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.
9447 Commits
32 Branches
301MB
Tree: 4440b1188e
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '4440b1188e'
${ noResults }
FFmpeg/libswscale/swscale.c

2930 lines
99KB
Raw Blame History 

  1. /*

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

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

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

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

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

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

  10.  *

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

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

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

  14.  * GNU General Public License for more details.

  15.  *

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

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

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

  19.  *

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

  21.  * under the LGPL license too

  22.  */

  23. 

  24. /*

  25.   supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09, 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 didnt write it down ...)

  42.  YV12 -> BGR16

  43.  YV12 -> YV12

  44.  BGR15 -> BGR16

  45.  BGR16 -> BGR16

  46.  YVU9 -> YV12

  47. 

  48. untested special converters

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

  50.   YV12/I420 -> YV12/I420

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

  52.   BGR24 -> BGR32 & RGB24 -> RGB32

  53.   BGR32 -> BGR24 & RGB32 -> RGB24

  54.   BGR24 -> YV12

  55. */

  56. 

  57. #include <inttypes.h>

  58. #include <string.h>

  59. #include <math.h>

  60. #include <stdio.h>

  61. #include <unistd.h>

  62. #include "config.h"

  63. #include <assert.h>

  64. #ifdef HAVE_SYS_MMAN_H

  65. #include <sys/mman.h>

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

  67. #define MAP_ANONYMOUS MAP_ANON

  68. #endif

  69. #endif

  70. #include "swscale.h"

  71. #include "swscale_internal.h"

  72. #include "x86_cpu.h"

  73. #include "bswap.h"

  74. #include "rgb2rgb.h"

  75. #ifdef USE_FASTMEMCPY

  76. #include "libvo/fastmemcpy.h"

  77. #endif

  78. 

  79. #undef MOVNTQ

  80. #undef PAVGB

  81. 

  82. //#undef HAVE_MMX2

  83. //#define HAVE_3DNOW

  84. //#undef HAVE_MMX

  85. //#undef ARCH_X86

  86. //#define WORDS_BIGENDIAN

  87. #define DITHER1XBPP

  88. 

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

  90. 

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

  92. 

  93. #ifdef MP_DEBUG

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

  95. #else

  96. #define ASSERT(x) ;

  97. #endif

  98. 

  99. #ifdef M_PI

  100. #define PI M_PI

  101. #else

  102. #define PI 3.14159265358979323846

  103. #endif

  104. 

  105. #define isSupportedIn(x)    (       \

  106.            (x)==PIX_FMT_YUV420P     \

  107.         || (x)==PIX_FMT_YUYV422     \

  108.         || (x)==PIX_FMT_UYVY422     \

  109.         || (x)==PIX_FMT_RGB32       \

  110.         || (x)==PIX_FMT_BGR24       \

  111.         || (x)==PIX_FMT_BGR565      \

  112.         || (x)==PIX_FMT_BGR555      \

  113.         || (x)==PIX_FMT_BGR32       \

  114.         || (x)==PIX_FMT_RGB24       \

  115.         || (x)==PIX_FMT_RGB565      \

  116.         || (x)==PIX_FMT_RGB555      \

  117.         || (x)==PIX_FMT_GRAY8       \

  118.         || (x)==PIX_FMT_YUV410P     \

  119.         || (x)==PIX_FMT_GRAY16BE    \

  120.         || (x)==PIX_FMT_GRAY16LE    \

  121.         || (x)==PIX_FMT_YUV444P     \

  122.         || (x)==PIX_FMT_YUV422P     \

  123.         || (x)==PIX_FMT_YUV411P     \

  124.         || (x)==PIX_FMT_PAL8        \

  125.         || (x)==PIX_FMT_BGR8        \

  126.         || (x)==PIX_FMT_RGB8        \

  127.         || (x)==PIX_FMT_BGR4_BYTE   \

  128.         || (x)==PIX_FMT_RGB4_BYTE   \

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

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

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

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

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

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

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

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

  198.         0x0103010301030103LL,

  199.         0x0200020002000200LL,};

  200. 

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

  202.         0x0602060206020602LL,

  203.         0x0004000400040004LL,};

  204. 

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

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

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

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

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

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

  211. 

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

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

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

  215. 

  216. #ifdef FAST_BGR2YV12

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

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

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

  220. #else

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

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

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

  224. #endif /* FAST_BGR2YV12 */

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

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

  227. static const uint64_t w1111        attribute_used __attribute__((aligned(8))) = 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. static AVClass sws_context_class = { "SWScaler", sws_context_to_name, NULL };

  246. 

  247. char *sws_format_name(enum PixelFormat format)

  248. {

  249.     switch (format) {

  250.         case PIX_FMT_YUV420P:

  251.             return "yuv420p";

  252.         case PIX_FMT_YUYV422:

  253.             return "yuyv422";

  254.         case PIX_FMT_RGB24:

  255.             return "rgb24";

  256.         case PIX_FMT_BGR24:

  257.             return "bgr24";

  258.         case PIX_FMT_YUV422P:

  259.             return "yuv422p";

  260.         case PIX_FMT_YUV444P:

  261.             return "yuv444p";

  262.         case PIX_FMT_RGB32:

  263.             return "rgb32";

  264.         case PIX_FMT_YUV410P:

  265.             return "yuv410p";

  266.         case PIX_FMT_YUV411P:

  267.             return "yuv411p";

  268.         case PIX_FMT_RGB565:

  269.             return "rgb565";

  270.         case PIX_FMT_RGB555:

  271.             return "rgb555";

  272.         case PIX_FMT_GRAY16BE:

  273.             return "gray16be";

  274.         case PIX_FMT_GRAY16LE:

  275.             return "gray16le";

  276.         case PIX_FMT_GRAY8:

  277.             return "gray8";

  278.         case PIX_FMT_MONOWHITE:

  279.             return "mono white";

  280.         case PIX_FMT_MONOBLACK:

  281.             return "mono black";

  282.         case PIX_FMT_PAL8:

  283.             return "Palette";

  284.         case PIX_FMT_YUVJ420P:

  285.             return "yuvj420p";

  286.         case PIX_FMT_YUVJ422P:

  287.             return "yuvj422p";

  288.         case PIX_FMT_YUVJ444P:

  289.             return "yuvj444p";

  290.         case PIX_FMT_XVMC_MPEG2_MC:

  291.             return "xvmc_mpeg2_mc";

  292.         case PIX_FMT_XVMC_MPEG2_IDCT:

  293.             return "xvmc_mpeg2_idct";

  294.         case PIX_FMT_UYVY422:

  295.             return "uyvy422";

  296.         case PIX_FMT_UYYVYY411:

  297.             return "uyyvyy411";

  298.         case PIX_FMT_RGB32_1:

  299.             return "rgb32x";

  300.         case PIX_FMT_BGR32_1:

  301.             return "bgr32x";

  302.         case PIX_FMT_BGR32:

  303.             return "bgr32";

  304.         case PIX_FMT_BGR565:

  305.             return "bgr565";

  306.         case PIX_FMT_BGR555:

  307.             return "bgr555";

  308.         case PIX_FMT_BGR8:

  309.             return "bgr8";

  310.         case PIX_FMT_BGR4:

  311.             return "bgr4";

  312.         case PIX_FMT_BGR4_BYTE:

  313.             return "bgr4 byte";

  314.         case PIX_FMT_RGB8:

  315.             return "rgb8";

  316.         case PIX_FMT_RGB4:

  317.             return "rgb4";

  318.         case PIX_FMT_RGB4_BYTE:

  319.             return "rgb4 byte";

  320.         case PIX_FMT_NV12:

  321.             return "nv12";

  322.         case PIX_FMT_NV21:

  323.             return "nv21";

  324.         default:

  325.             return "Unknown format";

  326.     }

  327. }

  328. 

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

  330. void in_asm_used_var_warning_killer()

  331. {

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

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

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

  335.     if (i) i=0;

  336. }

  337. #endif

  338. 

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

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

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

  342. {

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

  344.     int i;

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

  346.     {

  347.         int val=1<<18;

  348.         int j;

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

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

  351. 

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

  353.     }

  354. 

  355.     if (uDest != NULL)

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

  357.         {

  358.             int u=1<<18;

  359.             int v=1<<18;

  360.             int j;

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

  362.             {

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

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

  365.             }

  366. 

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

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

  369.         }

  370. }

  371. 

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

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

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

  375. {

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

  377.     int i;

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

  379.     {

  380.         int val=1<<18;

  381.         int j;

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

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

  384. 

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

  386.     }

  387. 

  388.     if (uDest == NULL)

  389.         return;

  390. 

  391.     if (dstFormat == PIX_FMT_NV12)

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

  393.         {

  394.             int u=1<<18;

  395.             int v=1<<18;

  396.             int j;

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

  398.             {

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

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

  401.             }

  402. 

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

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

  405.         }

  406.     else

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

  408.         {

  409.             int u=1<<18;

  410.             int v=1<<18;

  411.             int j;

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

  413.             {

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

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

  416.             }

  417. 

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

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

  420.         }

  421. }

  422. 

  423. #define YSCALE_YUV_2_PACKEDX_C(type) \

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

  425.         int j;\

  426.         int Y1 = 1<<18;\

  427.         int Y2 = 1<<18;\

  428.         int U  = 1<<18;\

  429.         int V  = 1<<18;\

  430.         type attribute_unused *r, *b, *g;\

  431.         const int i2= 2*i;\

  432.         \

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

  434.         {\

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

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

  437.         }\

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

  439.         {\

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

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

  442.         }\

  443.         Y1>>=19;\

  444.         Y2>>=19;\

  445.         U >>=19;\

  446.         V >>=19;\

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

  448.         {\

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

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

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

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

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

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

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

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

  457.         }

  458. 

  459. #define YSCALE_YUV_2_RGBX_C(type) \

  460.     YSCALE_YUV_2_PACKEDX_C(type)  \

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

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

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

  464. 

  465. #define YSCALE_YUV_2_PACKED2_C   \

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

  467.         const int i2= 2*i;       \

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

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

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

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

  472. 

  473. #define YSCALE_YUV_2_RGB2_C(type) \

  474.     YSCALE_YUV_2_PACKED2_C\

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

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

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

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

  479. 

  480. #define YSCALE_YUV_2_PACKED1_C \

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

  482.         const int i2= 2*i;\

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

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

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

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

  487. 

  488. #define YSCALE_YUV_2_RGB1_C(type) \

  489.     YSCALE_YUV_2_PACKED1_C\

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

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

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

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

  494. 

  495. #define YSCALE_YUV_2_PACKED1B_C \

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

  497.         const int i2= 2*i;\

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

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

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

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

  502. 

  503. #define YSCALE_YUV_2_RGB1B_C(type) \

  504.     YSCALE_YUV_2_PACKED1B_C\

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

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

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

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

  509. 

  510. #define YSCALE_YUV_2_ANYRGB_C(func, func2)\

  511.     switch(c->dstFormat)\

  512.     {\

  513.     case PIX_FMT_RGB32:\

  514.     case PIX_FMT_BGR32:\

  515.         func(uint32_t)\

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

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

  518.         }                \

  519.         break;\

  520.     case PIX_FMT_RGB24:\

  521.         func(uint8_t)\

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

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

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

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

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

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

  528.             dest+=6;\

  529.         }\

  530.         break;\

  531.     case PIX_FMT_BGR24:\

  532.         func(uint8_t)\

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

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

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

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

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

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

  539.             dest+=6;\

  540.         }\

  541.         break;\

  542.     case PIX_FMT_RGB565:\

  543.     case PIX_FMT_BGR565:\

  544.         {\

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

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

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

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

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

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

  551.             func(uint16_t)\

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

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

  554.             }\

  555.         }\

  556.         break;\

  557.     case PIX_FMT_RGB555:\

  558.     case PIX_FMT_BGR555:\

  559.         {\

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

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

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

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

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

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

  566.             func(uint16_t)\

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

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

  569.             }\

  570.         }\

  571.         break;\

  572.     case PIX_FMT_RGB8:\

  573.     case PIX_FMT_BGR8:\

  574.         {\

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

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

  577.             func(uint8_t)\

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

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

  580.             }\

  581.         }\

  582.         break;\

  583.     case PIX_FMT_RGB4:\

  584.     case PIX_FMT_BGR4:\

  585.         {\

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

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

  588.             func(uint8_t)\

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

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

  591.             }\

  592.         }\

  593.         break;\

  594.     case PIX_FMT_RGB4_BYTE:\

  595.     case PIX_FMT_BGR4_BYTE:\

  596.         {\

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

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

  599.             func(uint8_t)\

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

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

  602.             }\

  603.         }\

  604.         break;\

  605.     case PIX_FMT_MONOBLACK:\

  606.         {\

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

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

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

  610.                 int acc;\

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

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

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

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

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

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

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

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

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

  620.                 dest++;\

  621.             }\

  622. \

  623. /*\

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

  625. {\

  626.             int acc=0;\

  627.             int left=0;\

  628.             static int top[1024];\

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

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

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

  632.             int topLeft=0;\

  633.             int shift=0;\

  634.             int count=0;\

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

  636.             int error_new=0;\

  637.             int error_in3=0;\

  638.             int f=0;\

  639.             \

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

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

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

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

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

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

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

  647. \

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

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

  650.                 f= error_new - error_in3*4;\

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

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

  653. \

  654.                 topLeft= top[i];\

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

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

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

  658. \

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

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

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

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

  663.                 }\

  664.             }\

  665. }\

  666. */\

  667.         }\

  668.         break;\

  669.     case PIX_FMT_YUYV422:\

  670.         func2\

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

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

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

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

  675.         }                \

  676.         break;\

  677.     case PIX_FMT_UYVY422:\

  678.         func2\

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

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

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

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

  683.         }                \

  684.         break;\

  685.     }\

  686. 

  687. 

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

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

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

  691. {

  692.     int i;

  693.     switch(c->dstFormat)

  694.     {

  695.     case PIX_FMT_BGR32:

  696.     case PIX_FMT_RGB32:

  697.         YSCALE_YUV_2_RGBX_C(uint32_t)

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

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

  700.         }

  701.         break;

  702.     case PIX_FMT_RGB24:

  703.         YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

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

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

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

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

  710.             dest+=6;

  711.         }

  712.         break;

  713.     case PIX_FMT_BGR24:

  714.         YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

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

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

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

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

  721.             dest+=6;

  722.         }

  723.         break;

  724.     case PIX_FMT_RGB565:

  725.     case PIX_FMT_BGR565:

  726.         {

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

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

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

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

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

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

  733.             YSCALE_YUV_2_RGBX_C(uint16_t)

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

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

  736.             }

  737.         }

  738.         break;

  739.     case PIX_FMT_RGB555:

  740.     case PIX_FMT_BGR555:

  741.         {

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

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

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

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

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

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

  748.             YSCALE_YUV_2_RGBX_C(uint16_t)

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

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

  751.             }

  752.         }

  753.         break;

  754.     case PIX_FMT_RGB8:

  755.     case PIX_FMT_BGR8:

  756.         {

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

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

  759.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  762.             }

  763.         }

  764.         break;

  765.     case PIX_FMT_RGB4:

  766.     case PIX_FMT_BGR4:

  767.         {

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

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

  770.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  773.             }

  774.         }

  775.         break;

  776.     case PIX_FMT_RGB4_BYTE:

  777.     case PIX_FMT_BGR4_BYTE:

  778.         {

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

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

  781.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  784.             }

  785.         }

  786.         break;

  787.     case PIX_FMT_MONOBLACK:

  788.         {

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

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

  791.             int acc=0;

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

  793.                 int j;

  794.                 int Y1=1<<18;

  795.                 int Y2=1<<18;

  796. 

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

  798.                 {

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

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

  801.                 }

  802.                 Y1>>=19;

  803.                 Y2>>=19;

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

  805.                 {

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

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

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

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

  810.                 }

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

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

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

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

  815.                     dest++;

  816.                 }

  817.             }

  818.         }

  819.         break;

  820.     case PIX_FMT_YUYV422:

  821.         YSCALE_YUV_2_PACKEDX_C(void)

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

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

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

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

  826.         }

  827.         break;

  828.     case PIX_FMT_UYVY422:

  829.         YSCALE_YUV_2_PACKEDX_C(void)

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

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

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

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

  834.         }

  835.         break;

  836.     }

  837. }

  838. 

  839. 

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

  841. //Plain C versions

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

  843. #define COMPILE_C

  844. #endif

  845. 

  846. #ifdef ARCH_POWERPC

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

  848. #define COMPILE_ALTIVEC

  849. #endif //HAVE_ALTIVEC

  850. #endif //ARCH_POWERPC

  851. 

  852. #if defined(ARCH_X86)

  853. 

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

  855. #define COMPILE_MMX

  856. #endif

  857. 

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

  859. #define COMPILE_MMX2

  860. #endif

  861. 

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

  863. #define COMPILE_3DNOW

  864. #endif

  865. #endif //ARCH_X86 || ARCH_X86_64

  866. 

  867. #undef HAVE_MMX

  868. #undef HAVE_MMX2

  869. #undef HAVE_3DNOW

  870. 

  871. #ifdef COMPILE_C

  872. #undef HAVE_MMX

  873. #undef HAVE_MMX2

  874. #undef HAVE_3DNOW

  875. #undef HAVE_ALTIVEC

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

  877. #include "swscale_template.c"

  878. #endif

  879. 

  880. #ifdef ARCH_POWERPC

  881. #ifdef COMPILE_ALTIVEC

  882. #undef RENAME

  883. #define HAVE_ALTIVEC

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

  885. #include "swscale_template.c"

  886. #endif

  887. #endif //ARCH_POWERPC

  888. 

  889. #if defined(ARCH_X86)

  890. 

  891. //X86 versions

  892. /*

  893. #undef RENAME

  894. #undef HAVE_MMX

  895. #undef HAVE_MMX2

  896. #undef HAVE_3DNOW

  897. #define ARCH_X86

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

  899. #include "swscale_template.c"

  900. */

  901. //MMX versions

  902. #ifdef COMPILE_MMX

  903. #undef RENAME

  904. #define HAVE_MMX

  905. #undef HAVE_MMX2

  906. #undef HAVE_3DNOW

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

  908. #include "swscale_template.c"

  909. #endif

  910. 

  911. //MMX2 versions

  912. #ifdef COMPILE_MMX2

  913. #undef RENAME

  914. #define HAVE_MMX

  915. #define HAVE_MMX2

  916. #undef HAVE_3DNOW

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

  918. #include "swscale_template.c"

  919. #endif

  920. 

  921. //3DNOW versions

  922. #ifdef COMPILE_3DNOW

  923. #undef RENAME

  924. #define HAVE_MMX

  925. #undef HAVE_MMX2

  926. #define HAVE_3DNOW

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

  928. #include "swscale_template.c"

  929. #endif

  930. 

  931. #endif //ARCH_X86 || ARCH_X86_64

  932. 

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

  934. 

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

  936. {

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

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

  939.     else                return getSplineCoeff(        0.0,

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

  941.                                                     c + 3.0*d,

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

  943.                                             dist-1.0);

  944. }

  945. 

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

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

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

  949. {

  950.     int i;

  951.     int filterSize;

  952.     int filter2Size;

  953.     int minFilterSize;

  954.     double *filter=NULL;

  955.     double *filter2=NULL;

  956. #if defined(ARCH_X86)

  957.     if (flags & SWS_CPU_CAPS_MMX)

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

  959. #endif

  960. 

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

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

  963. 

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

  965.     {

  966.         int i;

  967.         filterSize= 1;

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

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

  970. 

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

  972.         {

  973.             filter[i*filterSize]=1;

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

  975.         }

  976. 

  977.     }

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

  979.     {

  980.         int i;

  981.         int xDstInSrc;

  982.         filterSize= 1;

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

  984. 

  985.         xDstInSrc= xInc/2 - 0x8000;

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

  987.         {

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

  989. 

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

  991.             filter[i]= 1.0;

  992.             xDstInSrc+= xInc;

  993.         }

  994.     }

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

  996.     {

  997.         int i;

  998.         int xDstInSrc;

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

  1000.         else if (flags&SWS_X      ) filterSize= 4;

  1001.         else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA

  1002.         filter= av_malloc(dstW*sizeof(double)*filterSize);

  1003. 

  1004.         xDstInSrc= xInc/2 - 0x8000;

  1005.         for (i=0; i<dstW; i++)

  1006.         {

  1007.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1008.             int j;

  1009. 

  1010.             (*filterPos)[i]= xx;

  1011.                 //Bilinear upscale / linear interpolate / Area averaging

  1012.                 for (j=0; j<filterSize; j++)

  1013.                 {

  1014.                     double d= FFABS((xx<<16) - xDstInSrc)/(double)(1<<16);

  1015.                     double coeff= 1.0 - d;

  1016.                     if (coeff<0) coeff=0;

  1017.                     filter[i*filterSize + j]= coeff;

  1018.                     xx++;

  1019.                 }

  1020.             xDstInSrc+= xInc;

  1021.         }

  1022.     }

  1023.     else

  1024.     {

  1025.         double xDstInSrc;

  1026.         double sizeFactor, filterSizeInSrc;

  1027.         const double xInc1= (double)xInc / (double)(1<<16);

  1028. 

  1029.         if      (flags&SWS_BICUBIC)      sizeFactor=  4.0;

  1030.         else if (flags&SWS_X)            sizeFactor=  8.0;

  1031.         else if (flags&SWS_AREA)         sizeFactor=  1.0; //downscale only, for upscale it is bilinear

  1032.         else if (flags&SWS_GAUSS)        sizeFactor=  8.0;   // infinite ;)

  1033.         else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? 2.0*param[0] : 6.0;

  1034.         else if (flags&SWS_SINC)         sizeFactor= 20.0; // infinite ;)

  1035.         else if (flags&SWS_SPLINE)       sizeFactor= 20.0;  // infinite ;)

  1036.         else if (flags&SWS_BILINEAR)     sizeFactor=  2.0;

  1037.         else {

  1038.             sizeFactor= 0.0; //GCC warning killer

  1039.             ASSERT(0)

  1040.         }

  1041. 

  1042.         if (xInc1 <= 1.0)       filterSizeInSrc= sizeFactor; // upscale

  1043.         else                    filterSizeInSrc= sizeFactor*srcW / (double)dstW;

  1044. 

  1045.         filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible

  1046.         if (filterSize > srcW-2) filterSize=srcW-2;

  1047. 

  1048.         filter= av_malloc(dstW*sizeof(double)*filterSize);

  1049. 

  1050.         xDstInSrc= xInc1 / 2.0 - 0.5;

  1051.         for (i=0; i<dstW; i++)

  1052.         {

  1053.             int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);

  1054.             int j;

  1055.             (*filterPos)[i]= xx;

  1056.             for (j=0; j<filterSize; j++)

  1057.             {

  1058.                 double d= FFABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;

  1059.                 double coeff;

  1060.                 if (flags & SWS_BICUBIC)

  1061.                 {

  1062.                     double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;

  1063.                     double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;

  1064. 

  1065.                     if (d<1.0)

  1066.                         coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;

  1067.                     else if (d<2.0)

  1068.                         coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;

  1069.                     else

  1070.                         coeff=0.0;

  1071.                 }

  1072. /*                else if (flags & SWS_X)

  1073.                 {

  1074.                     double p= param ? param*0.01 : 0.3;

  1075.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1076.                     coeff*= pow(2.0, - p*d*d);

  1077.                 }*/

  1078.                 else if (flags & SWS_X)

  1079.                 {

  1080.                     double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

  1081. 

  1082.                     if (d<1.0)

  1083.                         coeff = cos(d*PI);

  1084.                     else

  1085.                         coeff=-1.0;

  1086.                     if (coeff<0.0)      coeff= -pow(-coeff, A);

  1087.                     else                coeff=  pow( coeff, A);

  1088.                     coeff= coeff*0.5 + 0.5;

  1089.                 }

  1090.                 else if (flags & SWS_AREA)

  1091.                 {

  1092.                     double srcPixelSize= 1.0/xInc1;

  1093.                     if      (d + srcPixelSize/2 < 0.5) coeff= 1.0;

  1094.                     else if (d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5;

  1095.                     else coeff=0.0;

  1096.                 }

  1097.                 else if (flags & SWS_GAUSS)

  1098.                 {

  1099.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

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

  1101.                 }

  1102.                 else if (flags & SWS_SINC)

  1103.                 {

  1104.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1105.                 }

  1106.                 else if (flags & SWS_LANCZOS)

  1107.                 {

  1108.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1109.                     coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;

  1110.                     if (d>p) coeff=0;

  1111.                 }

  1112.                 else if (flags & SWS_BILINEAR)

  1113.                 {

  1114.                     coeff= 1.0 - d;

  1115.                     if (coeff<0) coeff=0;

  1116.                 }

  1117.                 else if (flags & SWS_SPLINE)

  1118.                 {

  1119.                     double p=-2.196152422706632;

  1120.                     coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);

  1121.                 }

  1122.                 else {

  1123.                     coeff= 0.0; //GCC warning killer

  1124.                     ASSERT(0)

  1125.                 }

  1126. 

  1127.                 filter[i*filterSize + j]= coeff;

  1128.                 xx++;

  1129.             }

  1130.             xDstInSrc+= xInc1;

  1131.         }

  1132.     }

  1133. 

  1134.     /* apply src & dst Filter to filter -> filter2

  1135.        av_free(filter);

  1136.     */

  1137.     ASSERT(filterSize>0)

  1138.     filter2Size= filterSize;

  1139.     if (srcFilter) filter2Size+= srcFilter->length - 1;

  1140.     if (dstFilter) filter2Size+= dstFilter->length - 1;

  1141.     ASSERT(filter2Size>0)

  1142.     filter2= av_malloc(filter2Size*dstW*sizeof(double));

  1143. 

  1144.     for (i=0; i<dstW; i++)

  1145.     {

  1146.         int j;

  1147.         SwsVector scaleFilter;

  1148.         SwsVector *outVec;

  1149. 

  1150.         scaleFilter.coeff= filter + i*filterSize;

  1151.         scaleFilter.length= filterSize;

  1152. 

  1153.         if (srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);

  1154.         else           outVec= &scaleFilter;

  1155. 

  1156.         ASSERT(outVec->length == filter2Size)

  1157.         //FIXME dstFilter

  1158. 

  1159.         for (j=0; j<outVec->length; j++)

  1160.         {

  1161.             filter2[i*filter2Size + j]= outVec->coeff[j];

  1162.         }

  1163. 

  1164.         (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

  1165. 

  1166.         if (outVec != &scaleFilter) sws_freeVec(outVec);

  1167.     }

  1168.     av_free(filter); filter=NULL;

  1169. 

  1170.     /* try to reduce the filter-size (step1 find size and shift left) */

  1171.     // Assume its near normalized (*0.5 or *2.0 is ok but * 0.001 is not)

  1172.     minFilterSize= 0;

  1173.     for (i=dstW-1; i>=0; i--)

  1174.     {

  1175.         int min= filter2Size;

  1176.         int j;

  1177.         double cutOff=0.0;

  1178. 

  1179.         /* get rid off near zero elements on the left by shifting left */

  1180.         for (j=0; j<filter2Size; j++)

  1181.         {

  1182.             int k;

  1183.             cutOff += FFABS(filter2[i*filter2Size]);

  1184. 

  1185.             if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;

  1186. 

  1187.             /* preserve Monotonicity because the core can't handle the filter otherwise */

  1188.             if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

  1189. 

  1190.             // Move filter coeffs left

  1191.             for (k=1; k<filter2Size; k++)

  1192.                 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];

  1193.             filter2[i*filter2Size + k - 1]= 0.0;

  1194.             (*filterPos)[i]++;

  1195.         }

  1196. 

  1197.         cutOff=0.0;

  1198.         /* count near zeros on the right */

  1199.         for (j=filter2Size-1; j>0; j--)

  1200.         {

  1201.             cutOff += FFABS(filter2[i*filter2Size + j]);

  1202. 

  1203.             if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;

  1204.             min--;

  1205.         }

  1206. 

  1207.         if (min>minFilterSize) minFilterSize= min;

  1208.     }

  1209. 

  1210.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1211.         // we can handle the special case 4,

  1212.         // so we don't want to go to the full 8

  1213.         if (minFilterSize < 5)

  1214.             filterAlign = 4;

  1215. 

  1216.         // we really don't want to waste our time

  1217.         // doing useless computation, so fall-back on

  1218.         // the scalar C code for very small filter.

  1219.         // vectorizing is worth it only if you have

  1220.         // decent-sized vector.

  1221.         if (minFilterSize < 3)

  1222.             filterAlign = 1;

  1223.     }

  1224. 

  1225.     if (flags & SWS_CPU_CAPS_MMX) {

  1226.         // special case for unscaled vertical filtering

  1227.         if (minFilterSize == 1 && filterAlign == 2)

  1228.             filterAlign= 1;

  1229.     }

  1230. 

  1231.     ASSERT(minFilterSize > 0)

  1232.     filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));

  1233.     ASSERT(filterSize > 0)

  1234.     filter= av_malloc(filterSize*dstW*sizeof(double));

  1235.     if (filterSize >= MAX_FILTER_SIZE)

  1236.         return -1;

  1237.     *outFilterSize= filterSize;

  1238. 

  1239.     if (flags&SWS_PRINT_INFO)

  1240.         av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);

  1241.     /* try to reduce the filter-size (step2 reduce it) */

  1242.     for (i=0; i<dstW; i++)

  1243.     {

  1244.         int j;

  1245. 

  1246.         for (j=0; j<filterSize; j++)

  1247.         {

  1248.             if (j>=filter2Size) filter[i*filterSize + j]= 0.0;

  1249.             else               filter[i*filterSize + j]= filter2[i*filter2Size + j];

  1250.         }

  1251.     }

  1252.     av_free(filter2); filter2=NULL;

  1253. 

  1254. 

  1255.     //FIXME try to align filterpos if possible

  1256. 

  1257.     //fix borders

  1258.     for (i=0; i<dstW; i++)

  1259.     {

  1260.         int j;

  1261.         if ((*filterPos)[i] < 0)

  1262.         {

  1263.             // Move filter coeffs left to compensate for filterPos

  1264.             for (j=1; j<filterSize; j++)

  1265.             {

  1266.                 int left= FFMAX(j + (*filterPos)[i], 0);

  1267.                 filter[i*filterSize + left] += filter[i*filterSize + j];

  1268.                 filter[i*filterSize + j]=0;

  1269.             }

  1270.             (*filterPos)[i]= 0;

  1271.         }

  1272. 

  1273.         if ((*filterPos)[i] + filterSize > srcW)

  1274.         {

  1275.             int shift= (*filterPos)[i] + filterSize - srcW;

  1276.             // Move filter coeffs right to compensate for filterPos

  1277.             for (j=filterSize-2; j>=0; j--)

  1278.             {

  1279.                 int right= FFMIN(j + shift, filterSize-1);

  1280.                 filter[i*filterSize +right] += filter[i*filterSize +j];

  1281.                 filter[i*filterSize +j]=0;

  1282.             }

  1283.             (*filterPos)[i]= srcW - filterSize;

  1284.         }

  1285.     }

  1286. 

  1287.     // Note the +1 is for the MMXscaler which reads over the end

  1288.     /* align at 16 for AltiVec (needed by hScale_altivec_real) */

  1289.     *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));

  1290. 

  1291.     /* Normalize & Store in outFilter */

  1292.     for (i=0; i<dstW; i++)

  1293.     {

  1294.         int j;

  1295.         double error=0;

  1296.         double sum=0;

  1297.         double scale= one;

  1298. 

  1299.         for (j=0; j<filterSize; j++)

  1300.         {

  1301.             sum+= filter[i*filterSize + j];

  1302.         }

  1303.         scale/= sum;

  1304.         for (j=0; j<*outFilterSize; j++)

  1305.         {

  1306.             double v= filter[i*filterSize + j]*scale + error;

  1307.             int intV= floor(v + 0.5);

  1308.             (*outFilter)[i*(*outFilterSize) + j]= intV;

  1309.             error = v - intV;

  1310.         }

  1311.     }

  1312. 

  1313.     (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end

  1314.     for (i=0; i<*outFilterSize; i++)

  1315.     {

  1316.         int j= dstW*(*outFilterSize);

  1317.         (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

  1318.     }

  1319. 

  1320.     av_free(filter);

  1321.     return 0;

  1322. }

  1323. 

  1324. #ifdef COMPILE_MMX2

  1325. static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)

  1326. {

  1327.     uint8_t *fragmentA;

  1328.     long imm8OfPShufW1A;

  1329.     long imm8OfPShufW2A;

  1330.     long fragmentLengthA;

  1331.     uint8_t *fragmentB;

  1332.     long imm8OfPShufW1B;

  1333.     long imm8OfPShufW2B;

  1334.     long fragmentLengthB;

  1335.     int fragmentPos;

  1336. 

  1337.     int xpos, i;

  1338. 

  1339.     // create an optimized horizontal scaling routine

  1340. 

  1341.     //code fragment

  1342. 

  1343.     asm volatile(

  1344.         "jmp                         9f                 \n\t"

  1345.     // Begin

  1346.         "0:                                             \n\t"

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

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

  1349.         "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"

  1350.         "punpcklbw                %%mm7, %%mm1          \n\t"

  1351.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1352.         "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"

  1353.         "1:                                             \n\t"

  1354.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1355.         "2:                                             \n\t"

  1356.         "psubw                    %%mm1, %%mm0          \n\t"

  1357.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1358.         "pmullw                   %%mm3, %%mm0          \n\t"

  1359.         "psllw                       $7, %%mm1          \n\t"

  1360.         "paddw                    %%mm1, %%mm0          \n\t"

  1361. 

  1362.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1363. 

  1364.         "add                         $8, %%"REG_a"      \n\t"

  1365.     // End

  1366.         "9:                                             \n\t"

  1367. //        "int $3                                         \n\t"

  1368.         "lea                         0b, %0             \n\t"

  1369.         "lea                         1b, %1             \n\t"

  1370.         "lea                         2b, %2             \n\t"

  1371.         "dec                         %1                 \n\t"

  1372.         "dec                         %2                 \n\t"

  1373.         "sub                         %0, %1             \n\t"

  1374.         "sub                         %0, %2             \n\t"

  1375.         "lea                         9b, %3             \n\t"

  1376.         "sub                         %0, %3             \n\t"

  1377. 

  1378. 

  1379.         :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),

  1380.         "=r" (fragmentLengthA)

  1381.     );

  1382. 

  1383.     asm volatile(

  1384.         "jmp                         9f                 \n\t"

  1385.     // Begin

  1386.         "0:                                             \n\t"

  1387.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1388.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1389.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1390.         "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"

  1391.         "1:                                             \n\t"

  1392.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1393.         "2:                                             \n\t"

  1394.         "psubw                    %%mm1, %%mm0          \n\t"

  1395.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1396.         "pmullw                   %%mm3, %%mm0          \n\t"

  1397.         "psllw                       $7, %%mm1          \n\t"

  1398.         "paddw                    %%mm1, %%mm0          \n\t"

  1399. 

  1400.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1401. 

  1402.         "add                         $8, %%"REG_a"      \n\t"

  1403.     // End

  1404.         "9:                                             \n\t"

  1405. //        "int                       $3                   \n\t"

  1406.         "lea                         0b, %0             \n\t"

  1407.         "lea                         1b, %1             \n\t"

  1408.         "lea                         2b, %2             \n\t"

  1409.         "dec                         %1                 \n\t"

  1410.         "dec                         %2                 \n\t"

  1411.         "sub                         %0, %1             \n\t"

  1412.         "sub                         %0, %2             \n\t"

  1413.         "lea                         9b, %3             \n\t"

  1414.         "sub                         %0, %3             \n\t"

  1415. 

  1416. 

  1417.         :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),

  1418.         "=r" (fragmentLengthB)

  1419.     );

  1420. 

  1421.     xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers

  1422.     fragmentPos=0;

  1423. 

  1424.     for (i=0; i<dstW/numSplits; i++)

  1425.     {

  1426.         int xx=xpos>>16;

  1427. 

  1428.         if ((i&3) == 0)

  1429.         {

  1430.             int a=0;

  1431.             int b=((xpos+xInc)>>16) - xx;

  1432.             int c=((xpos+xInc*2)>>16) - xx;

  1433.             int d=((xpos+xInc*3)>>16) - xx;

  1434. 

  1435.             filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;

  1436.             filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;

  1437.             filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;

  1438.             filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;

  1439.             filterPos[i/2]= xx;

  1440. 

  1441.             if (d+1<4)

  1442.             {

  1443.                 int maxShift= 3-(d+1);

  1444.                 int shift=0;

  1445. 

  1446.                 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);

  1447. 

  1448.                 funnyCode[fragmentPos + imm8OfPShufW1B]=

  1449.                     (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);

  1450.                 funnyCode[fragmentPos + imm8OfPShufW2B]=

  1451.                     a | (b<<2) | (c<<4) | (d<<6);

  1452. 

  1453.                 if (i+3>=dstW) shift=maxShift; //avoid overread

  1454.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

  1455. 

  1456.                 if (shift && i>=shift)

  1457.                 {

  1458.                     funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;

  1459.                     funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;

  1460.                     filterPos[i/2]-=shift;

  1461.                 }

  1462. 

  1463.                 fragmentPos+= fragmentLengthB;

  1464.             }

  1465.             else

  1466.             {

  1467.                 int maxShift= 3-d;

  1468.                 int shift=0;

  1469. 

  1470.                 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);

  1471. 

  1472.                 funnyCode[fragmentPos + imm8OfPShufW1A]=

  1473.                 funnyCode[fragmentPos + imm8OfPShufW2A]=

  1474.                     a | (b<<2) | (c<<4) | (d<<6);

  1475. 

  1476.                 if (i+4>=dstW) shift=maxShift; //avoid overread

  1477.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align

  1478. 

  1479.                 if (shift && i>=shift)

  1480.                 {

  1481.                     funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;

  1482.                     funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;

  1483.                     filterPos[i/2]-=shift;

  1484.                 }

  1485. 

  1486.                 fragmentPos+= fragmentLengthA;

  1487.             }

  1488. 

  1489.             funnyCode[fragmentPos]= RET;

  1490.         }

  1491.         xpos+=xInc;

  1492.     }

  1493.     filterPos[i/2]= xpos>>16; // needed to jump to the next part

  1494. }

  1495. #endif /* COMPILE_MMX2 */

  1496. 

  1497. static void globalInit(void){

  1498.     // generating tables:

  1499.     int i;

  1500.     for (i=0; i<768; i++){

  1501.         int c= av_clip_uint8(i-256);

  1502.         clip_table[i]=c;

  1503.     }

  1504. }

  1505. 

  1506. static SwsFunc getSwsFunc(int flags){

  1507. 

  1508. #if defined(RUNTIME_CPUDETECT) && defined (CONFIG_GPL)

  1509. #if defined(ARCH_X86)

  1510.     // ordered per speed fasterst first

  1511.     if (flags & SWS_CPU_CAPS_MMX2)

  1512.         return swScale_MMX2;

  1513.     else if (flags & SWS_CPU_CAPS_3DNOW)

  1514.         return swScale_3DNow;

  1515.     else if (flags & SWS_CPU_CAPS_MMX)

  1516.         return swScale_MMX;

  1517.     else

  1518.         return swScale_C;

  1519. 

  1520. #else

  1521. #ifdef ARCH_POWERPC

  1522.     if (flags & SWS_CPU_CAPS_ALTIVEC)

  1523.         return swScale_altivec;

  1524.     else

  1525.         return swScale_C;

  1526. #endif

  1527.     return swScale_C;

  1528. #endif /* defined(ARCH_X86) */

  1529. #else //RUNTIME_CPUDETECT

  1530. #ifdef HAVE_MMX2

  1531.     return swScale_MMX2;

  1532. #elif defined (HAVE_3DNOW)

  1533.     return swScale_3DNow;

  1534. #elif defined (HAVE_MMX)

  1535.     return swScale_MMX;

  1536. #elif defined (HAVE_ALTIVEC)

  1537.     return swScale_altivec;

  1538. #else

  1539.     return swScale_C;

  1540. #endif

  1541. #endif //!RUNTIME_CPUDETECT

  1542. }

  1543. 

  1544. static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1545.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1546.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1547.     /* Copy Y plane */

  1548.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1549.         memcpy(dst, src[0], srcSliceH*dstStride[0]);

  1550.     else

  1551.     {

  1552.         int i;

  1553.         uint8_t *srcPtr= src[0];

  1554.         uint8_t *dstPtr= dst;

  1555.         for (i=0; i<srcSliceH; i++)

  1556.         {

  1557.             memcpy(dstPtr, srcPtr, c->srcW);

  1558.             srcPtr+= srcStride[0];

  1559.             dstPtr+= dstStride[0];

  1560.         }

  1561.     }

  1562.     dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1563.     if (c->dstFormat == PIX_FMT_NV12)

  1564.         interleaveBytes( src[1],src[2],dst,c->srcW/2,srcSliceH/2,srcStride[1],srcStride[2],dstStride[0] );

  1565.     else

  1566.         interleaveBytes( src[2],src[1],dst,c->srcW/2,srcSliceH/2,srcStride[2],srcStride[1],dstStride[0] );

  1567. 

  1568.     return srcSliceH;

  1569. }

  1570. 

  1571. static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1572.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1573.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1574. 

  1575.     yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );

  1576. 

  1577.     return srcSliceH;

  1578. }

  1579. 

  1580. static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1581.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1582.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1583. 

  1584.     yv12touyvy( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );

  1585. 

  1586.     return srcSliceH;

  1587. }

  1588. 

  1589. /* {RGB,BGR}{15,16,24,32} -> {RGB,BGR}{15,16,24,32} */

  1590. static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1591.                           int srcSliceH, uint8_t* dst[], int dstStride[]){

  1592.     const int srcFormat= c->srcFormat;

  1593.     const int dstFormat= c->dstFormat;

  1594.     const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;

  1595.     const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;

  1596.     const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1597.     const int dstId= fmt_depth(dstFormat) >> 2;

  1598.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  1599. 

  1600.     /* BGR -> BGR */

  1601.     if (  (isBGR(srcFormat) && isBGR(dstFormat))

  1602.        || (isRGB(srcFormat) && isRGB(dstFormat))){

  1603.         switch(srcId | (dstId<<4)){

  1604.         case 0x34: conv= rgb16to15; break;

  1605.         case 0x36: conv= rgb24to15; break;

  1606.         case 0x38: conv= rgb32to15; break;

  1607.         case 0x43: conv= rgb15to16; break;

  1608.         case 0x46: conv= rgb24to16; break;

  1609.         case 0x48: conv= rgb32to16; break;

  1610.         case 0x63: conv= rgb15to24; break;

  1611.         case 0x64: conv= rgb16to24; break;

  1612.         case 0x68: conv= rgb32to24; break;

  1613.         case 0x83: conv= rgb15to32; break;

  1614.         case 0x84: conv= rgb16to32; break;

  1615.         case 0x86: conv= rgb24to32; break;

  1616.         default: av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",

  1617.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1618.         }

  1619.     }else if (  (isBGR(srcFormat) && isRGB(dstFormat))

  1620.              || (isRGB(srcFormat) && isBGR(dstFormat))){

  1621.         switch(srcId | (dstId<<4)){

  1622.         case 0x33: conv= rgb15tobgr15; break;

  1623.         case 0x34: conv= rgb16tobgr15; break;

  1624.         case 0x36: conv= rgb24tobgr15; break;

  1625.         case 0x38: conv= rgb32tobgr15; break;

  1626.         case 0x43: conv= rgb15tobgr16; break;

  1627.         case 0x44: conv= rgb16tobgr16; break;

  1628.         case 0x46: conv= rgb24tobgr16; break;

  1629.         case 0x48: conv= rgb32tobgr16; break;

  1630.         case 0x63: conv= rgb15tobgr24; break;

  1631.         case 0x64: conv= rgb16tobgr24; break;

  1632.         case 0x66: conv= rgb24tobgr24; break;

  1633.         case 0x68: conv= rgb32tobgr24; break;

  1634.         case 0x83: conv= rgb15tobgr32; break;

  1635.         case 0x84: conv= rgb16tobgr32; break;

  1636.         case 0x86: conv= rgb24tobgr32; break;

  1637.         case 0x88: conv= rgb32tobgr32; break;

  1638.         default: av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",

  1639.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1640.         }

  1641.     }else{

  1642.         av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",

  1643.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1644.     }

  1645. 

  1646.     if(conv)

  1647.     {

  1648.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp)

  1649.             conv(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  1650.         else

  1651.         {

  1652.             int i;

  1653.             uint8_t *srcPtr= src[0];

  1654.             uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1655. 

  1656.             for (i=0; i<srcSliceH; i++)

  1657.             {

  1658.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  1659.                 srcPtr+= srcStride[0];

  1660.                 dstPtr+= dstStride[0];

  1661.             }

  1662.         }

  1663.     }

  1664.     return srcSliceH;

  1665. }

  1666. 

  1667. static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

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

  1669. 

  1670.     rgb24toyv12(

  1671.         src[0],

  1672.         dst[0]+ srcSliceY    *dstStride[0],

  1673.         dst[1]+(srcSliceY>>1)*dstStride[1],

  1674.         dst[2]+(srcSliceY>>1)*dstStride[2],

  1675.         c->srcW, srcSliceH,

  1676.         dstStride[0], dstStride[1], srcStride[0]);

  1677.     return srcSliceH;

  1678. }

  1679. 

  1680. static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1681.                              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1682.     int i;

  1683. 

  1684.     /* copy Y */

  1685.     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)

  1686.         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);

  1687.     else{

  1688.         uint8_t *srcPtr= src[0];

  1689.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1690. 

  1691.         for (i=0; i<srcSliceH; i++)

  1692.         {

  1693.             memcpy(dstPtr, srcPtr, c->srcW);

  1694.             srcPtr+= srcStride[0];

  1695.             dstPtr+= dstStride[0];

  1696.         }

  1697.     }

  1698. 

  1699.     if (c->dstFormat==PIX_FMT_YUV420P){

  1700.         planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);

  1701.         planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);

  1702.     }else{

  1703.         planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);

  1704.         planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);

  1705.     }

  1706.     return srcSliceH;

  1707. }

  1708. 

  1709. /* unscaled copy like stuff (assumes nearly identical formats) */

  1710. static int simpleCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1711.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  1712. 

  1713.     if (isPacked(c->srcFormat))

  1714.     {

  1715.         if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1716.             memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  1717.         else

  1718.         {

  1719.             int i;

  1720.             uint8_t *srcPtr= src[0];

  1721.             uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1722.             int length=0;

  1723. 

  1724.             /* universal length finder */

  1725.             while(length+c->srcW <= FFABS(dstStride[0])

  1726.                && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  1727.             ASSERT(length!=0);

  1728. 

  1729.             for (i=0; i<srcSliceH; i++)

  1730.             {

  1731.                 memcpy(dstPtr, srcPtr, length);

  1732.                 srcPtr+= srcStride[0];

  1733.                 dstPtr+= dstStride[0];

  1734.             }

  1735.         }

  1736.     }

  1737.     else

  1738.     { /* Planar YUV or gray */

  1739.         int plane;

  1740.         for (plane=0; plane<3; plane++)

  1741.         {

  1742.             int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  1743.             int y=      plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  1744.             int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  1745. 

  1746.             if ((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)

  1747.             {

  1748.                 if (!isGray(c->dstFormat))

  1749.                     memset(dst[plane], 128, dstStride[plane]*height);

  1750.             }

  1751.             else

  1752.             {

  1753.                 if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)

  1754.                     memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);

  1755.                 else

  1756.                 {

  1757.                     int i;

  1758.                     uint8_t *srcPtr= src[plane];

  1759.                     uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  1760.                     for (i=0; i<height; i++)

  1761.                     {

  1762.                         memcpy(dstPtr, srcPtr, length);

  1763.                         srcPtr+= srcStride[plane];

  1764.                         dstPtr+= dstStride[plane];

  1765.                     }

  1766.                 }

  1767.             }

  1768.         }

  1769.     }

  1770.     return srcSliceH;

  1771. }

  1772. 

  1773. static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1774.                         int srcSliceH, uint8_t* dst[], int dstStride[]){

  1775. 

  1776.     int length= c->srcW;

  1777.     int y=      srcSliceY;

  1778.     int height= srcSliceH;

  1779.     int i, j;

  1780.     uint8_t *srcPtr= src[0];

  1781.     uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  1782. 

  1783.     if (!isGray(c->dstFormat)){

  1784.         int height= -((-srcSliceH)>>c->chrDstVSubSample);

  1785.         memset(dst[1], 128, dstStride[1]*height);

  1786.         memset(dst[2], 128, dstStride[2]*height);

  1787.     }

  1788.     if (c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;

  1789.     for (i=0; i<height; i++)

  1790.     {

  1791.         for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  1792.         srcPtr+= srcStride[0];

  1793.         dstPtr+= dstStride[0];

  1794.     }

  1795.     return srcSliceH;

  1796. }

  1797. 

  1798. static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1799.                         int srcSliceH, uint8_t* dst[], int dstStride[]){

  1800. 

  1801.     int length= c->srcW;

  1802.     int y=      srcSliceY;

  1803.     int height= srcSliceH;

  1804.     int i, j;

  1805.     uint8_t *srcPtr= src[0];

  1806.     uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  1807.     for (i=0; i<height; i++)

  1808.     {

  1809.         for (j=0; j<length; j++)

  1810.         {

  1811.             dstPtr[j<<1] = srcPtr[j];

  1812.             dstPtr[(j<<1)+1] = srcPtr[j];

  1813.         }

  1814.         srcPtr+= srcStride[0];

  1815.         dstPtr+= dstStride[0];

  1816.     }

  1817.     return srcSliceH;

  1818. }

  1819. 

  1820. static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1821.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  1822. 

  1823.     int length= c->srcW;

  1824.     int y=      srcSliceY;

  1825.     int height= srcSliceH;

  1826.     int i, j;

  1827.     uint16_t *srcPtr= src[0];

  1828.     uint16_t *dstPtr= dst[0] + dstStride[0]*y/2;

  1829.     for (i=0; i<height; i++)

  1830.     {

  1831.         for (j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);

  1832.         srcPtr+= srcStride[0]/2;

  1833.         dstPtr+= dstStride[0]/2;

  1834.     }

  1835.     return srcSliceH;

  1836. }

  1837. 

  1838. 

  1839. static void getSubSampleFactors(int *h, int *v, int format){

  1840.     switch(format){

  1841.     case PIX_FMT_UYVY422:

  1842.     case PIX_FMT_YUYV422:

  1843.         *h=1;

  1844.         *v=0;

  1845.         break;

  1846.     case PIX_FMT_YUV420P:

  1847.     case PIX_FMT_GRAY16BE:

  1848.     case PIX_FMT_GRAY16LE:

  1849.     case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented

  1850.     case PIX_FMT_NV12:

  1851.     case PIX_FMT_NV21:

  1852.         *h=1;

  1853.         *v=1;

  1854.         break;

  1855.     case PIX_FMT_YUV410P:

  1856.         *h=2;

  1857.         *v=2;

  1858.         break;

  1859.     case PIX_FMT_YUV444P:

  1860.         *h=0;

  1861.         *v=0;

  1862.         break;

  1863.     case PIX_FMT_YUV422P:

  1864.         *h=1;

  1865.         *v=0;

  1866.         break;

  1867.     case PIX_FMT_YUV411P:

  1868.         *h=2;

  1869.         *v=0;

  1870.         break;

  1871.     default:

  1872.         *h=0;

  1873.         *v=0;

  1874.         break;

  1875.     }

  1876. }

  1877. 

  1878. static uint16_t roundToInt16(int64_t f){

  1879.     int r= (f + (1<<15))>>16;

  1880.          if (r<-0x7FFF) return 0x8000;

  1881.     else if (r> 0x7FFF) return 0x7FFF;

  1882.     else                return r;

  1883. }

  1884. 

  1885. /**

  1886.  * @param inv_table the yuv2rgb coeffs, normally Inverse_Table_6_9[x]

  1887.  * @param fullRange if 1 then the luma range is 0..255 if 0 its 16..235

  1888.  * @return -1 if not supported

  1889.  */

  1890. int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){

  1891.     int64_t crv =  inv_table[0];

  1892.     int64_t cbu =  inv_table[1];

  1893.     int64_t cgu = -inv_table[2];

  1894.     int64_t cgv = -inv_table[3];

  1895.     int64_t cy  = 1<<16;

  1896.     int64_t oy  = 0;

  1897. 

  1898.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  1899.     memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);

  1900.     memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);

  1901. 

  1902.     c->brightness= brightness;

  1903.     c->contrast  = contrast;

  1904.     c->saturation= saturation;

  1905.     c->srcRange  = srcRange;

  1906.     c->dstRange  = dstRange;

  1907. 

  1908.     c->uOffset=   0x0400040004000400LL;

  1909.     c->vOffset=   0x0400040004000400LL;

  1910. 

  1911.     if (!srcRange){

  1912.         cy= (cy*255) / 219;

  1913.         oy= 16<<16;

  1914.     }else{

  1915.         crv= (crv*224) / 255;

  1916.         cbu= (cbu*224) / 255;

  1917.         cgu= (cgu*224) / 255;

  1918.         cgv= (cgv*224) / 255;

  1919.     }

  1920. 

  1921.     cy = (cy *contrast             )>>16;

  1922.     crv= (crv*contrast * saturation)>>32;

  1923.     cbu= (cbu*contrast * saturation)>>32;

  1924.     cgu= (cgu*contrast * saturation)>>32;

  1925.     cgv= (cgv*contrast * saturation)>>32;

  1926. 

  1927.     oy -= 256*brightness;

  1928. 

  1929.     c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;

  1930.     c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;

  1931.     c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;

  1932.     c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;

  1933.     c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;

  1934.     c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;

  1935. 

  1936.     yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);

  1937.     //FIXME factorize

  1938. 

  1939. #ifdef COMPILE_ALTIVEC

  1940.     if (c->flags & SWS_CPU_CAPS_ALTIVEC)

  1941.         yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);

  1942. #endif

  1943.     return 0;

  1944. }

  1945. 

  1946. /**

  1947.  * @return -1 if not supported

  1948.  */

  1949. int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){

  1950.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  1951. 

  1952.     *inv_table = c->srcColorspaceTable;

  1953.     *table     = c->dstColorspaceTable;

  1954.     *srcRange  = c->srcRange;

  1955.     *dstRange  = c->dstRange;

  1956.     *brightness= c->brightness;

  1957.     *contrast  = c->contrast;

  1958.     *saturation= c->saturation;

  1959. 

  1960.     return 0;

  1961. }

  1962. 

  1963. static int handle_jpeg(int *format)

  1964. {

  1965.     switch (*format) {

  1966.         case PIX_FMT_YUVJ420P:

  1967.             *format = PIX_FMT_YUV420P;

  1968.             return 1;

  1969.         case PIX_FMT_YUVJ422P:

  1970.             *format = PIX_FMT_YUV422P;

  1971.             return 1;

  1972.         case PIX_FMT_YUVJ444P:

  1973.             *format = PIX_FMT_YUV444P;

  1974.             return 1;

  1975.         default:

  1976.             return 0;

  1977.     }

  1978. }

  1979. 

  1980. SwsContext *sws_getContext(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat, int flags,

  1981.                            SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){

  1982. 

  1983.     SwsContext *c;

  1984.     int i;

  1985.     int usesVFilter, usesHFilter;

  1986.     int unscaled, needsDither;

  1987.     int srcRange, dstRange;

  1988.     SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};

  1989. #if defined(ARCH_X86)

  1990.     if (flags & SWS_CPU_CAPS_MMX)

  1991.         asm volatile("emms\n\t"::: "memory");

  1992. #endif

  1993. 

  1994. #if !defined(RUNTIME_CPUDETECT) || !defined (CONFIG_GPL) //ensure that the flags match the compiled variant if cpudetect is off

  1995.     flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC);

  1996. #ifdef HAVE_MMX2

  1997.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  1998. #elif defined (HAVE_3DNOW)

  1999.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  2000. #elif defined (HAVE_MMX)

  2001.     flags |= SWS_CPU_CAPS_MMX;

  2002. #elif defined (HAVE_ALTIVEC)

  2003.     flags |= SWS_CPU_CAPS_ALTIVEC;

  2004. #endif

  2005. #endif /* RUNTIME_CPUDETECT */

  2006.     if (clip_table[512] != 255) globalInit();

  2007.     if (rgb15to16 == NULL) sws_rgb2rgb_init(flags);

  2008. 

  2009.     unscaled = (srcW == dstW && srcH == dstH);

  2010.     needsDither= (isBGR(dstFormat) || isRGB(dstFormat))

  2011.         && (fmt_depth(dstFormat))<24

  2012.         && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

  2013. 

  2014.     srcRange = handle_jpeg(&srcFormat);

  2015.     dstRange = handle_jpeg(&dstFormat);

  2016. 

  2017.     if (!isSupportedIn(srcFormat))

  2018.     {

  2019.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input format\n", sws_format_name(srcFormat));

  2020.         return NULL;

  2021.     }

  2022.     if (!isSupportedOut(dstFormat))

  2023.     {

  2024.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output format\n", sws_format_name(dstFormat));

  2025.         return NULL;

  2026.     }

  2027. 

  2028.     /* sanity check */

  2029.     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

  2030.     {

  2031.         av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",

  2032.                srcW, srcH, dstW, dstH);

  2033.         return NULL;

  2034.     }

  2035. 

  2036.     if (!dstFilter) dstFilter= &dummyFilter;

  2037.     if (!srcFilter) srcFilter= &dummyFilter;

  2038. 

  2039.     c= av_mallocz(sizeof(SwsContext));

  2040. 

  2041.     c->av_class = &sws_context_class;

  2042.     c->srcW= srcW;

  2043.     c->srcH= srcH;

  2044.     c->dstW= dstW;

  2045.     c->dstH= dstH;

  2046.     c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;

  2047.     c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;

  2048.     c->flags= flags;

  2049.     c->dstFormat= dstFormat;

  2050.     c->srcFormat= srcFormat;

  2051.     c->vRounder= 4* 0x0001000100010001ULL;

  2052. 

  2053.     usesHFilter= usesVFilter= 0;

  2054.     if (dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesVFilter=1;

  2055.     if (dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesHFilter=1;

  2056.     if (dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesVFilter=1;

  2057.     if (dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesHFilter=1;

  2058.     if (srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesVFilter=1;

  2059.     if (srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesHFilter=1;

  2060.     if (srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesVFilter=1;

  2061.     if (srcFilter->chrH!=NULL && srcFilter->chrH->length>1) usesHFilter=1;

  2062. 

  2063.     getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);

  2064.     getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

  2065. 

  2066.     // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation

  2067.     if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

  2068. 

  2069.     // drop some chroma lines if the user wants it

  2070.     c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;

  2071.     c->chrSrcVSubSample+= c->vChrDrop;

  2072. 

  2073.     // drop every 2. pixel for chroma calculation unless user wants full chroma

  2074.     if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)

  2075.       && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8

  2076.       && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4

  2077.       && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE)

  2078.         c->chrSrcHSubSample=1;

  2079. 

  2080.     if (param){

  2081.         c->param[0] = param[0];

  2082.         c->param[1] = param[1];

  2083.     }else{

  2084.         c->param[0] =

  2085.         c->param[1] = SWS_PARAM_DEFAULT;

  2086.     }

  2087. 

  2088.     c->chrIntHSubSample= c->chrDstHSubSample;

  2089.     c->chrIntVSubSample= c->chrSrcVSubSample;

  2090. 

  2091.     // Note the -((-x)>>y) is so that we always round toward +inf.

  2092.     c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);

  2093.     c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);

  2094.     c->chrDstW= -((-dstW) >> c->chrDstHSubSample);

  2095.     c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

  2096. 

  2097.     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);

  2098. 

  2099.     /* unscaled special Cases */

  2100.     if (unscaled && !usesHFilter && !usesVFilter)

  2101.     {

  2102.         /* yv12_to_nv12 */

  2103.         if (srcFormat == PIX_FMT_YUV420P && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))

  2104.         {

  2105.             c->swScale= PlanarToNV12Wrapper;

  2106.         }

  2107. #ifdef CONFIG_GPL

  2108.         /* yuv2bgr */

  2109.         if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat)))

  2110.         {

  2111.             c->swScale= yuv2rgb_get_func_ptr(c);

  2112.         }

  2113. #endif

  2114. 

  2115.         if ( srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P )

  2116.         {

  2117.             c->swScale= yvu9toyv12Wrapper;

  2118.         }

  2119. 

  2120.         /* bgr24toYV12 */

  2121.         if (srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P)

  2122.             c->swScale= bgr24toyv12Wrapper;

  2123. 

  2124.         /* rgb/bgr -> rgb/bgr (no dither needed forms) */

  2125.         if (  (isBGR(srcFormat) || isRGB(srcFormat))

  2126.            && (isBGR(dstFormat) || isRGB(dstFormat))

  2127.            && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  2128.            && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  2129.            && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  2130.            && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  2131.            && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  2132.            && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  2133.            && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  2134.            && !needsDither)

  2135.              c->swScale= rgb2rgbWrapper;

  2136. 

  2137.         /* LQ converters if -sws 0 or -sws 4*/

  2138.         if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){

  2139.             /* rgb/bgr -> rgb/bgr (dither needed forms) */

  2140.             if ( (isBGR(srcFormat) || isRGB(srcFormat))

  2141.               && (isBGR(dstFormat) || isRGB(dstFormat))

  2142.               && needsDither)

  2143.                 c->swScale= rgb2rgbWrapper;

  2144. 

  2145.             /* yv12_to_yuy2 */

  2146.             if (srcFormat == PIX_FMT_YUV420P &&

  2147.                 (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422))

  2148.             {

  2149.                 if (dstFormat == PIX_FMT_YUYV422)

  2150.                     c->swScale= PlanarToYuy2Wrapper;

  2151.                 else

  2152.                     c->swScale= PlanarToUyvyWrapper;

  2153.             }

  2154.         }

  2155. 

  2156. #ifdef COMPILE_ALTIVEC

  2157.         if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  2158.             ((srcFormat == PIX_FMT_YUV420P &&

  2159.              (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422)))) {

  2160.           // unscaled YV12 -> packed YUV, we want speed

  2161.           if (dstFormat == PIX_FMT_YUYV422)

  2162.               c->swScale= yv12toyuy2_unscaled_altivec;

  2163.           else

  2164.               c->swScale= yv12touyvy_unscaled_altivec;

  2165.         }

  2166. #endif

  2167. 

  2168.         /* simple copy */

  2169.         if (  srcFormat == dstFormat

  2170.             || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  2171.             || (isPlanarYUV(dstFormat) && isGray(srcFormat)) )

  2172.         {

  2173.             c->swScale= simpleCopy;

  2174.         }

  2175. 

  2176.         /* gray16{le,be} conversions */

  2177.         if (isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))

  2178.         {

  2179.             c->swScale= gray16togray;

  2180.         }

  2181.         if ((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))

  2182.         {

  2183.             c->swScale= graytogray16;

  2184.         }

  2185.         if (srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))

  2186.         {

  2187.             c->swScale= gray16swap;

  2188.         }

  2189. 

  2190.         if (c->swScale){

  2191.             if (flags&SWS_PRINT_INFO)

  2192.                 av_log(c, AV_LOG_INFO, "SwScaler: using unscaled %s -> %s special converter\n",

  2193.                                 sws_format_name(srcFormat), sws_format_name(dstFormat));

  2194.             return c;

  2195.         }

  2196.     }

  2197. 

  2198.     if (flags & SWS_CPU_CAPS_MMX2)

  2199.     {

  2200.         c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;

  2201.         if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))

  2202.         {

  2203.             if (flags&SWS_PRINT_INFO)

  2204.                 av_log(c, AV_LOG_INFO, "SwScaler: output Width is not a multiple of 32 -> no MMX2 scaler\n");

  2205.         }

  2206.         if (usesHFilter) c->canMMX2BeUsed=0;

  2207.     }

  2208.     else

  2209.         c->canMMX2BeUsed=0;

  2210. 

  2211.     c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;

  2212.     c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

  2213. 

  2214.     // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst

  2215.     // but only for the FAST_BILINEAR mode otherwise do correct scaling

  2216.     // n-2 is the last chrominance sample available

  2217.     // this is not perfect, but noone shuld notice the difference, the more correct variant

  2218.     // would be like the vertical one, but that would require some special code for the

  2219.     // first and last pixel

  2220.     if (flags&SWS_FAST_BILINEAR)

  2221.     {

  2222.         if (c->canMMX2BeUsed)

  2223.         {

  2224.             c->lumXInc+= 20;

  2225.             c->chrXInc+= 20;

  2226.         }

  2227.         //we don't use the x86asm scaler if mmx is available

  2228.         else if (flags & SWS_CPU_CAPS_MMX)

  2229.         {

  2230.             c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;

  2231.             c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;

  2232.         }

  2233.     }

  2234. 

  2235.     /* precalculate horizontal scaler filter coefficients */

  2236.     {

  2237.         const int filterAlign=

  2238.             (flags & SWS_CPU_CAPS_MMX) ? 4 :

  2239.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2240.             1;

  2241. 

  2242.         initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,

  2243.                    srcW      ,       dstW, filterAlign, 1<<14,

  2244.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2245.                    srcFilter->lumH, dstFilter->lumH, c->param);

  2246.         initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,

  2247.                    c->chrSrcW, c->chrDstW, filterAlign, 1<<14,

  2248.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2249.                    srcFilter->chrH, dstFilter->chrH, c->param);

  2250. 

  2251. #define MAX_FUNNY_CODE_SIZE 10000

  2252. #if defined(COMPILE_MMX2)

  2253. // can't downscale !!!

  2254.         if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))

  2255.         {

  2256. #ifdef MAP_ANONYMOUS

  2257.             c->funnyYCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2258.             c->funnyUVCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2259. #else

  2260.             c->funnyYCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2261.             c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2262. #endif

  2263. 

  2264.             c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));

  2265.             c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));

  2266.             c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));

  2267.             c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));

  2268. 

  2269.             initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);

  2270.             initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);

  2271.         }

  2272. #endif /* defined(COMPILE_MMX2) */

  2273.     } // Init Horizontal stuff

  2274. 

  2275. 

  2276. 

  2277.     /* precalculate vertical scaler filter coefficients */

  2278.     {

  2279.         const int filterAlign=

  2280.             (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :

  2281.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2282.             1;

  2283. 

  2284.         initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,

  2285.                    srcH      ,        dstH, filterAlign, (1<<12)-4,

  2286.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2287.                    srcFilter->lumV, dstFilter->lumV, c->param);

  2288.         initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,

  2289.                    c->chrSrcH, c->chrDstH, filterAlign, (1<<12)-4,

  2290.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2291.                    srcFilter->chrV, dstFilter->chrV, c->param);

  2292. 

  2293. #ifdef HAVE_ALTIVEC

  2294.         c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);

  2295.         c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);

  2296. 

  2297.         for (i=0;i<c->vLumFilterSize*c->dstH;i++) {

  2298.             int j;

  2299.             short *p = (short *)&c->vYCoeffsBank[i];

  2300.             for (j=0;j<8;j++)

  2301.                 p[j] = c->vLumFilter[i];

  2302.         }

  2303. 

  2304.         for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {

  2305.             int j;

  2306.             short *p = (short *)&c->vCCoeffsBank[i];

  2307.             for (j=0;j<8;j++)

  2308.                 p[j] = c->vChrFilter[i];

  2309.         }

  2310. #endif

  2311.     }

  2312. 

  2313.     // Calculate Buffer Sizes so that they won't run out while handling these damn slices

  2314.     c->vLumBufSize= c->vLumFilterSize;

  2315.     c->vChrBufSize= c->vChrFilterSize;

  2316.     for (i=0; i<dstH; i++)

  2317.     {

  2318.         int chrI= i*c->chrDstH / dstH;

  2319.         int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,

  2320.                            ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

  2321. 

  2322.         nextSlice>>= c->chrSrcVSubSample;

  2323.         nextSlice<<= c->chrSrcVSubSample;

  2324.         if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)

  2325.             c->vLumBufSize= nextSlice - c->vLumFilterPos[i   ];

  2326.         if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))

  2327.             c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];

  2328.     }

  2329. 

  2330.     // allocate pixbufs (we use dynamic allocation because otherwise we would need to

  2331.     c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));

  2332.     c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));

  2333.     //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)

  2334.     /* align at 16 bytes for AltiVec */

  2335.     for (i=0; i<c->vLumBufSize; i++)

  2336.         c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(4000);

  2337.     for (i=0; i<c->vChrBufSize; i++)

  2338.         c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc(8000);

  2339. 

  2340.     //try to avoid drawing green stuff between the right end and the stride end

  2341.     for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, 8000);

  2342. 

  2343.     ASSERT(c->chrDstH <= dstH)

  2344. 

  2345.     if (flags&SWS_PRINT_INFO)

  2346.     {

  2347. #ifdef DITHER1XBPP

  2348.         char *dither= " dithered";

  2349. #else

  2350.         char *dither= "";

  2351. #endif

  2352.         if (flags&SWS_FAST_BILINEAR)

  2353.             av_log(c, AV_LOG_INFO, "SwScaler: FAST_BILINEAR scaler, ");

  2354.         else if (flags&SWS_BILINEAR)

  2355.             av_log(c, AV_LOG_INFO, "SwScaler: BILINEAR scaler, ");

  2356.         else if (flags&SWS_BICUBIC)

  2357.             av_log(c, AV_LOG_INFO, "SwScaler: BICUBIC scaler, ");

  2358.         else if (flags&SWS_X)

  2359.             av_log(c, AV_LOG_INFO, "SwScaler: Experimental scaler, ");

  2360.         else if (flags&SWS_POINT)

  2361.             av_log(c, AV_LOG_INFO, "SwScaler: Nearest Neighbor / POINT scaler, ");

  2362.         else if (flags&SWS_AREA)

  2363.             av_log(c, AV_LOG_INFO, "SwScaler: Area Averageing scaler, ");

  2364.         else if (flags&SWS_BICUBLIN)

  2365.             av_log(c, AV_LOG_INFO, "SwScaler: luma BICUBIC / chroma BILINEAR scaler, ");

  2366.         else if (flags&SWS_GAUSS)

  2367.             av_log(c, AV_LOG_INFO, "SwScaler: Gaussian scaler, ");

  2368.         else if (flags&SWS_SINC)

  2369.             av_log(c, AV_LOG_INFO, "SwScaler: Sinc scaler, ");

  2370.         else if (flags&SWS_LANCZOS)

  2371.             av_log(c, AV_LOG_INFO, "SwScaler: Lanczos scaler, ");

  2372.         else if (flags&SWS_SPLINE)

  2373.             av_log(c, AV_LOG_INFO, "SwScaler: Bicubic spline scaler, ");

  2374.         else

  2375.             av_log(c, AV_LOG_INFO, "SwScaler: ehh flags invalid?! ");

  2376. 

  2377.         if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)

  2378.             av_log(c, AV_LOG_INFO, "from %s to%s %s ",

  2379.                    sws_format_name(srcFormat), dither, sws_format_name(dstFormat));

  2380.         else

  2381.             av_log(c, AV_LOG_INFO, "from %s to %s ",

  2382.                    sws_format_name(srcFormat), sws_format_name(dstFormat));

  2383. 

  2384.         if (flags & SWS_CPU_CAPS_MMX2)

  2385.             av_log(c, AV_LOG_INFO, "using MMX2\n");

  2386.         else if (flags & SWS_CPU_CAPS_3DNOW)

  2387.             av_log(c, AV_LOG_INFO, "using 3DNOW\n");

  2388.         else if (flags & SWS_CPU_CAPS_MMX)

  2389.             av_log(c, AV_LOG_INFO, "using MMX\n");

  2390.         else if (flags & SWS_CPU_CAPS_ALTIVEC)

  2391.             av_log(c, AV_LOG_INFO, "using AltiVec\n");

  2392.         else

  2393.             av_log(c, AV_LOG_INFO, "using C\n");

  2394.     }

  2395. 

  2396.     if (flags & SWS_PRINT_INFO)

  2397.     {

  2398.         if (flags & SWS_CPU_CAPS_MMX)

  2399.         {

  2400.             if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))

  2401.                 av_log(c, AV_LOG_VERBOSE, "SwScaler: using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");

  2402.             else

  2403.             {

  2404.                 if (c->hLumFilterSize==4)

  2405.                     av_log(c, AV_LOG_VERBOSE, "SwScaler: using 4-tap MMX scaler for horizontal luminance scaling\n");

  2406.                 else if (c->hLumFilterSize==8)

  2407.                     av_log(c, AV_LOG_VERBOSE, "SwScaler: using 8-tap MMX scaler for horizontal luminance scaling\n");

  2408.                 else

  2409.                     av_log(c, AV_LOG_VERBOSE, "SwScaler: using n-tap MMX scaler for horizontal luminance scaling\n");

  2410. 

  2411.                 if (c->hChrFilterSize==4)

  2412.                     av_log(c, AV_LOG_VERBOSE, "SwScaler: using 4-tap MMX scaler for horizontal chrominance scaling\n");

  2413.                 else if (c->hChrFilterSize==8)

  2414.                     av_log(c, AV_LOG_VERBOSE, "SwScaler: using 8-tap MMX scaler for horizontal chrominance scaling\n");

  2415.                 else

  2416.                     av_log(c, AV_LOG_VERBOSE, "SwScaler: using n-tap MMX scaler for horizontal chrominance scaling\n");

  2417.             }

  2418.         }

  2419.         else

  2420.         {

  2421. #if defined(ARCH_X86)

  2422.             av_log(c, AV_LOG_VERBOSE, "SwScaler: using X86-Asm scaler for horizontal scaling\n");

  2423. #else

  2424.             if (flags & SWS_FAST_BILINEAR)

  2425.                 av_log(c, AV_LOG_VERBOSE, "SwScaler: using FAST_BILINEAR C scaler for horizontal scaling\n");

  2426.             else

  2427.                 av_log(c, AV_LOG_VERBOSE, "SwScaler: using C scaler for horizontal scaling\n");

  2428. #endif

  2429.         }

  2430.         if (isPlanarYUV(dstFormat))

  2431.         {

  2432.             if (c->vLumFilterSize==1)

  2433.                 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2434.             else

  2435.                 av_log(c, AV_LOG_VERBOSE, "SwScaler: using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2436.         }

  2437.         else

  2438.         {

  2439.             if (c->vLumFilterSize==1 && c->vChrFilterSize==2)

  2440.                 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"

  2441.                        "SwScaler:       2-tap scaler for vertical chrominance scaling (BGR)\n",(flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2442.             else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)

  2443.                 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2444.             else

  2445.                 av_log(c, AV_LOG_VERBOSE, "SwScaler: using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2446.         }

  2447. 

  2448.         if (dstFormat==PIX_FMT_BGR24)

  2449.             av_log(c, AV_LOG_VERBOSE, "SwScaler: using %s YV12->BGR24 Converter\n",

  2450.                    (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));

  2451.         else if (dstFormat==PIX_FMT_RGB32)

  2452.             av_log(c, AV_LOG_VERBOSE, "SwScaler: using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2453.         else if (dstFormat==PIX_FMT_BGR565)

  2454.             av_log(c, AV_LOG_VERBOSE, "SwScaler: using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2455.         else if (dstFormat==PIX_FMT_BGR555)

  2456.             av_log(c, AV_LOG_VERBOSE, "SwScaler: using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2457. 

  2458.         av_log(c, AV_LOG_VERBOSE, "SwScaler: %dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);

  2459.     }

  2460.     if (flags & SWS_PRINT_INFO)

  2461.     {

  2462.         av_log(c, AV_LOG_DEBUG, "SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2463.                c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);

  2464.         av_log(c, AV_LOG_DEBUG, "SwScaler:Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2465.                c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);

  2466.     }

  2467. 

  2468.     c->swScale= getSwsFunc(flags);

  2469.     return c;

  2470. }

  2471. 

  2472. /**

  2473.  * swscale warper, so we don't need to export the SwsContext.

  2474.  * assumes planar YUV to be in YUV order instead of YVU

  2475.  */

  2476. int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2477.               int srcSliceH, uint8_t* dst[], int dstStride[]){

  2478.     int i;

  2479.     uint8_t* src2[4]= {src[0], src[1], src[2]};

  2480.     uint32_t pal[256];

  2481.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  2482.         av_log(c, AV_LOG_ERROR, "swScaler: slices start in the middle!\n");

  2483.         return 0;

  2484.     }

  2485.     if (c->sliceDir == 0) {

  2486.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  2487.     }

  2488. 

  2489.     if (c->srcFormat == PIX_FMT_PAL8){

  2490.         for (i=0; i<256; i++){

  2491.             int p= ((uint32_t*)(src[1]))[i];

  2492.             int r= (p>>16)&0xFF;

  2493.             int g= (p>> 8)&0xFF;

  2494.             int b=  p     &0xFF;

  2495.             int y= av_clip_uint8(((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16 );

  2496.             int u= av_clip_uint8(((RU*r + GU*g + BU*b)>>RGB2YUV_SHIFT) + 128);

  2497.             int v= av_clip_uint8(((RV*r + GV*g + BV*b)>>RGB2YUV_SHIFT) + 128);

  2498.             pal[i]= y + (u<<8) + (v<<16);

  2499.         }

  2500.         src2[1]= pal;

  2501.     }

  2502. 

  2503.     // copy strides, so they can safely be modified

  2504.     if (c->sliceDir == 1) {

  2505.         // slices go from top to bottom

  2506.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2]};

  2507.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2]};

  2508.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);

  2509.     } else {

  2510.         // slices go from bottom to top => we flip the image internally

  2511.         uint8_t* dst2[4]= {dst[0] + (c->dstH-1)*dstStride[0],

  2512.                            dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],

  2513.                            dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2]};

  2514.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2]};

  2515.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2]};

  2516. 

  2517.         src2[0] += (srcSliceH-1)*srcStride[0];

  2518.         if (c->srcFormat != PIX_FMT_PAL8)

  2519.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  2520.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  2521. 

  2522.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  2523.     }

  2524. }

  2525. 

  2526. /**

  2527.  * swscale warper, so we don't need to export the SwsContext

  2528.  */

  2529. int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2530.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  2531.     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);

  2532. }

  2533. 

  2534. SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,

  2535.                                 float lumaSharpen, float chromaSharpen,

  2536.                                 float chromaHShift, float chromaVShift,

  2537.                                 int verbose)

  2538. {

  2539.     SwsFilter *filter= av_malloc(sizeof(SwsFilter));

  2540. 

  2541.     if (lumaGBlur!=0.0){

  2542.         filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);

  2543.         filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);

  2544.     }else{

  2545.         filter->lumH= sws_getIdentityVec();

  2546.         filter->lumV= sws_getIdentityVec();

  2547.     }

  2548. 

  2549.     if (chromaGBlur!=0.0){

  2550.         filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);

  2551.         filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);

  2552.     }else{

  2553.         filter->chrH= sws_getIdentityVec();

  2554.         filter->chrV= sws_getIdentityVec();

  2555.     }

  2556. 

  2557.     if (chromaSharpen!=0.0){

  2558.         SwsVector *id= sws_getIdentityVec();

  2559.         sws_scaleVec(filter->chrH, -chromaSharpen);

  2560.         sws_scaleVec(filter->chrV, -chromaSharpen);

  2561.         sws_addVec(filter->chrH, id);

  2562.         sws_addVec(filter->chrV, id);

  2563.         sws_freeVec(id);

  2564.     }

  2565. 

  2566.     if (lumaSharpen!=0.0){

  2567.         SwsVector *id= sws_getIdentityVec();

  2568.         sws_scaleVec(filter->lumH, -lumaSharpen);

  2569.         sws_scaleVec(filter->lumV, -lumaSharpen);

  2570.         sws_addVec(filter->lumH, id);

  2571.         sws_addVec(filter->lumV, id);

  2572.         sws_freeVec(id);

  2573.     }

  2574. 

  2575.     if (chromaHShift != 0.0)

  2576.         sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));

  2577. 

  2578.     if (chromaVShift != 0.0)

  2579.         sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));

  2580. 

  2581.     sws_normalizeVec(filter->chrH, 1.0);

  2582.     sws_normalizeVec(filter->chrV, 1.0);

  2583.     sws_normalizeVec(filter->lumH, 1.0);

  2584.     sws_normalizeVec(filter->lumV, 1.0);

  2585. 

  2586.     if (verbose) sws_printVec(filter->chrH);

  2587.     if (verbose) sws_printVec(filter->lumH);

  2588. 

  2589.     return filter;

  2590. }

  2591. 

  2592. /**

  2593.  * returns a normalized gaussian curve used to filter stuff

  2594.  * quality=3 is high quality, lowwer is lowwer quality

  2595.  */

  2596. SwsVector *sws_getGaussianVec(double variance, double quality){

  2597.     const int length= (int)(variance*quality + 0.5) | 1;

  2598.     int i;

  2599.     double *coeff= av_malloc(length*sizeof(double));

  2600.     double middle= (length-1)*0.5;

  2601.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2602. 

  2603.     vec->coeff= coeff;

  2604.     vec->length= length;

  2605. 

  2606.     for (i=0; i<length; i++)

  2607.     {

  2608.         double dist= i-middle;

  2609.         coeff[i]= exp( -dist*dist/(2*variance*variance) ) / sqrt(2*variance*PI);

  2610.     }

  2611. 

  2612.     sws_normalizeVec(vec, 1.0);

  2613. 

  2614.     return vec;

  2615. }

  2616. 

  2617. SwsVector *sws_getConstVec(double c, int length){

  2618.     int i;

  2619.     double *coeff= av_malloc(length*sizeof(double));

  2620.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2621. 

  2622.     vec->coeff= coeff;

  2623.     vec->length= length;

  2624. 

  2625.     for (i=0; i<length; i++)

  2626.         coeff[i]= c;

  2627. 

  2628.     return vec;

  2629. }

  2630. 

  2631. 

  2632. SwsVector *sws_getIdentityVec(void){

  2633.     return sws_getConstVec(1.0, 1);

  2634. }

  2635. 

  2636. double sws_dcVec(SwsVector *a){

  2637.     int i;

  2638.     double sum=0;

  2639. 

  2640.     for (i=0; i<a->length; i++)

  2641.         sum+= a->coeff[i];

  2642. 

  2643.     return sum;

  2644. }

  2645. 

  2646. void sws_scaleVec(SwsVector *a, double scalar){

  2647.     int i;

  2648. 

  2649.     for (i=0; i<a->length; i++)

  2650.         a->coeff[i]*= scalar;

  2651. }

  2652. 

  2653. void sws_normalizeVec(SwsVector *a, double height){

  2654.     sws_scaleVec(a, height/sws_dcVec(a));

  2655. }

  2656. 

  2657. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){

  2658.     int length= a->length + b->length - 1;

  2659.     double *coeff= av_malloc(length*sizeof(double));

  2660.     int i, j;

  2661.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2662. 

  2663.     vec->coeff= coeff;

  2664.     vec->length= length;

  2665. 

  2666.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2667. 

  2668.     for (i=0; i<a->length; i++)

  2669.     {

  2670.         for (j=0; j<b->length; j++)

  2671.         {

  2672.             coeff[i+j]+= a->coeff[i]*b->coeff[j];

  2673.         }

  2674.     }

  2675. 

  2676.     return vec;

  2677. }

  2678. 

  2679. static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){

  2680.     int length= FFMAX(a->length, b->length);

  2681.     double *coeff= av_malloc(length*sizeof(double));

  2682.     int i;

  2683.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2684. 

  2685.     vec->coeff= coeff;

  2686.     vec->length= length;

  2687. 

  2688.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2689. 

  2690.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  2691.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];

  2692. 

  2693.     return vec;

  2694. }

  2695. 

  2696. static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){

  2697.     int length= FFMAX(a->length, b->length);

  2698.     double *coeff= av_malloc(length*sizeof(double));

  2699.     int i;

  2700.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2701. 

  2702.     vec->coeff= coeff;

  2703.     vec->length= length;

  2704. 

  2705.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2706. 

  2707.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  2708.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];

  2709. 

  2710.     return vec;

  2711. }

  2712. 

  2713. /* shift left / or right if "shift" is negative */

  2714. static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){

  2715.     int length= a->length + FFABS(shift)*2;

  2716.     double *coeff= av_malloc(length*sizeof(double));

  2717.     int i;

  2718.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2719. 

  2720.     vec->coeff= coeff;

  2721.     vec->length= length;

  2722. 

  2723.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2724. 

  2725.     for (i=0; i<a->length; i++)

  2726.     {

  2727.         coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];

  2728.     }

  2729. 

  2730.     return vec;

  2731. }

  2732. 

  2733. void sws_shiftVec(SwsVector *a, int shift){

  2734.     SwsVector *shifted= sws_getShiftedVec(a, shift);

  2735.     av_free(a->coeff);

  2736.     a->coeff= shifted->coeff;

  2737.     a->length= shifted->length;

  2738.     av_free(shifted);

  2739. }

  2740. 

  2741. void sws_addVec(SwsVector *a, SwsVector *b){

  2742.     SwsVector *sum= sws_sumVec(a, b);

  2743.     av_free(a->coeff);

  2744.     a->coeff= sum->coeff;

  2745.     a->length= sum->length;

  2746.     av_free(sum);

  2747. }

  2748. 

  2749. void sws_subVec(SwsVector *a, SwsVector *b){

  2750.     SwsVector *diff= sws_diffVec(a, b);

  2751.     av_free(a->coeff);

  2752.     a->coeff= diff->coeff;

  2753.     a->length= diff->length;

  2754.     av_free(diff);

  2755. }

  2756. 

  2757. void sws_convVec(SwsVector *a, SwsVector *b){

  2758.     SwsVector *conv= sws_getConvVec(a, b);

  2759.     av_free(a->coeff);

  2760.     a->coeff= conv->coeff;

  2761.     a->length= conv->length;

  2762.     av_free(conv);

  2763. }

  2764. 

  2765. SwsVector *sws_cloneVec(SwsVector *a){

  2766.     double *coeff= av_malloc(a->length*sizeof(double));

  2767.     int i;

  2768.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2769. 

  2770.     vec->coeff= coeff;

  2771.     vec->length= a->length;

  2772. 

  2773.     for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];

  2774. 

  2775.     return vec;

  2776. }

  2777. 

  2778. void sws_printVec(SwsVector *a){

  2779.     int i;

  2780.     double max=0;

  2781.     double min=0;

  2782.     double range;

  2783. 

  2784.     for (i=0; i<a->length; i++)

  2785.         if (a->coeff[i]>max) max= a->coeff[i];

  2786. 

  2787.     for (i=0; i<a->length; i++)

  2788.         if (a->coeff[i]<min) min= a->coeff[i];

  2789. 

  2790.     range= max - min;

  2791. 

  2792.     for (i=0; i<a->length; i++)

  2793.     {

  2794.         int x= (int)((a->coeff[i]-min)*60.0/range +0.5);

  2795.         av_log(NULL, AV_LOG_DEBUG, "%1.3f ", a->coeff[i]);

  2796.         for (;x>0; x--) av_log(NULL, AV_LOG_DEBUG, " ");

  2797.         av_log(NULL, AV_LOG_DEBUG, "|\n");

  2798.     }

  2799. }

  2800. 

  2801. void sws_freeVec(SwsVector *a){

  2802.     if (!a) return;

  2803.     av_free(a->coeff);

  2804.     a->coeff=NULL;

  2805.     a->length=0;

  2806.     av_free(a);

  2807. }

  2808. 

  2809. void sws_freeFilter(SwsFilter *filter){

  2810.     if (!filter) return;

  2811. 

  2812.     if (filter->lumH) sws_freeVec(filter->lumH);

  2813.     if (filter->lumV) sws_freeVec(filter->lumV);

  2814.     if (filter->chrH) sws_freeVec(filter->chrH);

  2815.     if (filter->chrV) sws_freeVec(filter->chrV);

  2816.     av_free(filter);

  2817. }

  2818. 

  2819. 

  2820. void sws_freeContext(SwsContext *c){

  2821.     int i;

  2822.     if (!c) return;

  2823. 

  2824.     if (c->lumPixBuf)

  2825.     {

  2826.         for (i=0; i<c->vLumBufSize; i++)

  2827.         {

  2828.             av_free(c->lumPixBuf[i]);

  2829.             c->lumPixBuf[i]=NULL;

  2830.         }

  2831.         av_free(c->lumPixBuf);

  2832.         c->lumPixBuf=NULL;

  2833.     }

  2834. 

  2835.     if (c->chrPixBuf)

  2836.     {

  2837.         for (i=0; i<c->vChrBufSize; i++)

  2838.         {

  2839.             av_free(c->chrPixBuf[i]);

  2840.             c->chrPixBuf[i]=NULL;

  2841.         }

  2842.         av_free(c->chrPixBuf);

  2843.         c->chrPixBuf=NULL;

  2844.     }

  2845. 

  2846.     av_free(c->vLumFilter);

  2847.     c->vLumFilter = NULL;

  2848.     av_free(c->vChrFilter);

  2849.     c->vChrFilter = NULL;

  2850.     av_free(c->hLumFilter);

  2851.     c->hLumFilter = NULL;

  2852.     av_free(c->hChrFilter);

  2853.     c->hChrFilter = NULL;

  2854. #ifdef HAVE_ALTIVEC

  2855.     av_free(c->vYCoeffsBank);

  2856.     c->vYCoeffsBank = NULL;

  2857.     av_free(c->vCCoeffsBank);

  2858.     c->vCCoeffsBank = NULL;

  2859. #endif

  2860. 

  2861.     av_free(c->vLumFilterPos);

  2862.     c->vLumFilterPos = NULL;

  2863.     av_free(c->vChrFilterPos);

  2864.     c->vChrFilterPos = NULL;

  2865.     av_free(c->hLumFilterPos);

  2866.     c->hLumFilterPos = NULL;

  2867.     av_free(c->hChrFilterPos);

  2868.     c->hChrFilterPos = NULL;

  2869. 

  2870. #if defined(ARCH_X86) && defined(CONFIG_GPL)

  2871. #ifdef MAP_ANONYMOUS

  2872.     if (c->funnyYCode) munmap(c->funnyYCode, MAX_FUNNY_CODE_SIZE);

  2873.     if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);

  2874. #else

  2875.     av_free(c->funnyYCode);

  2876.     av_free(c->funnyUVCode);

  2877. #endif

  2878.     c->funnyYCode=NULL;

  2879.     c->funnyUVCode=NULL;

  2880. #endif /* defined(ARCH_X86) */

  2881. 

  2882.     av_free(c->lumMmx2Filter);

  2883.     c->lumMmx2Filter=NULL;

  2884.     av_free(c->chrMmx2Filter);

  2885.     c->chrMmx2Filter=NULL;

  2886.     av_free(c->lumMmx2FilterPos);

  2887.     c->lumMmx2FilterPos=NULL;

  2888.     av_free(c->chrMmx2FilterPos);

  2889.     c->chrMmx2FilterPos=NULL;

  2890.     av_free(c->yuvTable);

  2891.     c->yuvTable=NULL;

  2892. 

  2893.     av_free(c);

  2894. }

  2895. 

  2896. /**

  2897.  * Checks if context is valid or reallocs a new one instead.

  2898.  * If context is NULL, just calls sws_getContext() to get a new one.

  2899.  * Otherwise, checks if the parameters are the same already saved in context.

  2900.  * If that is the case, returns the current context.

  2901.  * Otherwise, frees context and gets a new one.

  2902.  *

  2903.  * Be warned that srcFilter, dstFilter are not checked, they are

  2904.  * asumed to remain valid.

  2905.  */

  2906. struct SwsContext *sws_getCachedContext(struct SwsContext *context,

  2907.                                         int srcW, int srcH, int srcFormat,

  2908.                                         int dstW, int dstH, int dstFormat, int flags,

  2909.                                         SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)

  2910. {

  2911.     if (context != NULL) {

  2912.         if ((context->srcW != srcW) || (context->srcH != srcH) ||

  2913.             (context->srcFormat != srcFormat) ||

  2914.             (context->dstW != dstW) || (context->dstH != dstH) ||

  2915.             (context->dstFormat != dstFormat) || (context->flags != flags) ||

  2916.             (context->param != param))

  2917.         {

  2918.             sws_freeContext(context);

  2919.             context = NULL;

  2920.         }

  2921.     }

  2922.     if (context == NULL) {

  2923.         return sws_getContext(srcW, srcH, srcFormat,

  2924.                               dstW, dstH, dstFormat, flags,

  2925.                               srcFilter, dstFilter, param);

  2926.     }

  2927.     return context;

  2928. }