], 128, dstStride[2]*height);

  1811.     }

  1812.     if (c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;

  1813.     for (i=0; i<height; i++)

  1814.     {

  1815.         for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  1816.         srcPtr+= srcStride[0];

  1817.         dstPtr+= dstStride[0];

  1818.     }

  1819.     return srcSliceH;

  1820. }

  1821. 

  1822. static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1823.                         int srcSliceH, uint8_t* dst[], int dstStride[]){

  1824. 

  1825.     int length= c->srcW;

  1826.     int y=      srcSliceY;

  1827.     int height= srcSliceH;

  1828.     int i, j;

  1829.     uint8_t *srcPtr= src[0];

  1830.     uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  1831.     for (i=0; i<height; i++)

  1832.     {

  1833.         for (j=0; j<length; j++)

  1834.         {

  1835.             dstPtr[j<<1] = srcPtr[j];

  1836.             dstPtr[(j<<1)+1] = srcPtr[j];

  1837.         }

  1838.         srcPtr+= srcStride[0];

  1839.         dstPtr+= dstStride[0];

  1840.     }

  1841.     return srcSliceH;

  1842. }

  1843. 

  1844. static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1845.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  1846. 

  1847.     int length= c->srcW;

  1848.     int y=      srcSliceY;

  1849.     int height= srcSliceH;

  1850.     int i, j;

  1851.     uint16_t *srcPtr= (uint16_t*)src[0];

  1852.     uint16_t *dstPtr= (uint16_t*)(dst[0] + dstStride[0]*y/2);

  1853.     for (i=0; i<height; i++)

  1854.     {

  1855.         for (j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);

  1856.         srcPtr+= srcStride[0]/2;

  1857.         dstPtr+= dstStride[0]/2;

  1858.     }

  1859.     return srcSliceH;

  1860. }

  1861. 

  1862. 

  1863. static void getSubSampleFactors(int *h, int *v, int format){

  1864.     switch(format){

  1865.     case PIX_FMT_UYVY422:

  1866.     case PIX_FMT_YUYV422:

  1867.         *h=1;

  1868.         *v=0;

  1869.         break;

  1870.     case PIX_FMT_YUV420P:

  1871.     case PIX_FMT_YUVA420P:

  1872.     case PIX_FMT_GRAY16BE:

  1873.     case PIX_FMT_GRAY16LE:

  1874.     case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented

  1875.     case PIX_FMT_NV12:

  1876.     case PIX_FMT_NV21:

  1877.         *h=1;

  1878.         *v=1;

  1879.         break;

  1880.     case PIX_FMT_YUV440P:

  1881.         *h=0;

  1882.         *v=1;

  1883.         break;

  1884.     case PIX_FMT_YUV410P:

  1885.         *h=2;

  1886.         *v=2;

  1887.         break;

  1888.     case PIX_FMT_YUV444P:

  1889.         *h=0;

  1890.         *v=0;

  1891.         break;

  1892.     case PIX_FMT_YUV422P:

  1893.         *h=1;

  1894.         *v=0;

  1895.         break;

  1896.     case PIX_FMT_YUV411P:

  1897.         *h=2;

  1898.         *v=0;

  1899.         break;

  1900.     default:

  1901.         *h=0;

  1902.         *v=0;

  1903.         break;

  1904.     }

  1905. }

  1906. 

  1907. static uint16_t roundToInt16(int64_t f){

  1908.     int r= (f + (1<<15))>>16;

  1909.          if (r<-0x7FFF) return 0x8000;

  1910.     else if (r> 0x7FFF) return 0x7FFF;

  1911.     else                return r;

  1912. }

  1913. 

  1914. /**

  1915.  * @param inv_table the yuv2rgb coeffs, normally Inverse_Table_6_9[x]

  1916.  * @param fullRange if 1 then the luma range is 0..255 if 0 it is 16..235

  1917.  * @return -1 if not supported

  1918.  */

  1919. int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){

  1920.     int64_t crv =  inv_table[0];

  1921.     int64_t cbu =  inv_table[1];

  1922.     int64_t cgu = -inv_table[2];

  1923.     int64_t cgv = -inv_table[3];

  1924.     int64_t cy  = 1<<16;

  1925.     int64_t oy  = 0;

  1926. 

  1927.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  1928.     memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);

  1929.     memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);

  1930. 

  1931.     c->brightness= brightness;

  1932.     c->contrast  = contrast;

  1933.     c->saturation= saturation;

  1934.     c->srcRange  = srcRange;

  1935.     c->dstRange  = dstRange;

  1936. 

  1937.     c->uOffset=   0x0400040004000400LL;

  1938.     c->vOffset=   0x0400040004000400LL;

  1939. 

  1940.     if (!srcRange){

  1941.         cy= (cy*255) / 219;

  1942.         oy= 16<<16;

  1943.     }else{

  1944.         crv= (crv*224) / 255;

  1945.         cbu= (cbu*224) / 255;

  1946.         cgu= (cgu*224) / 255;

  1947.         cgv= (cgv*224) / 255;

  1948.     }

  1949. 

  1950.     cy = (cy *contrast             )>>16;

  1951.     crv= (crv*contrast * saturation)>>32;

  1952.     cbu= (cbu*contrast * saturation)>>32;

  1953.     cgu= (cgu*contrast * saturation)>>32;

  1954.     cgv= (cgv*contrast * saturation)>>32;

  1955. 

  1956.     oy -= 256*brightness;

  1957. 

  1958.     c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;

  1959.     c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;

  1960.     c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;

  1961.     c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;

  1962.     c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;

  1963.     c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;

  1964. 

  1965.     yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);

  1966.     //FIXME factorize

  1967. 

  1968. #ifdef COMPILE_ALTIVEC

  1969.     if (c->flags & SWS_CPU_CAPS_ALTIVEC)

  1970.         yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);

  1971. #endif

  1972.     return 0;

  1973. }

  1974. 

  1975. /**

  1976.  * @return -1 if not supported

  1977.  */

  1978. int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){

  1979.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  1980. 

  1981.     *inv_table = c->srcColorspaceTable;

  1982.     *table     = c->dstColorspaceTable;

  1983.     *srcRange  = c->srcRange;

  1984.     *dstRange  = c->dstRange;

  1985.     *brightness= c->brightness;

  1986.     *contrast  = c->contrast;

  1987.     *saturation= c->saturation;

  1988. 

  1989.     return 0;

  1990. }

  1991. 

  1992. static int handle_jpeg(int *format)

  1993. {

  1994.     switch (*format) {

  1995.         case PIX_FMT_YUVJ420P:

  1996.             *format = PIX_FMT_YUV420P;

  1997.             return 1;

  1998.         case PIX_FMT_YUVJ422P:

  1999.             *format = PIX_FMT_YUV422P;

  2000.             return 1;

  2001.         case PIX_FMT_YUVJ444P:

  2002.             *format = PIX_FMT_YUV444P;

  2003.             return 1;

  2004.         case PIX_FMT_YUVJ440P:

  2005.             *format = PIX_FMT_YUV440P;

  2006.             return 1;

  2007.         default:

  2008.             return 0;

  2009.     }

  2010. }

  2011. 

  2012. SwsContext *sws_getContext(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat, int flags,

  2013.                            SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){

  2014. 

  2015.     SwsContext *c;

  2016.     int i;

  2017.     int usesVFilter, usesHFilter;

  2018.     int unscaled, needsDither;

  2019.     int srcRange, dstRange;

  2020.     SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};

  2021. #if defined(ARCH_X86)

  2022.     if (flags & SWS_CPU_CAPS_MMX)

  2023.         asm volatile("emms\n\t"::: "memory");

  2024. #endif

  2025. 

  2026. #if !defined(RUNTIME_CPUDETECT) || !defined (CONFIG_GPL) //ensure that the flags match the compiled variant if cpudetect is off

  2027.     flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);

  2028. #ifdef HAVE_MMX2

  2029.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  2030. #elif defined (HAVE_3DNOW)

  2031.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  2032. #elif defined (HAVE_MMX)

  2033.     flags |= SWS_CPU_CAPS_MMX;

  2034. #elif defined (HAVE_ALTIVEC)

  2035.     flags |= SWS_CPU_CAPS_ALTIVEC;

  2036. #elif defined (ARCH_BFIN)

  2037.     flags |= SWS_CPU_CAPS_BFIN;

  2038. #endif

  2039. #endif /* RUNTIME_CPUDETECT */

  2040.     if (clip_table[512] != 255) globalInit();

  2041.     if (!rgb15to16) sws_rgb2rgb_init(flags);

  2042. 

  2043.     unscaled = (srcW == dstW && srcH == dstH);

  2044.     needsDither= (isBGR(dstFormat) || isRGB(dstFormat))

  2045.         && (fmt_depth(dstFormat))<24

  2046.         && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

  2047. 

  2048.     srcRange = handle_jpeg(&srcFormat);

  2049.     dstRange = handle_jpeg(&dstFormat);

  2050. 

  2051.     if (!isSupportedIn(srcFormat))

  2052.     {

  2053.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));

  2054.         return NULL;

  2055.     }

  2056.     if (!isSupportedOut(dstFormat))

  2057.     {

  2058.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));

  2059.         return NULL;

  2060.     }

  2061. 

  2062.     /* sanity check */

  2063.     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

  2064.     {

  2065.         av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",

  2066.                srcW, srcH, dstW, dstH);

  2067.         return NULL;

  2068.     }

  2069.     if(srcW > VOFW || dstW > VOFW){

  2070.         av_log(NULL, AV_LOG_ERROR, "swScaler: Compile time max width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");

  2071.         return NULL;

  2072.     }

  2073. 

  2074.     if (!dstFilter) dstFilter= &dummyFilter;

  2075.     if (!srcFilter) srcFilter= &dummyFilter;

  2076. 

  2077.     c= av_mallocz(sizeof(SwsContext));

  2078. 

  2079.     c->av_class = &sws_context_class;

  2080.     c->srcW= srcW;

  2081.     c->srcH= srcH;

  2082.     c->dstW= dstW;

  2083.     c->dstH= dstH;

  2084.     c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;

  2085.     c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;

  2086.     c->flags= flags;

  2087.     c->dstFormat= dstFormat;

  2088.     c->srcFormat= srcFormat;

  2089.     c->vRounder= 4* 0x0001000100010001ULL;

  2090. 

  2091.     usesHFilter= usesVFilter= 0;

  2092.     if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;

  2093.     if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;

  2094.     if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;

  2095.     if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;

  2096.     if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;

  2097.     if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;

  2098.     if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;

  2099.     if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;

  2100. 

  2101.     getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);

  2102.     getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

  2103. 

  2104.     // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation

  2105.     if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

  2106. 

  2107.     // drop some chroma lines if the user wants it

  2108.     c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;

  2109.     c->chrSrcVSubSample+= c->vChrDrop;

  2110. 

  2111.     // drop every 2. pixel for chroma calculation unless user wants full chroma

  2112.     if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)

  2113.       && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8

  2114.       && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4

  2115.       && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE)

  2116.         c->chrSrcHSubSample=1;

  2117. 

  2118.     if (param){

  2119.         c->param[0] = param[0];

  2120.         c->param[1] = param[1];

  2121.     }else{

  2122.         c->param[0] =

  2123.         c->param[1] = SWS_PARAM_DEFAULT;

  2124.     }

  2125. 

  2126.     c->chrIntHSubSample= c->chrDstHSubSample;

  2127.     c->chrIntVSubSample= c->chrSrcVSubSample;

  2128. 

  2129.     // Note the -((-x)>>y) is so that we always round toward +inf.

  2130.     c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);

  2131.     c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);

  2132.     c->chrDstW= -((-dstW) >> c->chrDstHSubSample);

  2133.     c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

  2134. 

  2135.     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);

  2136. 

  2137.     /* unscaled special Cases */

  2138.     if (unscaled && !usesHFilter && !usesVFilter)

  2139.     {

  2140.         /* yv12_to_nv12 */

  2141.         if (srcFormat == PIX_FMT_YUV420P && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))

  2142.         {

  2143.             c->swScale= PlanarToNV12Wrapper;

  2144.         }

  2145. #ifdef CONFIG_GPL

  2146.         /* yuv2bgr */

  2147.         if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat)))

  2148.         {

  2149.             c->swScale= yuv2rgb_get_func_ptr(c);

  2150.         }

  2151. #endif

  2152. 

  2153.         if (srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P)

  2154.         {

  2155.             c->swScale= yvu9toyv12Wrapper;

  2156.         }

  2157. 

  2158.         /* bgr24toYV12 */

  2159.         if (srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P)

  2160.             c->swScale= bgr24toyv12Wrapper;

  2161. 

  2162.         /* rgb/bgr -> rgb/bgr (no dither needed forms) */

  2163.         if (  (isBGR(srcFormat) || isRGB(srcFormat))

  2164.            && (isBGR(dstFormat) || isRGB(dstFormat))

  2165.            && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  2166.            && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  2167.            && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  2168.            && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  2169.            && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  2170.            && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  2171.            && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  2172.            && !needsDither)

  2173.              c->swScale= rgb2rgbWrapper;

  2174. 

  2175.         /* LQ converters if -sws 0 or -sws 4*/

  2176.         if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){

  2177.             /* rgb/bgr -> rgb/bgr (dither needed forms) */

  2178.             if ( (isBGR(srcFormat) || isRGB(srcFormat))

  2179.               && (isBGR(dstFormat) || isRGB(dstFormat))

  2180.               && needsDither)

  2181.                 c->swScale= rgb2rgbWrapper;

  2182. 

  2183.             /* yv12_to_yuy2 */

  2184.             if (srcFormat == PIX_FMT_YUV420P &&

  2185.                 (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422))

  2186.             {

  2187.                 if (dstFormat == PIX_FMT_YUYV422)

  2188.                     c->swScale= PlanarToYuy2Wrapper;

  2189.                 else

  2190.                     c->swScale= PlanarToUyvyWrapper;

  2191.             }

  2192.         }

  2193. 

  2194. #ifdef COMPILE_ALTIVEC

  2195.         if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  2196.             ((srcFormat == PIX_FMT_YUV420P &&

  2197.              (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422)))) {

  2198.           // unscaled YV12 -> packed YUV, we want speed

  2199.           if (dstFormat == PIX_FMT_YUYV422)

  2200.               c->swScale= yv12toyuy2_unscaled_altivec;

  2201.           else

  2202.               c->swScale= yv12touyvy_unscaled_altivec;

  2203.         }

  2204. #endif

  2205. 

  2206.         /* simple copy */

  2207.         if (  srcFormat == dstFormat

  2208.             || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  2209.             || (isPlanarYUV(dstFormat) && isGray(srcFormat)))

  2210.         {

  2211.             if (isPacked(c->srcFormat))

  2212.                 c->swScale= packedCopy;

  2213.             else /* Planar YUV or gray */

  2214.                 c->swScale= planarCopy;

  2215.         }

  2216. 

  2217.         /* gray16{le,be} conversions */

  2218.         if (isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))

  2219.         {

  2220.             c->swScale= gray16togray;

  2221.         }

  2222.         if ((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))

  2223.         {

  2224.             c->swScale= graytogray16;

  2225.         }

  2226.         if (srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))

  2227.         {

  2228.             c->swScale= gray16swap;

  2229.         }

  2230. 

  2231. #ifdef ARCH_BFIN

  2232.         if (flags & SWS_CPU_CAPS_BFIN)

  2233.             ff_bfin_get_unscaled_swscale (c);

  2234. #endif

  2235. 

  2236.         if (c->swScale){

  2237.             if (flags&SWS_PRINT_INFO)

  2238.                 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",

  2239.                                 sws_format_name(srcFormat), sws_format_name(dstFormat));

  2240.             return c;

  2241.         }

  2242.     }

  2243. 

  2244.     if (flags & SWS_CPU_CAPS_MMX2)

  2245.     {

  2246.         c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;

  2247.         if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))

  2248.         {

  2249.             if (flags&SWS_PRINT_INFO)

  2250.                 av_log(c, AV_LOG_INFO, "output Width is not a multiple of 32 -> no MMX2 scaler\n");

  2251.         }

  2252.         if (usesHFilter) c->canMMX2BeUsed=0;

  2253.     }

  2254.     else

  2255.         c->canMMX2BeUsed=0;

  2256. 

  2257.     c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;

  2258.     c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

  2259. 

  2260.     // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst

  2261.     // but only for the FAST_BILINEAR mode otherwise do correct scaling

  2262.     // n-2 is the last chrominance sample available

  2263.     // this is not perfect, but no one should notice the difference, the more correct variant

  2264.     // would be like the vertical one, but that would require some special code for the

  2265.     // first and last pixel

  2266.     if (flags&SWS_FAST_BILINEAR)

  2267.     {

  2268.         if (c->canMMX2BeUsed)

  2269.         {

  2270.             c->lumXInc+= 20;

  2271.             c->chrXInc+= 20;

  2272.         }

  2273.         //we don't use the x86asm scaler if mmx is available

  2274.         else if (flags & SWS_CPU_CAPS_MMX)

  2275.         {

  2276.             c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;

  2277.             c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;

  2278.         }

  2279.     }

  2280. 

  2281.     /* precalculate horizontal scaler filter coefficients */

  2282.     {

  2283.         const int filterAlign=

  2284.             (flags & SWS_CPU_CAPS_MMX) ? 4 :

  2285.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2286.             1;

  2287. 

  2288.         initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,

  2289.                    srcW      ,       dstW, filterAlign, 1<<14,

  2290.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2291.                    srcFilter->lumH, dstFilter->lumH, c->param);

  2292.         initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,

  2293.                    c->chrSrcW, c->chrDstW, filterAlign, 1<<14,

  2294.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2295.                    srcFilter->chrH, dstFilter->chrH, c->param);

  2296. 

  2297. #define MAX_FUNNY_CODE_SIZE 10000

  2298. #if defined(COMPILE_MMX2)

  2299. // can't downscale !!!

  2300.         if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))

  2301.         {

  2302. #ifdef MAP_ANONYMOUS

  2303.             c->funnyYCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2304.             c->funnyUVCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2305. #else

  2306.             c->funnyYCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2307.             c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2308. #endif

  2309. 

  2310.             c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));

  2311.             c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));

  2312.             c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));

  2313.             c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));

  2314. 

  2315.             initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);

  2316.             initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);

  2317.         }

  2318. #endif /* defined(COMPILE_MMX2) */

  2319.     } // Init Horizontal stuff

  2320. 

  2321. 

  2322. 

  2323.     /* precalculate vertical scaler filter coefficients */

  2324.     {

  2325.         const int filterAlign=

  2326.             (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :

  2327.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2328.             1;

  2329. 

  2330.         initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,

  2331.                    srcH      ,        dstH, filterAlign, (1<<12)-4,

  2332.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2333.                    srcFilter->lumV, dstFilter->lumV, c->param);

  2334.         initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,

  2335.                    c->chrSrcH, c->chrDstH, filterAlign, (1<<12)-4,

  2336.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2337.                    srcFilter->chrV, dstFilter->chrV, c->param);

  2338. 

  2339. #ifdef HAVE_ALTIVEC

  2340.         c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);

  2341.         c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);

  2342. 

  2343.         for (i=0;i<c->vLumFilterSize*c->dstH;i++) {

  2344.             int j;

  2345.             short *p = (short *)&c->vYCoeffsBank[i];

  2346.             for (j=0;j<8;j++)

  2347.                 p[j] = c->vLumFilter[i];

  2348.         }

  2349. 

  2350.         for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {

  2351.             int j;

  2352.             short *p = (short *)&c->vCCoeffsBank[i];

  2353.             for (j=0;j<8;j++)

  2354.                 p[j] = c->vChrFilter[i];

  2355.         }

  2356. #endif

  2357.     }

  2358. 

  2359.     // Calculate Buffer Sizes so that they won't run out while handling these damn slices

  2360.     c->vLumBufSize= c->vLumFilterSize;

  2361.     c->vChrBufSize= c->vChrFilterSize;

  2362.     for (i=0; i<dstH; i++)

  2363.     {

  2364.         int chrI= i*c->chrDstH / dstH;

  2365.         int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,

  2366.                            ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

  2367. 

  2368.         nextSlice>>= c->chrSrcVSubSample;

  2369.         nextSlice<<= c->chrSrcVSubSample;

  2370.         if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)

  2371.             c->vLumBufSize= nextSlice - c->vLumFilterPos[i];

  2372.         if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))

  2373.             c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];

  2374.     }

  2375. 

  2376.     // allocate pixbufs (we use dynamic allocation because otherwise we would need to

  2377.     c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));

  2378.     c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));

  2379.     //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)

  2380.     /* align at 16 bytes for AltiVec */

  2381.     for (i=0; i<c->vLumBufSize; i++)

  2382.         c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);

  2383.     for (i=0; i<c->vChrBufSize; i++)

  2384.         c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);

  2385. 

  2386.     //try to avoid drawing green stuff between the right end and the stride end

  2387.     for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);

  2388. 

  2389.     assert(2*VOFW == VOF);

  2390. 

  2391.     ASSERT(c->chrDstH <= dstH)

  2392. 

  2393.     if (flags&SWS_PRINT_INFO)

  2394.     {

  2395. #ifdef DITHER1XBPP

  2396.         const char *dither= " dithered";

  2397. #else

  2398.         const char *dither= "";

  2399. #endif

  2400.         if (flags&SWS_FAST_BILINEAR)

  2401.             av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");

  2402.         else if (flags&SWS_BILINEAR)

  2403.             av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");

  2404.         else if (flags&SWS_BICUBIC)

  2405.             av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");

  2406.         else if (flags&SWS_X)

  2407.             av_log(c, AV_LOG_INFO, "Experimental scaler, ");

  2408.         else if (flags&SWS_POINT)

  2409.             av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");

  2410.         else if (flags&SWS_AREA)

  2411.             av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");

  2412.         else if (flags&SWS_BICUBLIN)

  2413.             av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");

  2414.         else if (flags&SWS_GAUSS)

  2415.             av_log(c, AV_LOG_INFO, "Gaussian scaler, ");

  2416.         else if (flags&SWS_SINC)

  2417.             av_log(c, AV_LOG_INFO, "Sinc scaler, ");

  2418.         else if (flags&SWS_LANCZOS)

  2419.             av_log(c, AV_LOG_INFO, "Lanczos scaler, ");

  2420.         else if (flags&SWS_SPLINE)

  2421.             av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");

  2422.         else

  2423.             av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");

  2424. 

  2425.         if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)

  2426.             av_log(c, AV_LOG_INFO, "from %s to%s %s ",

  2427.                    sws_format_name(srcFormat), dither, sws_format_name(dstFormat));

  2428.         else

  2429.             av_log(c, AV_LOG_INFO, "from %s to %s ",

  2430.                    sws_format_name(srcFormat), sws_format_name(dstFormat));

  2431. 

  2432.         if (flags & SWS_CPU_CAPS_MMX2)

  2433.             av_log(c, AV_LOG_INFO, "using MMX2\n");

  2434.         else if (flags & SWS_CPU_CAPS_3DNOW)

  2435.             av_log(c, AV_LOG_INFO, "using 3DNOW\n");

  2436.         else if (flags & SWS_CPU_CAPS_MMX)

  2437.             av_log(c, AV_LOG_INFO, "using MMX\n");

  2438.         else if (flags & SWS_CPU_CAPS_ALTIVEC)

  2439.             av_log(c, AV_LOG_INFO, "using AltiVec\n");

  2440.         else

  2441.             av_log(c, AV_LOG_INFO, "using C\n");

  2442.     }

  2443. 

  2444.     if (flags & SWS_PRINT_INFO)

  2445.     {

  2446.         if (flags & SWS_CPU_CAPS_MMX)

  2447.         {

  2448.             if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))

  2449.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");

  2450.             else

  2451.             {

  2452.                 if (c->hLumFilterSize==4)

  2453.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");

  2454.                 else if (c->hLumFilterSize==8)

  2455.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");

  2456.                 else

  2457.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");

  2458. 

  2459.                 if (c->hChrFilterSize==4)

  2460.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");

  2461.                 else if (c->hChrFilterSize==8)

  2462.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");

  2463.                 else

  2464.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");

  2465.             }

  2466.         }

  2467.         else

  2468.         {

  2469. #if defined(ARCH_X86)

  2470.             av_log(c, AV_LOG_VERBOSE, "using X86-Asm scaler for horizontal scaling\n");

  2471. #else

  2472.             if (flags & SWS_FAST_BILINEAR)

  2473.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");

  2474.             else

  2475.                 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");

  2476. #endif

  2477.         }

  2478.         if (isPlanarYUV(dstFormat))

  2479.         {

  2480.             if (c->vLumFilterSize==1)

  2481.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2482.             else

  2483.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2484.         }

  2485.         else

  2486.         {

  2487.             if (c->vLumFilterSize==1 && c->vChrFilterSize==2)

  2488.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"

  2489.                        "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2490.             else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)

  2491.                 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2492.             else

  2493.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2494.         }

  2495. 

  2496.         if (dstFormat==PIX_FMT_BGR24)

  2497.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 Converter\n",

  2498.                    (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));

  2499.         else if (dstFormat==PIX_FMT_RGB32)

  2500.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2501.         else if (dstFormat==PIX_FMT_BGR565)

  2502.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2503.         else if (dstFormat==PIX_FMT_BGR555)

  2504.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2505. 

  2506.         av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);

  2507.     }

  2508.     if (flags & SWS_PRINT_INFO)

  2509.     {

  2510.         av_log(c, AV_LOG_DEBUG, "Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2511.                c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);

  2512.         av_log(c, AV_LOG_DEBUG, "Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2513.                c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);

  2514.     }

  2515. 

  2516.     c->swScale= getSwsFunc(flags);

  2517.     return c;

  2518. }

  2519. 

  2520. /**

  2521.  * swscale wrapper, so we don't need to export the SwsContext.

  2522.  * assumes planar YUV to be in YUV order instead of YVU

  2523.  */

  2524. int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2525.               int srcSliceH, uint8_t* dst[], int dstStride[]){

  2526.     int i;

  2527.     uint8_t* src2[4]= {src[0], src[1], src[2]};

  2528.     uint32_t pal[256];

  2529.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  2530.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  2531.         return 0;

  2532.     }

  2533.     if (c->sliceDir == 0) {

  2534.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  2535.     }

  2536. 

  2537.     if (c->srcFormat == PIX_FMT_PAL8){

  2538.         for (i=0; i<256; i++){

  2539.             int p= ((uint32_t*)(src[1]))[i];

  2540.             int r= (p>>16)&0xFF;

  2541.             int g= (p>> 8)&0xFF;

  2542.             int b=  p     &0xFF;

  2543.             int y= av_clip_uint8(((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16 );

  2544.             int u= av_clip_uint8(((RU*r + GU*g + BU*b)>>RGB2YUV_SHIFT) + 128);

  2545.             int v= av_clip_uint8(((RV*r + GV*g + BV*b)>>RGB2YUV_SHIFT) + 128);

  2546.             pal[i]= y + (u<<8) + (v<<16);

  2547.         }

  2548.         src2[1]= (uint8_t*)pal;

  2549.     }

  2550. 

  2551.     // copy strides, so they can safely be modified

  2552.     if (c->sliceDir == 1) {

  2553.         // slices go from top to bottom

  2554.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2]};

  2555.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2]};

  2556.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);

  2557.     } else {

  2558.         // slices go from bottom to top => we flip the image internally

  2559.         uint8_t* dst2[4]= {dst[0] + (c->dstH-1)*dstStride[0],

  2560.                            dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],

  2561.                            dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2]};

  2562.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2]};

  2563.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2]};

  2564. 

  2565.         src2[0] += (srcSliceH-1)*srcStride[0];

  2566.         if (c->srcFormat != PIX_FMT_PAL8)

  2567.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  2568.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  2569. 

  2570.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  2571.     }

  2572. }

  2573. 

  2574. /**

  2575.  * swscale wrapper, so we don't need to export the SwsContext

  2576.  */

  2577. int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2578.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  2579.     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);

  2580. }

  2581. 

  2582. SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,

  2583.                                 float lumaSharpen, float chromaSharpen,

  2584.                                 float chromaHShift, float chromaVShift,

  2585.                                 int verbose)

  2586. {

  2587.     SwsFilter *filter= av_malloc(sizeof(SwsFilter));

  2588. 

  2589.     if (lumaGBlur!=0.0){

  2590.         filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);

  2591.         filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);

  2592.     }else{

  2593.         filter->lumH= sws_getIdentityVec();

  2594.         filter->lumV= sws_getIdentityVec();

  2595.     }

  2596. 

  2597.     if (chromaGBlur!=0.0){

  2598.         filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);

  2599.         filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);

  2600.     }else{

  2601.         filter->chrH= sws_getIdentityVec();

  2602.         filter->chrV= sws_getIdentityVec();

  2603.     }

  2604. 

  2605.     if (chromaSharpen!=0.0){

  2606.         SwsVector *id= sws_getIdentityVec();

  2607.         sws_scaleVec(filter->chrH, -chromaSharpen);

  2608.         sws_scaleVec(filter->chrV, -chromaSharpen);

  2609.         sws_addVec(filter->chrH, id);

  2610.         sws_addVec(filter->chrV, id);

  2611.         sws_freeVec(id);

  2612.     }

  2613. 

  2614.     if (lumaSharpen!=0.0){

  2615.         SwsVector *id= sws_getIdentityVec();

  2616.         sws_scaleVec(filter->lumH, -lumaSharpen);

  2617.         sws_scaleVec(filter->lumV, -lumaSharpen);

  2618.         sws_addVec(filter->lumH, id);

  2619.         sws_addVec(filter->lumV, id);

  2620.         sws_freeVec(id);

  2621.     }

  2622. 

  2623.     if (chromaHShift != 0.0)

  2624.         sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));

  2625. 

  2626.     if (chromaVShift != 0.0)

  2627.         sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));

  2628. 

  2629.     sws_normalizeVec(filter->chrH, 1.0);

  2630.     sws_normalizeVec(filter->chrV, 1.0);

  2631.     sws_normalizeVec(filter->lumH, 1.0);

  2632.     sws_normalizeVec(filter->lumV, 1.0);

  2633. 

  2634.     if (verbose) sws_printVec(filter->chrH);

  2635.     if (verbose) sws_printVec(filter->lumH);

  2636. 

  2637.     return filter;

  2638. }

  2639. 

  2640. /**

  2641.  * returns a normalized gaussian curve used to filter stuff

  2642.  * quality=3 is high quality, lowwer is lowwer quality

  2643.  */

  2644. SwsVector *sws_getGaussianVec(double variance, double quality){

  2645.     const int length= (int)(variance*quality + 0.5) | 1;

  2646.     int i;

  2647.     double *coeff= av_malloc(length*sizeof(double));

  2648.     double middle= (length-1)*0.5;

  2649.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2650. 

  2651.     vec->coeff= coeff;

  2652.     vec->length= length;

  2653. 

  2654.     for (i=0; i<length; i++)

  2655.     {

  2656.         double dist= i-middle;

  2657.         coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);

  2658.     }

  2659. 

  2660.     sws_normalizeVec(vec, 1.0);

  2661. 

  2662.     return vec;

  2663. }

  2664. 

  2665. SwsVector *sws_getConstVec(double c, int length){

  2666.     int i;

  2667.     double *coeff= av_malloc(length*sizeof(double));

  2668.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2669. 

  2670.     vec->coeff= coeff;

  2671.     vec->length= length;

  2672. 

  2673.     for (i=0; i<length; i++)

  2674.         coeff[i]= c;

  2675. 

  2676.     return vec;

  2677. }

  2678. 

  2679. 

  2680. SwsVector *sws_getIdentityVec(void){

  2681.     return sws_getConstVec(1.0, 1);

  2682. }

  2683. 

  2684. double sws_dcVec(SwsVector *a){

  2685.     int i;

  2686.     double sum=0;

  2687. 

  2688.     for (i=0; i<a->length; i++)

  2689.         sum+= a->coeff[i];

  2690. 

  2691.     return sum;

  2692. }

  2693. 

  2694. void sws_scaleVec(SwsVector *a, double scalar){

  2695.     int i;

  2696. 

  2697.     for (i=0; i<a->length; i++)

  2698.         a->coeff[i]*= scalar;

  2699. }

  2700. 

  2701. void sws_normalizeVec(SwsVector *a, double height){

  2702.     sws_scaleVec(a, height/sws_dcVec(a));

  2703. }

  2704. 

  2705. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){

  2706.     int length= a->length + b->length - 1;

  2707.     double *coeff= av_malloc(length*sizeof(double));

  2708.     int i, j;

  2709.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2710. 

  2711.     vec->coeff= coeff;

  2712.     vec->length= length;

  2713. 

  2714.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2715. 

  2716.     for (i=0; i<a->length; i++)

  2717.     {

  2718.         for (j=0; j<b->length; j++)

  2719.         {

  2720.             coeff[i+j]+= a->coeff[i]*b->coeff[j];

  2721.         }

  2722.     }

  2723. 

  2724.     return vec;

  2725. }

  2726. 

  2727. static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){

  2728.     int length= FFMAX(a->length, b->length);

  2729.     double *coeff= av_malloc(length*sizeof(double));

  2730.     int i;

  2731.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2732. 

  2733.     vec->coeff= coeff;

  2734.     vec->length= length;

  2735. 

  2736.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2737. 

  2738.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  2739.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];

  2740. 

  2741.     return vec;

  2742. }

  2743. 

  2744. static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){

  2745.     int length= FFMAX(a->length, b->length);

  2746.     double *coeff= av_malloc(length*sizeof(double));

  2747.     int i;

  2748.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2749. 

  2750.     vec->coeff= coeff;

  2751.     vec->length= length;

  2752. 

  2753.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2754. 

  2755.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  2756.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];

  2757. 

  2758.     return vec;

  2759. }

  2760. 

  2761. /* shift left / or right if "shift" is negative */

  2762. static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){

  2763.     int length= a->length + FFABS(shift)*2;

  2764.     double *coeff= av_malloc(length*sizeof(double));

  2765.     int i;

  2766.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2767. 

  2768.     vec->coeff= coeff;

  2769.     vec->length= length;

  2770. 

  2771.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2772. 

  2773.     for (i=0; i<a->length; i++)

  2774.     {

  2775.         coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];

  2776.     }

  2777. 

  2778.     return vec;

  2779. }

  2780. 

  2781. void sws_shiftVec(SwsVector *a, int shift){

  2782.     SwsVector *shifted= sws_getShiftedVec(a, shift);

  2783.     av_free(a->coeff);

  2784.     a->coeff= shifted->coeff;

  2785.     a->length= shifted->length;

  2786.     av_free(shifted);

  2787. }

  2788. 

  2789. void sws_addVec(SwsVector *a, SwsVector *b){

  2790.     SwsVector *sum= sws_sumVec(a, b);

  2791.     av_free(a->coeff);

  2792.     a->coeff= sum->coeff;

  2793.     a->length= sum->length;

  2794.     av_free(sum);

  2795. }

  2796. 

  2797. void sws_subVec(SwsVector *a, SwsVector *b){

  2798.     SwsVector *diff= sws_diffVec(a, b);

  2799.     av_free(a->coeff);

  2800.     a->coeff= diff->coeff;

  2801.     a->length= diff->length;

  2802.     av_free(diff);

  2803. }

  2804. 

  2805. void sws_convVec(SwsVector *a, SwsVector *b){

  2806.     SwsVector *conv= sws_getConvVec(a, b);

  2807.     av_free(a->coeff);

  2808.     a->coeff= conv->coeff;

  2809.     a->length= conv->length;

  2810.     av_free(conv);

  2811. }

  2812. 

  2813. SwsVector *sws_cloneVec(SwsVector *a){

  2814.     double *coeff= av_malloc(a->length*sizeof(double));

  2815.     int i;

  2816.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2817. 

  2818.     vec->coeff= coeff;

  2819.     vec->length= a->length;

  2820. 

  2821.     for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];

  2822. 

  2823.     return vec;

  2824. }

  2825. 

  2826. void sws_printVec(SwsVector *a){

  2827.     int i;

  2828.     double max=0;

  2829.     double min=0;

  2830.     double range;

  2831. 

  2832.     for (i=0; i<a->length; i++)

  2833.         if (a->coeff[i]>max) max= a->coeff[i];

  2834. 

  2835.     for (i=0; i<a->length; i++)

  2836.         if (a->coeff[i]<min) min= a->coeff[i];

  2837. 

  2838.     range= max - min;

  2839. 

  2840.     for (i=0; i<a->length; i++)

  2841.     {

  2842.         int x= (int)((a->coeff[i]-min)*60.0/range +0.5);

  2843.         av_log(NULL, AV_LOG_DEBUG, "%1.3f ", a->coeff[i]);

  2844.         for (;x>0; x--) av_log(NULL, AV_LOG_DEBUG, " ");

  2845.         av_log(NULL, AV_LOG_DEBUG, "|\n");

  2846.     }

  2847. }

  2848. 

  2849. void sws_freeVec(SwsVector *a){

  2850.     if (!a) return;

  2851.     av_free(a->coeff);

  2852.     a->coeff=NULL;

  2853.     a->length=0;

  2854.     av_free(a);

  2855. }

  2856. 

  2857. void sws_freeFilter(SwsFilter *filter){

  2858.     if (!filter) return;

  2859. 

  2860.     if (filter->lumH) sws_freeVec(filter->lumH);

  2861.     if (filter->lumV) sws_freeVec(filter->lumV);

  2862.     if (filter->chrH) sws_freeVec(filter->chrH);

  2863.     if (filter->chrV) sws_freeVec(filter->chrV);

  2864.     av_free(filter);

  2865. }

  2866. 

  2867. 

  2868. void sws_freeContext(SwsContext *c){

  2869.     int i;

  2870.     if (!c) return;

  2871. 

  2872.     if (c->lumPixBuf)

  2873.     {

  2874.         for (i=0; i<c->vLumBufSize; i++)

  2875.         {

  2876.             av_free(c->lumPixBuf[i]);

  2877.             c->lumPixBuf[i]=NULL;

  2878.         }

  2879.         av_free(c->lumPixBuf);

  2880.         c->lumPixBuf=NULL;

  2881.     }

  2882. 

  2883.     if (c->chrPixBuf)

  2884.     {

  2885.         for (i=0; i<c->vChrBufSize; i++)

  2886.         {

  2887.             av_free(c->chrPixBuf[i]);

  2888.             c->chrPixBuf[i]=NULL;

  2889.         }

  2890.         av_free(c->chrPixBuf);

  2891.         c->chrPixBuf=NULL;

  2892.     }

  2893. 

  2894.     av_free(c->vLumFilter);

  2895.     c->vLumFilter = NULL;

  2896.     av_free(c->vChrFilter);

  2897.     c->vChrFilter = NULL;

  2898.     av_free(c->hLumFilter);

  2899.     c->hLumFilter = NULL;

  2900.     av_free(c->hChrFilter);

  2901.     c->hChrFilter = NULL;

  2902. #ifdef HAVE_ALTIVEC

  2903.     av_free(c->vYCoeffsBank);

  2904.     c->vYCoeffsBank = NULL;

  2905.     av_free(c->vCCoeffsBank);

  2906.     c->vCCoeffsBank = NULL;

  2907. #endif

  2908. 

  2909.     av_free(c->vLumFilterPos);

  2910.     c->vLumFilterPos = NULL;

  2911.     av_free(c->vChrFilterPos);

  2912.     c->vChrFilterPos = NULL;

  2913.     av_free(c->hLumFilterPos);

  2914.     c->hLumFilterPos = NULL;

  2915.     av_free(c->hChrFilterPos);

  2916.     c->hChrFilterPos = NULL;

  2917. 

  2918. #if defined(ARCH_X86) && defined(CONFIG_GPL)

  2919. #ifdef MAP_ANONYMOUS

  2920.     if (c->funnyYCode) munmap(c->funnyYCode, MAX_FUNNY_CODE_SIZE);

  2921.     if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);

  2922. #else

  2923.     av_free(c->funnyYCode);

  2924.     av_free(c->funnyUVCode);

  2925. #endif

  2926.     c->funnyYCode=NULL;

  2927.     c->funnyUVCode=NULL;

  2928. #endif /* defined(ARCH_X86) */

  2929. 

  2930.     av_free(c->lumMmx2Filter);

  2931.     c->lumMmx2Filter=NULL;

  2932.     av_free(c->chrMmx2Filter);

  2933.     c->chrMmx2Filter=NULL;

  2934.     av_free(c->lumMmx2FilterPos);

  2935.     c->lumMmx2FilterPos=NULL;

  2936.     av_free(c->chrMmx2FilterPos);

  2937.     c->chrMmx2FilterPos=NULL;

  2938.     av_free(c->yuvTable);

  2939.     c->yuvTable=NULL;

  2940. 

  2941.     av_free(c);

  2942. }

  2943. 

  2944. /**

  2945.  * Checks if context is valid or reallocs a new one instead.

  2946.  * If context is NULL, just calls sws_getContext() to get a new one.

  2947.  * Otherwise, checks if the parameters are the same already saved in context.

  2948.  * If that is the case, returns the current context.

  2949.  * Otherwise, frees context and gets a new one.

  2950.  *

  2951.  * Be warned that srcFilter, dstFilter are not checked, they are

  2952.  * asumed to remain valid.

  2953.  */

  2954. struct SwsContext *sws_getCachedContext(struct SwsContext *context,

  2955.                                         int srcW, int srcH, int srcFormat,

  2956.                                         int dstW, int dstH, int dstFormat, int flags,

  2957.                                         SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)

  2958. {

  2959.     static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};

  2960. 

  2961.     if (!param)

  2962.         param = default_param;

  2963. 

  2964.     if (context) {

  2965.         if (context->srcW != srcW || context->srcH != srcH ||

  2966.             context->srcFormat != srcFormat ||

  2967.             context->dstW != dstW || context->dstH != dstH ||

  2968.             context->dstFormat != dstFormat || context->flags != flags ||

  2969.             context->param[0] != param[0] || context->param[1] != param[1])

  2970.         {

  2971.             sws_freeContext(context);

  2972.             context = NULL;

  2973.         }

  2974.     }

  2975.     if (!context) {

  2976.         return sws_getContext(srcW, srcH, srcFormat,

  2977.                               dstW, dstH, dstFormat, flags,

  2978.                               srcFilter, dstFilter, param);

  2979.     }

  2980.     return context;

  2981. }