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.
15839 Commits
32 Branches
1.2GB
Tree: dd1c8f3e6e
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'dd1c8f3e6e'
${ noResults }
FFmpeg/libswscale/swscale.c

3064 lines
102KB
Raw Blame History 

  1. /*

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

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

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

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

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

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

  10.  *

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

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

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

  14.  * GNU General Public License for more details.

  15.  *

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

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

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

  19.  *

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

  21.  * under the LGPL license too

  22.  */

  23. 

  24. /*

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

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

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

  28. 

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

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

  31.   x -> x

  32.   YUV9 -> YV12

  33.   YUV9/YV12 -> Y800

  34.   Y800 -> YUV9/YV12

  35.   BGR24 -> BGR32 & RGB24 -> RGB32

  36.   BGR32 -> BGR24 & RGB32 -> RGB24

  37.   BGR15 -> BGR16

  38. */

  39. 

  40. /*

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

  42.  YV12 -> BGR16

  43.  YV12 -> YV12

  44.  BGR15 -> BGR16

  45.  BGR16 -> BGR16

  46.  YVU9 -> YV12

  47. 

  48. untested special converters

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

  50.   YV12/I420 -> YV12/I420

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

  52.   BGR24 -> BGR32 & RGB24 -> RGB32

  53.   BGR32 -> BGR24 & RGB32 -> RGB24

  54.   BGR24 -> YV12

  55. */

  56. 

  57. #define _SVID_SOURCE //needed for MAP_ANONYMOUS

  58. #include <inttypes.h>

  59. #include <string.h>

  60. #include <math.h>

  61. #include <stdio.h>

  62. #include <unistd.h>

  63. #include "config.h"

  64. #include <assert.h>

  65. #ifdef HAVE_SYS_MMAN_H

  66. #include <sys/mman.h>

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

  68. #define MAP_ANONYMOUS MAP_ANON

  69. #endif

  70. #endif

  71. #include "swscale.h"

  72. #include "swscale_internal.h"

  73. #include "rgb2rgb.h"

  74. #include "libavutil/x86_cpu.h"

  75. #include "libavutil/bswap.h"

  76. 

  77. unsigned swscale_version(void)

  78. {

  79.     return LIBSWSCALE_VERSION_INT;

  80. }

  81. 

  82. #undef MOVNTQ

  83. #undef PAVGB

  84. 

  85. //#undef HAVE_MMX2

  86. //#define HAVE_3DNOW

  87. //#undef HAVE_MMX

  88. //#undef ARCH_X86

  89. //#define WORDS_BIGENDIAN

  90. #define DITHER1XBPP

  91. 

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

  93. 

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

  95. 

  96. #ifdef M_PI

  97. #define PI M_PI

  98. #else

  99. #define PI 3.14159265358979323846

  100. #endif

  101. 

  102. #define isSupportedIn(x)    (       \

  103.            (x)==PIX_FMT_YUV420P     \

  104.         || (x)==PIX_FMT_YUVA420P    \

  105.         || (x)==PIX_FMT_YUYV422     \

  106.         || (x)==PIX_FMT_UYVY422     \

  107.         || (x)==PIX_FMT_RGB32       \

  108.         || (x)==PIX_FMT_RGB32_1     \

  109.         || (x)==PIX_FMT_BGR24       \

  110.         || (x)==PIX_FMT_BGR565      \

  111.         || (x)==PIX_FMT_BGR555      \

  112.         || (x)==PIX_FMT_BGR32       \

  113.         || (x)==PIX_FMT_BGR32_1     \

  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.         || (x)==PIX_FMT_YUV440P     \

  130.     )

  131. #define isSupportedOut(x)   (       \

  132.            (x)==PIX_FMT_YUV420P     \

  133.         || (x)==PIX_FMT_YUYV422     \

  134.         || (x)==PIX_FMT_UYVY422     \

  135.         || (x)==PIX_FMT_YUV444P     \

  136.         || (x)==PIX_FMT_YUV422P     \

  137.         || (x)==PIX_FMT_YUV411P     \

  138.         || isRGB(x)                 \

  139.         || isBGR(x)                 \

  140.         || (x)==PIX_FMT_NV12        \

  141.         || (x)==PIX_FMT_NV21        \

  142.         || (x)==PIX_FMT_GRAY16BE    \

  143.         || (x)==PIX_FMT_GRAY16LE    \

  144.         || (x)==PIX_FMT_GRAY8       \

  145.         || (x)==PIX_FMT_YUV410P     \

  146.     )

  147. #define isPacked(x)         (       \

  148.            (x)==PIX_FMT_PAL8        \

  149.         || (x)==PIX_FMT_YUYV422     \

  150.         || (x)==PIX_FMT_UYVY422     \

  151.         || isRGB(x)                 \

  152.         || isBGR(x)                 \

  153.     )

  154. 

  155. #define RGB2YUV_SHIFT 16

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

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

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

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

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

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

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

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

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

  165. 

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

  167. 

  168. /*

  169. NOTES

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

  171. 

  172. TODO

  173. more intelligent misalignment avoidance for the horizontal scaler

  174. write special vertical cubic upscale version

  175. Optimize C code (yv12 / minmax)

  176. add support for packed pixel yuv input & output

  177. add support for Y8 output

  178. optimize bgr24 & bgr32

  179. add BGR4 output support

  180. write special BGR->BGR scaler

  181. */

  182. 

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

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

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

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

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

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

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

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

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

  192. 

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

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

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

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

  197. 

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

  199.         0x0103010301030103LL,

  200.         0x0200020002000200LL,};

  201. 

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

  203.         0x0602060206020602LL,

  204.         0x0004000400040004LL,};

  205. 

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

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

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

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

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

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

  212. 

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

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

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

  216. 

  217. #ifdef FAST_BGR2YV12

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

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

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

  221. #else

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

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

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

  225. #endif /* FAST_BGR2YV12 */

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

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

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

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

  230. 

  231. // clipping helper table for C implementations:

  232. static unsigned char clip_table[768];

  233. 

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

  235. 

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

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

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

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

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

  241. 

  242. const char *sws_format_name(enum PixelFormat format)

  243. {

  244.     switch (format) {

  245.         case PIX_FMT_YUV420P:

  246.             return "yuv420p";

  247.         case PIX_FMT_YUVA420P:

  248.             return "yuva420p";

  249.         case PIX_FMT_YUYV422:

  250.             return "yuyv422";

  251.         case PIX_FMT_RGB24:

  252.             return "rgb24";

  253.         case PIX_FMT_BGR24:

  254.             return "bgr24";

  255.         case PIX_FMT_YUV422P:

  256.             return "yuv422p";

  257.         case PIX_FMT_YUV444P:

  258.             return "yuv444p";

  259.         case PIX_FMT_RGB32:

  260.             return "rgb32";

  261.         case PIX_FMT_YUV410P:

  262.             return "yuv410p";

  263.         case PIX_FMT_YUV411P:

  264.             return "yuv411p";

  265.         case PIX_FMT_RGB565:

  266.             return "rgb565";

  267.         case PIX_FMT_RGB555:

  268.             return "rgb555";

  269.         case PIX_FMT_GRAY16BE:

  270.             return "gray16be";

  271.         case PIX_FMT_GRAY16LE:

  272.             return "gray16le";

  273.         case PIX_FMT_GRAY8:

  274.             return "gray8";

  275.         case PIX_FMT_MONOWHITE:

  276.             return "mono white";

  277.         case PIX_FMT_MONOBLACK:

  278.             return "mono black";

  279.         case PIX_FMT_PAL8:

  280.             return "Palette";

  281.         case PIX_FMT_YUVJ420P:

  282.             return "yuvj420p";

  283.         case PIX_FMT_YUVJ422P:

  284.             return "yuvj422p";

  285.         case PIX_FMT_YUVJ444P:

  286.             return "yuvj444p";

  287.         case PIX_FMT_XVMC_MPEG2_MC:

  288.             return "xvmc_mpeg2_mc";

  289.         case PIX_FMT_XVMC_MPEG2_IDCT:

  290.             return "xvmc_mpeg2_idct";

  291.         case PIX_FMT_UYVY422:

  292.             return "uyvy422";

  293.         case PIX_FMT_UYYVYY411:

  294.             return "uyyvyy411";

  295.         case PIX_FMT_RGB32_1:

  296.             return "rgb32x";

  297.         case PIX_FMT_BGR32_1:

  298.             return "bgr32x";

  299.         case PIX_FMT_BGR32:

  300.             return "bgr32";

  301.         case PIX_FMT_BGR565:

  302.             return "bgr565";

  303.         case PIX_FMT_BGR555:

  304.             return "bgr555";

  305.         case PIX_FMT_BGR8:

  306.             return "bgr8";

  307.         case PIX_FMT_BGR4:

  308.             return "bgr4";

  309.         case PIX_FMT_BGR4_BYTE:

  310.             return "bgr4 byte";

  311.         case PIX_FMT_RGB8:

  312.             return "rgb8";

  313.         case PIX_FMT_RGB4:

  314.             return "rgb4";

  315.         case PIX_FMT_RGB4_BYTE:

  316.             return "rgb4 byte";

  317.         case PIX_FMT_NV12:

  318.             return "nv12";

  319.         case PIX_FMT_NV21:

  320.             return "nv21";

  321.         case PIX_FMT_YUV440P:

  322.             return "yuv440p";

  323.         default:

  324.             return "Unknown format";

  325.     }

  326. }

  327. 

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

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

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

  331. {

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

  333.     int i;

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

  335.     {

  336.         int val=1<<18;

  337.         int j;

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

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

  340. 

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

  342.     }

  343. 

  344.     if (uDest)

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

  346.         {

  347.             int u=1<<18;

  348.             int v=1<<18;

  349.             int j;

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

  351.             {

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

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

  354.             }

  355. 

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

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

  358.         }

  359. }

  360. 

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

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

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

  364. {

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

  366.     int i;

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

  368.     {

  369.         int val=1<<18;

  370.         int j;

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

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

  373. 

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

  375.     }

  376. 

  377.     if (!uDest)

  378.         return;

  379. 

  380.     if (dstFormat == PIX_FMT_NV12)

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

  382.         {

  383.             int u=1<<18;

  384.             int v=1<<18;

  385.             int j;

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

  387.             {

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

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

  390.             }

  391. 

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

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

  394.         }

  395.     else

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

  397.         {

  398.             int u=1<<18;

  399.             int v=1<<18;

  400.             int j;

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

  402.             {

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

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

  405.             }

  406. 

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

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

  409.         }

  410. }

  411. 

  412. #define YSCALE_YUV_2_PACKEDX_C(type) \

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

  414.         int j;\

  415.         int Y1 = 1<<18;\

  416.         int Y2 = 1<<18;\

  417.         int U  = 1<<18;\

  418.         int V  = 1<<18;\

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

  420.         const int i2= 2*i;\

  421.         \

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

  423.         {\

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

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

  426.         }\

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

  428.         {\

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

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

  431.         }\

  432.         Y1>>=19;\

  433.         Y2>>=19;\

  434.         U >>=19;\

  435.         V >>=19;\

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

  437.         {\

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

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

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

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

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

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

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

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

  446.         }

  447. 

  448. #define YSCALE_YUV_2_GRAY16_C(type) \

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

  450.         int j;\

  451.         int Y1 = 1<<18;\

  452.         int Y2 = 1<<18;\

  453.         int U  = 1<<18;\

  454.         int V  = 1<<18;\

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

  456.         const int i2= 2*i;\

  457.         \

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

  459.         {\

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

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

  462.         }\

  463.         Y1>>=11;\

  464.         Y2>>=11;\

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

  466.         {\

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

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

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

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

  471.         }

  472. 

  473. #define YSCALE_YUV_2_RGBX_C(type) \

  474.     YSCALE_YUV_2_PACKEDX_C(type)  \

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

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

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

  478. 

  479. #define YSCALE_YUV_2_PACKED2_C   \

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

  481.         const int i2= 2*i;       \

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

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

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

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

  486. 

  487. #define YSCALE_YUV_2_GRAY16_2_C   \

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

  489.         const int i2= 2*i;       \

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

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

  492. 

  493. #define YSCALE_YUV_2_RGB2_C(type) \

  494.     YSCALE_YUV_2_PACKED2_C\

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

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

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

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

  499. 

  500. #define YSCALE_YUV_2_PACKED1_C \

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

  502.         const int i2= 2*i;\

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

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

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

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

  507. 

  508. #define YSCALE_YUV_2_GRAY16_1_C \

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

  510.         const int i2= 2*i;\

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

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

  513. 

  514. #define YSCALE_YUV_2_RGB1_C(type) \

  515.     YSCALE_YUV_2_PACKED1_C\

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

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

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

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

  520. 

  521. #define YSCALE_YUV_2_PACKED1B_C \

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

  523.         const int i2= 2*i;\

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

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

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

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

  528. 

  529. #define YSCALE_YUV_2_RGB1B_C(type) \

  530.     YSCALE_YUV_2_PACKED1B_C\

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

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

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

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

  535. 

  536. #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16)\

  537.     switch(c->dstFormat)\

  538.     {\

  539.     case PIX_FMT_RGB32:\

  540.     case PIX_FMT_BGR32:\

  541.     case PIX_FMT_RGB32_1:\

  542.     case PIX_FMT_BGR32_1:\

  543.         func(uint32_t)\

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

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

  546.         }                \

  547.         break;\

  548.     case PIX_FMT_RGB24:\

  549.         func(uint8_t)\

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

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

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

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

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

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

  556.             dest+=6;\

  557.         }\

  558.         break;\

  559.     case PIX_FMT_BGR24:\

  560.         func(uint8_t)\

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

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

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

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

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

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

  567.             dest+=6;\

  568.         }\

  569.         break;\

  570.     case PIX_FMT_RGB565:\

  571.     case PIX_FMT_BGR565:\

  572.         {\

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

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

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

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

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

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

  579.             func(uint16_t)\

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

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

  582.             }\

  583.         }\

  584.         break;\

  585.     case PIX_FMT_RGB555:\

  586.     case PIX_FMT_BGR555:\

  587.         {\

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

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

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

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

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

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

  594.             func(uint16_t)\

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

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

  597.             }\

  598.         }\

  599.         break;\

  600.     case PIX_FMT_RGB8:\

  601.     case PIX_FMT_BGR8:\

  602.         {\

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

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

  605.             func(uint8_t)\

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

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

  608.             }\

  609.         }\

  610.         break;\

  611.     case PIX_FMT_RGB4:\

  612.     case PIX_FMT_BGR4:\

  613.         {\

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

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

  616.             func(uint8_t)\

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

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

  619.             }\

  620.         }\

  621.         break;\

  622.     case PIX_FMT_RGB4_BYTE:\

  623.     case PIX_FMT_BGR4_BYTE:\

  624.         {\

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

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

  627.             func(uint8_t)\

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

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

  630.             }\

  631.         }\

  632.         break;\

  633.     case PIX_FMT_MONOBLACK:\

  634.         {\

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

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

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

  638.                 int acc;\

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

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

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

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

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

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

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

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

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

  648.                 dest++;\

  649.             }\

  650. \

  651. /*\

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

  653. {\

  654.             int acc=0;\

  655.             int left=0;\

  656.             static int top[1024];\

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

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

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

  660.             int topLeft=0;\

  661.             int shift=0;\

  662.             int count=0;\

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

  664.             int error_new=0;\

  665.             int error_in3=0;\

  666.             int f=0;\

  667.             \

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

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

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

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

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

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

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

  675. \

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

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

  678.                 f= error_new - error_in3*4;\

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

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

  681. \

  682.                 topLeft= top[i];\

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

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

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

  686. \

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

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

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

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

  691.                 }\

  692.             }\

  693. }\

  694. */\

  695.         }\

  696.         break;\

  697.     case PIX_FMT_YUYV422:\

  698.         func2\

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

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

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

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

  703.         }                \

  704.         break;\

  705.     case PIX_FMT_UYVY422:\

  706.         func2\

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

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

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

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

  711.         }                \

  712.         break;\

  713.     case PIX_FMT_GRAY16BE:\

  714.         func_g16\

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

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

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

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

  719.         }                \

  720.         break;\

  721.     case PIX_FMT_GRAY16LE:\

  722.         func_g16\

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

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

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

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

  727.         }                \

  728.         break;\

  729.     }\

  730. 

  731. 

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

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

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

  735. {

  736.     int i;

  737.     switch(c->dstFormat)

  738.     {

  739.     case PIX_FMT_BGR32:

  740.     case PIX_FMT_RGB32:

  741.     case PIX_FMT_BGR32_1:

  742.     case PIX_FMT_RGB32_1:

  743.         YSCALE_YUV_2_RGBX_C(uint32_t)

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

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

  746.         }

  747.         break;

  748.     case PIX_FMT_RGB24:

  749.         YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

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

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

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

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

  756.             dest+=6;

  757.         }

  758.         break;

  759.     case PIX_FMT_BGR24:

  760.         YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

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

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

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

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

  767.             dest+=6;

  768.         }

  769.         break;

  770.     case PIX_FMT_RGB565:

  771.     case PIX_FMT_BGR565:

  772.         {

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

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

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

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

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

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

  779.             YSCALE_YUV_2_RGBX_C(uint16_t)

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

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

  782.             }

  783.         }

  784.         break;

  785.     case PIX_FMT_RGB555:

  786.     case PIX_FMT_BGR555:

  787.         {

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

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

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

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

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

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

  794.             YSCALE_YUV_2_RGBX_C(uint16_t)

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

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

  797.             }

  798.         }

  799.         break;

  800.     case PIX_FMT_RGB8:

  801.     case PIX_FMT_BGR8:

  802.         {

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

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

  805.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  808.             }

  809.         }

  810.         break;

  811.     case PIX_FMT_RGB4:

  812.     case PIX_FMT_BGR4:

  813.         {

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

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

  816.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  819.             }

  820.         }

  821.         break;

  822.     case PIX_FMT_RGB4_BYTE:

  823.     case PIX_FMT_BGR4_BYTE:

  824.         {

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

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

  827.             YSCALE_YUV_2_RGBX_C(uint8_t)

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

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

  830.             }

  831.         }

  832.         break;

  833.     case PIX_FMT_MONOBLACK:

  834.         {

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

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

  837.             int acc=0;

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

  839.                 int j;

  840.                 int Y1=1<<18;

  841.                 int Y2=1<<18;

  842. 

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

  844.                 {

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

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

  847.                 }

  848.                 Y1>>=19;

  849.                 Y2>>=19;

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

  851.                 {

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

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

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

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

  856.                 }

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

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

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

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

  861.                     dest++;

  862.                 }

  863.             }

  864.         }

  865.         break;

  866.     case PIX_FMT_YUYV422:

  867.         YSCALE_YUV_2_PACKEDX_C(void)

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

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

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

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

  872.         }

  873.         break;

  874.     case PIX_FMT_UYVY422:

  875.         YSCALE_YUV_2_PACKEDX_C(void)

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

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

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

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

  880.         }

  881.         break;

  882.     case PIX_FMT_GRAY16BE:

  883.         YSCALE_YUV_2_GRAY16_C(void)

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

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

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

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

  888.         }

  889.         break;

  890.     case PIX_FMT_GRAY16LE:

  891.         YSCALE_YUV_2_GRAY16_C(void)

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

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

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

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

  896.         }

  897.         break;

  898.     }

  899. }

  900. 

  901. 

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

  903. //Plain C versions

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

  905. #define COMPILE_C

  906. #endif

  907. 

  908. #ifdef ARCH_POWERPC

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

  910. #define COMPILE_ALTIVEC

  911. #endif //HAVE_ALTIVEC

  912. #endif //ARCH_POWERPC

  913. 

  914. #if defined(ARCH_X86)

  915. 

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

  917. #define COMPILE_MMX

  918. #endif

  919. 

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

  921. #define COMPILE_MMX2

  922. #endif

  923. 

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

  925. #define COMPILE_3DNOW

  926. #endif

  927. #endif //ARCH_X86 || ARCH_X86_64

  928. 

  929. #undef HAVE_MMX

  930. #undef HAVE_MMX2

  931. #undef HAVE_3DNOW

  932. 

  933. #ifdef COMPILE_C

  934. #undef HAVE_MMX

  935. #undef HAVE_MMX2

  936. #undef HAVE_3DNOW

  937. #undef HAVE_ALTIVEC

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

  939. #include "swscale_template.c"

  940. #endif

  941. 

  942. #ifdef COMPILE_ALTIVEC

  943. #undef RENAME

  944. #define HAVE_ALTIVEC

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

  946. #include "swscale_template.c"

  947. #endif

  948. 

  949. #if defined(ARCH_X86)

  950. 

  951. //X86 versions

  952. /*

  953. #undef RENAME

  954. #undef HAVE_MMX

  955. #undef HAVE_MMX2

  956. #undef HAVE_3DNOW

  957. #define ARCH_X86

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

  959. #include "swscale_template.c"

  960. */

  961. //MMX versions

  962. #ifdef COMPILE_MMX

  963. #undef RENAME

  964. #define HAVE_MMX

  965. #undef HAVE_MMX2

  966. #undef HAVE_3DNOW

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

  968. #include "swscale_template.c"

  969. #endif

  970. 

  971. //MMX2 versions

  972. #ifdef COMPILE_MMX2

  973. #undef RENAME

  974. #define HAVE_MMX

  975. #define HAVE_MMX2

  976. #undef HAVE_3DNOW

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

  978. #include "swscale_template.c"

  979. #endif

  980. 

  981. //3DNOW versions

  982. #ifdef COMPILE_3DNOW

  983. #undef RENAME

  984. #define HAVE_MMX

  985. #undef HAVE_MMX2

  986. #define HAVE_3DNOW

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

  988. #include "swscale_template.c"

  989. #endif

  990. 

  991. #endif //ARCH_X86 || ARCH_X86_64

  992. 

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

  994. 

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

  996. {

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

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

  999.     else                return getSplineCoeff(        0.0,

  1000.                                              b+ 2.0*c + 3.0*d,

  1001.                                                     c + 3.0*d,

  1002.                                             -b- 3.0*c - 6.0*d,

  1003.                                             dist-1.0);

  1004. }

  1005. 

  1006. static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,

  1007.                              int srcW, int dstW, int filterAlign, int one, int flags,

  1008.                              SwsVector *srcFilter, SwsVector *dstFilter, double param[2])

  1009. {

  1010.     int i;

  1011.     int filterSize;

  1012.     int filter2Size;

  1013.     int minFilterSize;

  1014.     double *filter=NULL;

  1015.     double *filter2=NULL;

  1016.     int ret= -1;

  1017. #if defined(ARCH_X86)

  1018.     if (flags & SWS_CPU_CAPS_MMX)

  1019.         asm volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)

  1020. #endif

  1021. 

  1022.     // Note the +1 is for the MMXscaler which reads over the end

  1023.     *filterPos = av_malloc((dstW+1)*sizeof(int16_t));

  1024. 

  1025.     if (FFABS(xInc - 0x10000) <10) // unscaled

  1026.     {

  1027.         int i;

  1028.         filterSize= 1;

  1029.         filter= av_malloc(dstW*sizeof(double)*filterSize);

  1030.         for (i=0; i<dstW*filterSize; i++) filter[i]=0;

  1031. 

  1032.         for (i=0; i<dstW; i++)

  1033.         {

  1034.             filter[i*filterSize]=1;

  1035.             (*filterPos)[i]=i;

  1036.         }

  1037. 

  1038.     }

  1039.     else if (flags&SWS_POINT) // lame looking point sampling mode

  1040.     {

  1041.         int i;

  1042.         int xDstInSrc;

  1043.         filterSize= 1;

  1044.         filter= av_malloc(dstW*sizeof(double)*filterSize);

  1045. 

  1046.         xDstInSrc= xInc/2 - 0x8000;

  1047.         for (i=0; i<dstW; i++)

  1048.         {

  1049.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1050. 

  1051.             (*filterPos)[i]= xx;

  1052.             filter[i]= 1.0;

  1053.             xDstInSrc+= xInc;

  1054.         }

  1055.     }

  1056.     else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale

  1057.     {

  1058.         int i;

  1059.         int xDstInSrc;

  1060.         if      (flags&SWS_BICUBIC) filterSize= 4;

  1061.         else if (flags&SWS_X      ) filterSize= 4;

  1062.         else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA

  1063.         filter= av_malloc(dstW*sizeof(double)*filterSize);

  1064. 

  1065.         xDstInSrc= xInc/2 - 0x8000;

  1066.         for (i=0; i<dstW; i++)

  1067.         {

  1068.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1069.             int j;

  1070. 

  1071.             (*filterPos)[i]= xx;

  1072.                 //Bilinear upscale / linear interpolate / Area averaging

  1073.                 for (j=0; j<filterSize; j++)

  1074.                 {

  1075.                     double d= FFABS((xx<<16) - xDstInSrc)/(double)(1<<16);

  1076.                     double coeff= 1.0 - d;

  1077.                     if (coeff<0) coeff=0;

  1078.                     filter[i*filterSize + j]= coeff;

  1079.                     xx++;

  1080.                 }

  1081.             xDstInSrc+= xInc;

  1082.         }

  1083.     }

  1084.     else

  1085.     {

  1086.         double xDstInSrc;

  1087.         double sizeFactor, filterSizeInSrc;

  1088.         const double xInc1= (double)xInc / (double)(1<<16);

  1089. 

  1090.         if      (flags&SWS_BICUBIC)      sizeFactor=  4.0;

  1091.         else if (flags&SWS_X)            sizeFactor=  8.0;

  1092.         else if (flags&SWS_AREA)         sizeFactor=  1.0; //downscale only, for upscale it is bilinear

  1093.         else if (flags&SWS_GAUSS)        sizeFactor=  8.0;   // infinite ;)

  1094.         else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? 2.0*param[0] : 6.0;

  1095.         else if (flags&SWS_SINC)         sizeFactor= 20.0; // infinite ;)

  1096.         else if (flags&SWS_SPLINE)       sizeFactor= 20.0;  // infinite ;)

  1097.         else if (flags&SWS_BILINEAR)     sizeFactor=  2.0;

  1098.         else {

  1099.             sizeFactor= 0.0; //GCC warning killer

  1100.             assert(0);

  1101.         }

  1102. 

  1103.         if (xInc1 <= 1.0)       filterSizeInSrc= sizeFactor; // upscale

  1104.         else                    filterSizeInSrc= sizeFactor*srcW / (double)dstW;

  1105. 

  1106.         filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible

  1107.         if (filterSize > srcW-2) filterSize=srcW-2;

  1108. 

  1109.         filter= av_malloc(dstW*sizeof(double)*filterSize);

  1110. 

  1111.         xDstInSrc= xInc1 / 2.0 - 0.5;

  1112.         for (i=0; i<dstW; i++)

  1113.         {

  1114.             int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);

  1115.             int j;

  1116.             (*filterPos)[i]= xx;

  1117.             for (j=0; j<filterSize; j++)

  1118.             {

  1119.                 double d= FFABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;

  1120.                 double coeff;

  1121.                 if (flags & SWS_BICUBIC)

  1122.                 {

  1123.                     double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;

  1124.                     double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;

  1125. 

  1126.                     if (d<1.0)

  1127.                         coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;

  1128.                     else if (d<2.0)

  1129.                         coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;

  1130.                     else

  1131.                         coeff=0.0;

  1132.                 }

  1133. /*                else if (flags & SWS_X)

  1134.                 {

  1135.                     double p= param ? param*0.01 : 0.3;

  1136.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1137.                     coeff*= pow(2.0, - p*d*d);

  1138.                 }*/

  1139.                 else if (flags & SWS_X)

  1140.                 {

  1141.                     double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

  1142. 

  1143.                     if (d<1.0)

  1144.                         coeff = cos(d*PI);

  1145.                     else

  1146.                         coeff=-1.0;

  1147.                     if (coeff<0.0)      coeff= -pow(-coeff, A);

  1148.                     else                coeff=  pow( coeff, A);

  1149.                     coeff= coeff*0.5 + 0.5;

  1150.                 }

  1151.                 else if (flags & SWS_AREA)

  1152.                 {

  1153.                     double srcPixelSize= 1.0/xInc1;

  1154.                     if      (d + srcPixelSize/2 < 0.5) coeff= 1.0;

  1155.                     else if (d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5;

  1156.                     else coeff=0.0;

  1157.                 }

  1158.                 else if (flags & SWS_GAUSS)

  1159.                 {

  1160.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1161.                     coeff = pow(2.0, - p*d*d);

  1162.                 }

  1163.                 else if (flags & SWS_SINC)

  1164.                 {

  1165.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1166.                 }

  1167.                 else if (flags & SWS_LANCZOS)

  1168.                 {

  1169.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1170.                     coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;

  1171.                     if (d>p) coeff=0;

  1172.                 }

  1173.                 else if (flags & SWS_BILINEAR)

  1174.                 {

  1175.                     coeff= 1.0 - d;

  1176.                     if (coeff<0) coeff=0;

  1177.                 }

  1178.                 else if (flags & SWS_SPLINE)

  1179.                 {

  1180.                     double p=-2.196152422706632;

  1181.                     coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);

  1182.                 }

  1183.                 else {

  1184.                     coeff= 0.0; //GCC warning killer

  1185.                     assert(0);

  1186.                 }

  1187. 

  1188.                 filter[i*filterSize + j]= coeff;

  1189.                 xx++;

  1190.             }

  1191.             xDstInSrc+= xInc1;

  1192.         }

  1193.     }

  1194. 

  1195.     /* apply src & dst Filter to filter -> filter2

  1196.        av_free(filter);

  1197.     */

  1198.     assert(filterSize>0);

  1199.     filter2Size= filterSize;

  1200.     if (srcFilter) filter2Size+= srcFilter->length - 1;

  1201.     if (dstFilter) filter2Size+= dstFilter->length - 1;

  1202.     assert(filter2Size>0);

  1203.     filter2= av_malloc(filter2Size*dstW*sizeof(double));

  1204. 

  1205.     for (i=0; i<dstW; i++)

  1206.     {

  1207.         int j;

  1208.         SwsVector scaleFilter;

  1209.         SwsVector *outVec;

  1210. 

  1211.         scaleFilter.coeff= filter + i*filterSize;

  1212.         scaleFilter.length= filterSize;

  1213. 

  1214.         if (srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);

  1215.         else           outVec= &scaleFilter;

  1216. 

  1217.         assert(outVec->length == filter2Size);

  1218.         //FIXME dstFilter

  1219. 

  1220.         for (j=0; j<outVec->length; j++)

  1221.         {

  1222.             filter2[i*filter2Size + j]= outVec->coeff[j];

  1223.         }

  1224. 

  1225.         (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

  1226. 

  1227.         if (outVec != &scaleFilter) sws_freeVec(outVec);

  1228.     }

  1229.     av_freep(&filter);

  1230. 

  1231.     /* try to reduce the filter-size (step1 find size and shift left) */

  1232.     // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).

  1233.     minFilterSize= 0;

  1234.     for (i=dstW-1; i>=0; i--)

  1235.     {

  1236.         int min= filter2Size;

  1237.         int j;

  1238.         double cutOff=0.0;

  1239. 

  1240.         /* get rid off near zero elements on the left by shifting left */

  1241.         for (j=0; j<filter2Size; j++)

  1242.         {

  1243.             int k;

  1244.             cutOff += FFABS(filter2[i*filter2Size]);

  1245. 

  1246.             if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;

  1247. 

  1248.             /* preserve monotonicity because the core can't handle the filter otherwise */

  1249.             if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

  1250. 

  1251.             // Move filter coeffs left

  1252.             for (k=1; k<filter2Size; k++)

  1253.                 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];

  1254.             filter2[i*filter2Size + k - 1]= 0.0;

  1255.             (*filterPos)[i]++;

  1256.         }

  1257. 

  1258.         cutOff=0.0;

  1259.         /* count near zeros on the right */

  1260.         for (j=filter2Size-1; j>0; j--)

  1261.         {

  1262.             cutOff += FFABS(filter2[i*filter2Size + j]);

  1263. 

  1264.             if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;

  1265.             min--;

  1266.         }

  1267. 

  1268.         if (min>minFilterSize) minFilterSize= min;

  1269.     }

  1270. 

  1271.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1272.         // we can handle the special case 4,

  1273.         // so we don't want to go to the full 8

  1274.         if (minFilterSize < 5)

  1275.             filterAlign = 4;

  1276. 

  1277.         // we really don't want to waste our time

  1278.         // doing useless computation, so fall-back on

  1279.         // the scalar C code for very small filter.

  1280.         // vectorizing is worth it only if you have

  1281.         // decent-sized vector.

  1282.         if (minFilterSize < 3)

  1283.             filterAlign = 1;

  1284.     }

  1285. 

  1286.     if (flags & SWS_CPU_CAPS_MMX) {

  1287.         // special case for unscaled vertical filtering

  1288.         if (minFilterSize == 1 && filterAlign == 2)

  1289.             filterAlign= 1;

  1290.     }

  1291. 

  1292.     assert(minFilterSize > 0);

  1293.     filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));

  1294.     assert(filterSize > 0);

  1295.     filter= av_malloc(filterSize*dstW*sizeof(double));

  1296.     if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)

  1297.         goto error;

  1298.     *outFilterSize= filterSize;

  1299. 

  1300.     if (flags&SWS_PRINT_INFO)

  1301.         av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);

  1302.     /* try to reduce the filter-size (step2 reduce it) */

  1303.     for (i=0; i<dstW; i++)

  1304.     {

  1305.         int j;

  1306. 

  1307.         for (j=0; j<filterSize; j++)

  1308.         {

  1309.             if (j>=filter2Size) filter[i*filterSize + j]= 0.0;

  1310.             else               filter[i*filterSize + j]= filter2[i*filter2Size + j];

  1311.         }

  1312.     }

  1313. 

  1314. 

  1315.     //FIXME try to align filterpos if possible

  1316. 

  1317.     //fix borders

  1318.     for (i=0; i<dstW; i++)

  1319.     {

  1320.         int j;

  1321.         if ((*filterPos)[i] < 0)

  1322.         {

  1323.             // Move filter coeffs left to compensate for filterPos

  1324.             for (j=1; j<filterSize; j++)

  1325.             {

  1326.                 int left= FFMAX(j + (*filterPos)[i], 0);

  1327.                 filter[i*filterSize + left] += filter[i*filterSize + j];

  1328.                 filter[i*filterSize + j]=0;

  1329.             }

  1330.             (*filterPos)[i]= 0;

  1331.         }

  1332. 

  1333.         if ((*filterPos)[i] + filterSize > srcW)

  1334.         {

  1335.             int shift= (*filterPos)[i] + filterSize - srcW;

  1336.             // Move filter coeffs right to compensate for filterPos

  1337.             for (j=filterSize-2; j>=0; j--)

  1338.             {

  1339.                 int right= FFMIN(j + shift, filterSize-1);

  1340.                 filter[i*filterSize +right] += filter[i*filterSize +j];

  1341.                 filter[i*filterSize +j]=0;

  1342.             }

  1343.             (*filterPos)[i]= srcW - filterSize;

  1344.         }

  1345.     }

  1346. 

  1347.     // Note the +1 is for the MMXscaler which reads over the end

  1348.     /* align at 16 for AltiVec (needed by hScale_altivec_real) */

  1349.     *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));

  1350. 

  1351.     /* Normalize & Store in outFilter */

  1352.     for (i=0; i<dstW; i++)

  1353.     {

  1354.         int j;

  1355.         double error=0;

  1356.         double sum=0;

  1357.         double scale= one;

  1358. 

  1359.         for (j=0; j<filterSize; j++)

  1360.         {

  1361.             sum+= filter[i*filterSize + j];

  1362.         }

  1363.         scale/= sum;

  1364.         for (j=0; j<*outFilterSize; j++)

  1365.         {

  1366.             double v= filter[i*filterSize + j]*scale + error;

  1367.             int intV= floor(v + 0.5);

  1368.             (*outFilter)[i*(*outFilterSize) + j]= intV;

  1369.             error = v - intV;

  1370.         }

  1371.     }

  1372. 

  1373.     (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end

  1374.     for (i=0; i<*outFilterSize; i++)

  1375.     {

  1376.         int j= dstW*(*outFilterSize);

  1377.         (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

  1378.     }

  1379. 

  1380.     ret=0;

  1381. error:

  1382.     av_free(filter);

  1383.     av_free(filter2);

  1384.     return ret;

  1385. }

  1386. 

  1387. #ifdef COMPILE_MMX2

  1388. static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)

  1389. {

  1390.     uint8_t *fragmentA;

  1391.     long imm8OfPShufW1A;

  1392.     long imm8OfPShufW2A;

  1393.     long fragmentLengthA;

  1394.     uint8_t *fragmentB;

  1395.     long imm8OfPShufW1B;

  1396.     long imm8OfPShufW2B;

  1397.     long fragmentLengthB;

  1398.     int fragmentPos;

  1399. 

  1400.     int xpos, i;

  1401. 

  1402.     // create an optimized horizontal scaling routine

  1403. 

  1404.     //code fragment

  1405. 

  1406.     asm volatile(

  1407.         "jmp                         9f                 \n\t"

  1408.     // Begin

  1409.         "0:                                             \n\t"

  1410.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1411.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1412.         "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"

  1413.         "punpcklbw                %%mm7, %%mm1          \n\t"

  1414.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1415.         "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"

  1416.         "1:                                             \n\t"

  1417.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1418.         "2:                                             \n\t"

  1419.         "psubw                    %%mm1, %%mm0          \n\t"

  1420.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1421.         "pmullw                   %%mm3, %%mm0          \n\t"

  1422.         "psllw                       $7, %%mm1          \n\t"

  1423.         "paddw                    %%mm1, %%mm0          \n\t"

  1424. 

  1425.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1426. 

  1427.         "add                         $8, %%"REG_a"      \n\t"

  1428.     // End

  1429.         "9:                                             \n\t"

  1430. //        "int $3                                         \n\t"

  1431.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1432.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1433.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1434.         "dec                         %1                 \n\t"

  1435.         "dec                         %2                 \n\t"

  1436.         "sub                         %0, %1             \n\t"

  1437.         "sub                         %0, %2             \n\t"

  1438.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1439.         "sub                         %0, %3             \n\t"

  1440. 

  1441. 

  1442.         :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),

  1443.         "=r" (fragmentLengthA)

  1444.     );

  1445. 

  1446.     asm volatile(

  1447.         "jmp                         9f                 \n\t"

  1448.     // Begin

  1449.         "0:                                             \n\t"

  1450.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1451.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1452.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1453.         "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"

  1454.         "1:                                             \n\t"

  1455.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1456.         "2:                                             \n\t"

  1457.         "psubw                    %%mm1, %%mm0          \n\t"

  1458.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1459.         "pmullw                   %%mm3, %%mm0          \n\t"

  1460.         "psllw                       $7, %%mm1          \n\t"

  1461.         "paddw                    %%mm1, %%mm0          \n\t"

  1462. 

  1463.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1464. 

  1465.         "add                         $8, %%"REG_a"      \n\t"

  1466.     // End

  1467.         "9:                                             \n\t"

  1468. //        "int                       $3                   \n\t"

  1469.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1470.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1471.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1472.         "dec                         %1                 \n\t"

  1473.         "dec                         %2                 \n\t"

  1474.         "sub                         %0, %1             \n\t"

  1475.         "sub                         %0, %2             \n\t"

  1476.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1477.         "sub                         %0, %3             \n\t"

  1478. 

  1479. 

  1480.         :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),

  1481.         "=r" (fragmentLengthB)

  1482.     );

  1483. 

  1484.     xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers

  1485.     fragmentPos=0;

  1486. 

  1487.     for (i=0; i<dstW/numSplits; i++)

  1488.     {

  1489.         int xx=xpos>>16;

  1490. 

  1491.         if ((i&3) == 0)

  1492.         {

  1493.             int a=0;

  1494.             int b=((xpos+xInc)>>16) - xx;

  1495.             int c=((xpos+xInc*2)>>16) - xx;

  1496.             int d=((xpos+xInc*3)>>16) - xx;

  1497. 

  1498.             filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;

  1499.             filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;

  1500.             filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;

  1501.             filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;

  1502.             filterPos[i/2]= xx;

  1503. 

  1504.             if (d+1<4)

  1505.             {

  1506.                 int maxShift= 3-(d+1);

  1507.                 int shift=0;

  1508. 

  1509.                 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);

  1510. 

  1511.                 funnyCode[fragmentPos + imm8OfPShufW1B]=

  1512.                     (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);

  1513.                 funnyCode[fragmentPos + imm8OfPShufW2B]=

  1514.                     a | (b<<2) | (c<<4) | (d<<6);

  1515. 

  1516.                 if (i+3>=dstW) shift=maxShift; //avoid overread

  1517.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

  1518. 

  1519.                 if (shift && i>=shift)

  1520.                 {

  1521.                     funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;

  1522.                     funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;

  1523.                     filterPos[i/2]-=shift;

  1524.                 }

  1525. 

  1526.                 fragmentPos+= fragmentLengthB;

  1527.             }

  1528.             else

  1529.             {

  1530.                 int maxShift= 3-d;

  1531.                 int shift=0;

  1532. 

  1533.                 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);

  1534. 

  1535.                 funnyCode[fragmentPos + imm8OfPShufW1A]=

  1536.                 funnyCode[fragmentPos + imm8OfPShufW2A]=

  1537.                     a | (b<<2) | (c<<4) | (d<<6);

  1538. 

  1539.                 if (i+4>=dstW) shift=maxShift; //avoid overread

  1540.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align

  1541. 

  1542.                 if (shift && i>=shift)

  1543.                 {

  1544.                     funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;

  1545.                     funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;

  1546.                     filterPos[i/2]-=shift;

  1547.                 }

  1548. 

  1549.                 fragmentPos+= fragmentLengthA;

  1550.             }

  1551. 

  1552.             funnyCode[fragmentPos]= RET;

  1553.         }

  1554.         xpos+=xInc;

  1555.     }

  1556.     filterPos[i/2]= xpos>>16; // needed to jump to the next part

  1557. }

  1558. #endif /* COMPILE_MMX2 */

  1559. 

  1560. static void globalInit(void){

  1561.     // generating tables:

  1562.     int i;

  1563.     for (i=0; i<768; i++){

  1564.         int c= av_clip_uint8(i-256);

  1565.         clip_table[i]=c;

  1566.     }

  1567. }

  1568. 

  1569. static SwsFunc getSwsFunc(int flags){

  1570. 

  1571. #if defined(RUNTIME_CPUDETECT) && defined (CONFIG_GPL)

  1572. #if defined(ARCH_X86)

  1573.     // ordered per speed fastest first

  1574.     if (flags & SWS_CPU_CAPS_MMX2)

  1575.         return swScale_MMX2;

  1576.     else if (flags & SWS_CPU_CAPS_3DNOW)

  1577.         return swScale_3DNow;

  1578.     else if (flags & SWS_CPU_CAPS_MMX)

  1579.         return swScale_MMX;

  1580.     else

  1581.         return swScale_C;

  1582. 

  1583. #else

  1584. #ifdef ARCH_POWERPC

  1585.     if (flags & SWS_CPU_CAPS_ALTIVEC)

  1586.         return swScale_altivec;

  1587.     else

  1588.         return swScale_C;

  1589. #endif

  1590.     return swScale_C;

  1591. #endif /* defined(ARCH_X86) */

  1592. #else //RUNTIME_CPUDETECT

  1593. #ifdef HAVE_MMX2

  1594.     return swScale_MMX2;

  1595. #elif defined (HAVE_3DNOW)

  1596.     return swScale_3DNow;

  1597. #elif defined (HAVE_MMX)

  1598.     return swScale_MMX;

  1599. #elif defined (HAVE_ALTIVEC)

  1600.     return swScale_altivec;

  1601. #else

  1602.     return swScale_C;

  1603. #endif

  1604. #endif //!RUNTIME_CPUDETECT

  1605. }

  1606. 

  1607. static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1608.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1609.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1610.     /* Copy Y plane */

  1611.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1612.         memcpy(dst, src[0], srcSliceH*dstStride[0]);

  1613.     else

  1614.     {

  1615.         int i;

  1616.         uint8_t *srcPtr= src[0];

  1617.         uint8_t *dstPtr= dst;

  1618.         for (i=0; i<srcSliceH; i++)

  1619.         {

  1620.             memcpy(dstPtr, srcPtr, c->srcW);

  1621.             srcPtr+= srcStride[0];

  1622.             dstPtr+= dstStride[0];

  1623.         }

  1624.     }

  1625.     dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1626.     if (c->dstFormat == PIX_FMT_NV12)

  1627.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  1628.     else

  1629.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  1630. 

  1631.     return srcSliceH;

  1632. }

  1633. 

  1634. static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1635.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1636.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1637. 

  1638.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1639. 

  1640.     return srcSliceH;

  1641. }

  1642. 

  1643. static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1644.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1645.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1646. 

  1647.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1648. 

  1649.     return srcSliceH;

  1650. }

  1651. 

  1652. static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1653.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[]){

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

  1655. 

  1656.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1657. 

  1658.     return srcSliceH;

  1659. }

  1660. 

  1661. static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1662.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1663.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1664. 

  1665.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1666. 

  1667.     return srcSliceH;

  1668. }

  1669. 

  1670. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  1671. static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1672.                           int srcSliceH, uint8_t* dst[], int dstStride[]){

  1673.     const int srcFormat= c->srcFormat;

  1674.     const int dstFormat= c->dstFormat;

  1675.     const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;

  1676.     const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;

  1677.     const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1678.     const int dstId= fmt_depth(dstFormat) >> 2;

  1679.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  1680. 

  1681.     /* BGR -> BGR */

  1682.     if (  (isBGR(srcFormat) && isBGR(dstFormat))

  1683.        || (isRGB(srcFormat) && isRGB(dstFormat))){

  1684.         switch(srcId | (dstId<<4)){

  1685.         case 0x34: conv= rgb16to15; break;

  1686.         case 0x36: conv= rgb24to15; break;

  1687.         case 0x38: conv= rgb32to15; break;

  1688.         case 0x43: conv= rgb15to16; break;

  1689.         case 0x46: conv= rgb24to16; break;

  1690.         case 0x48: conv= rgb32to16; break;

  1691.         case 0x63: conv= rgb15to24; break;

  1692.         case 0x64: conv= rgb16to24; break;

  1693.         case 0x68: conv= rgb32to24; break;

  1694.         case 0x83: conv= rgb15to32; break;

  1695.         case 0x84: conv= rgb16to32; break;

  1696.         case 0x86: conv= rgb24to32; break;

  1697.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1698.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1699.         }

  1700.     }else if (  (isBGR(srcFormat) && isRGB(dstFormat))

  1701.              || (isRGB(srcFormat) && isBGR(dstFormat))){

  1702.         switch(srcId | (dstId<<4)){

  1703.         case 0x33: conv= rgb15tobgr15; break;

  1704.         case 0x34: conv= rgb16tobgr15; break;

  1705.         case 0x36: conv= rgb24tobgr15; break;

  1706.         case 0x38: conv= rgb32tobgr15; break;

  1707.         case 0x43: conv= rgb15tobgr16; break;

  1708.         case 0x44: conv= rgb16tobgr16; break;

  1709.         case 0x46: conv= rgb24tobgr16; break;

  1710.         case 0x48: conv= rgb32tobgr16; break;

  1711.         case 0x63: conv= rgb15tobgr24; break;

  1712.         case 0x64: conv= rgb16tobgr24; break;

  1713.         case 0x66: conv= rgb24tobgr24; break;

  1714.         case 0x68: conv= rgb32tobgr24; break;

  1715.         case 0x83: conv= rgb15tobgr32; break;

  1716.         case 0x84: conv= rgb16tobgr32; break;

  1717.         case 0x86: conv= rgb24tobgr32; break;

  1718.         case 0x88: conv= rgb32tobgr32; break;

  1719.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1720.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1721.         }

  1722.     }else{

  1723.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1724.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1725.     }

  1726. 

  1727.     if(conv)

  1728.     {

  1729.         uint8_t *srcPtr= src[0];

  1730.         if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)

  1731.             srcPtr += ALT32_CORR;

  1732. 

  1733.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  1734.             conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  1735.         else

  1736.         {

  1737.             int i;

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

  1739. 

  1740.             for (i=0; i<srcSliceH; i++)

  1741.             {

  1742.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  1743.                 srcPtr+= srcStride[0];

  1744.                 dstPtr+= dstStride[0];

  1745.             }

  1746.         }

  1747.     }

  1748.     return srcSliceH;

  1749. }

  1750. 

  1751. static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1752.                               int srcSliceH, uint8_t* dst[], int dstStride[]){

  1753. 

  1754.     rgb24toyv12(

  1755.         src[0],

  1756.         dst[0]+ srcSliceY    *dstStride[0],

  1757.         dst[1]+(srcSliceY>>1)*dstStride[1],

  1758.         dst[2]+(srcSliceY>>1)*dstStride[2],

  1759.         c->srcW, srcSliceH,

  1760.         dstStride[0], dstStride[1], srcStride[0]);

  1761.     return srcSliceH;

  1762. }

  1763. 

  1764. static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1765.                              int srcSliceH, uint8_t* dst[], int dstStride[]){

  1766.     int i;

  1767. 

  1768.     /* copy Y */

  1769.     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)

  1770.         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);

  1771.     else{

  1772.         uint8_t *srcPtr= src[0];

  1773.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1774. 

  1775.         for (i=0; i<srcSliceH; i++)

  1776.         {

  1777.             memcpy(dstPtr, srcPtr, c->srcW);

  1778.             srcPtr+= srcStride[0];

  1779.             dstPtr+= dstStride[0];

  1780.         }

  1781.     }

  1782. 

  1783.     if (c->dstFormat==PIX_FMT_YUV420P){

  1784.         planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);

  1785.         planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);

  1786.     }else{

  1787.         planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);

  1788.         planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);

  1789.     }

  1790.     return srcSliceH;

  1791. }

  1792. 

  1793. /* unscaled copy like stuff (assumes nearly identical formats) */

  1794. static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1795.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  1796. {

  1797.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1798.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  1799.     else

  1800.     {

  1801.         int i;

  1802.         uint8_t *srcPtr= src[0];

  1803.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1804.         int length=0;

  1805. 

  1806.         /* universal length finder */

  1807.         while(length+c->srcW <= FFABS(dstStride[0])

  1808.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  1809.         assert(length!=0);

  1810. 

  1811.         for (i=0; i<srcSliceH; i++)

  1812.         {

  1813.             memcpy(dstPtr, srcPtr, length);

  1814.             srcPtr+= srcStride[0];

  1815.             dstPtr+= dstStride[0];

  1816.         }

  1817.     }

  1818.     return srcSliceH;

  1819. }

  1820. 

  1821. static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1822.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  1823. {

  1824.     int plane;

  1825.     for (plane=0; plane<3; plane++)

  1826.     {

  1827.         int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  1828.         int y=      plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  1829.         int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  1830. 

  1831.         if ((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)

  1832.         {

  1833.             if (!isGray(c->dstFormat))

  1834.                 memset(dst[plane], 128, dstStride[plane]*height);

  1835.         }

  1836.         else

  1837.         {

  1838.             if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)

  1839.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);

  1840.             else

  1841.             {

  1842.                 int i;

  1843.                 uint8_t *srcPtr= src[plane];

  1844.                 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  1845.                 for (i=0; i<height; i++)

  1846.                 {

  1847.                     memcpy(dstPtr, srcPtr, length);

  1848.                     srcPtr+= srcStride[plane];

  1849.                     dstPtr+= dstStride[plane];

  1850.                 }

  1851.             }

  1852.         }

  1853.     }

  1854.     return srcSliceH;

  1855. }

  1856. 

  1857. static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1858.                         int srcSliceH, uint8_t* dst[], int dstStride[]){

  1859. 

  1860.     int length= c->srcW;

  1861.     int y=      srcSliceY;

  1862.     int height= srcSliceH;

  1863.     int i, j;

  1864.     uint8_t *srcPtr= src[0];

  1865.     uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  1866. 

  1867.     if (!isGray(c->dstFormat)){

  1868.         int height= -((-srcSliceH)>>c->chrDstVSubSample);

  1869.         memset(dst[1], 128, dstStride[1]*height);

  1870.         memset(dst[2], 128, dstStride[2]*height);

  1871.     }

  1872.     if (c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;

  1873.     for (i=0; i<height; i++)

  1874.     {

  1875.         for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  1876.         srcPtr+= srcStride[0];

  1877.         dstPtr+= dstStride[0];

  1878.     }

  1879.     return srcSliceH;

  1880. }

  1881. 

  1882. static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1883.                         int srcSliceH, uint8_t* dst[], int dstStride[]){

  1884. 

  1885.     int length= c->srcW;

  1886.     int y=      srcSliceY;

  1887.     int height= srcSliceH;

  1888.     int i, j;

  1889.     uint8_t *srcPtr= src[0];

  1890.     uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  1891.     for (i=0; i<height; i++)

  1892.     {

  1893.         for (j=0; j<length; j++)

  1894.         {

  1895.             dstPtr[j<<1] = srcPtr[j];

  1896.             dstPtr[(j<<1)+1] = srcPtr[j];

  1897.         }

  1898.         srcPtr+= srcStride[0];

  1899.         dstPtr+= dstStride[0];

  1900.     }

  1901.     return srcSliceH;

  1902. }

  1903. 

  1904. static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1905.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  1906. 

  1907.     int length= c->srcW;

  1908.     int y=      srcSliceY;

  1909.     int height= srcSliceH;

  1910.     int i, j;

  1911.     uint16_t *srcPtr= (uint16_t*)src[0];

  1912.     uint16_t *dstPtr= (uint16_t*)(dst[0] + dstStride[0]*y/2);

  1913.     for (i=0; i<height; i++)

  1914.     {

  1915.         for (j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);

  1916.         srcPtr+= srcStride[0]/2;

  1917.         dstPtr+= dstStride[0]/2;

  1918.     }

  1919.     return srcSliceH;

  1920. }

  1921. 

  1922. 

  1923. static void getSubSampleFactors(int *h, int *v, int format){

  1924.     switch(format){

  1925.     case PIX_FMT_UYVY422:

  1926.     case PIX_FMT_YUYV422:

  1927.         *h=1;

  1928.         *v=0;

  1929.         break;

  1930.     case PIX_FMT_YUV420P:

  1931.     case PIX_FMT_YUVA420P:

  1932.     case PIX_FMT_GRAY16BE:

  1933.     case PIX_FMT_GRAY16LE:

  1934.     case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented

  1935.     case PIX_FMT_NV12:

  1936.     case PIX_FMT_NV21:

  1937.         *h=1;

  1938.         *v=1;

  1939.         break;

  1940.     case PIX_FMT_YUV440P:

  1941.         *h=0;

  1942.         *v=1;

  1943.         break;

  1944.     case PIX_FMT_YUV410P:

  1945.         *h=2;

  1946.         *v=2;

  1947.         break;

  1948.     case PIX_FMT_YUV444P:

  1949.         *h=0;

  1950.         *v=0;

  1951.         break;

  1952.     case PIX_FMT_YUV422P:

  1953.         *h=1;

  1954.         *v=0;

  1955.         break;

  1956.     case PIX_FMT_YUV411P:

  1957.         *h=2;

  1958.         *v=0;

  1959.         break;

  1960.     default:

  1961.         *h=0;

  1962.         *v=0;

  1963.         break;

  1964.     }

  1965. }

  1966. 

  1967. static uint16_t roundToInt16(int64_t f){

  1968.     int r= (f + (1<<15))>>16;

  1969.          if (r<-0x7FFF) return 0x8000;

  1970.     else if (r> 0x7FFF) return 0x7FFF;

  1971.     else                return r;

  1972. }

  1973. 

  1974. /**

  1975.  * @param inv_table the yuv2rgb coeffs, normally Inverse_Table_6_9[x]

  1976.  * @param fullRange if 1 then the luma range is 0..255 if 0 it is 16..235

  1977.  * @return -1 if not supported

  1978.  */

  1979. int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){

  1980.     int64_t crv =  inv_table[0];

  1981.     int64_t cbu =  inv_table[1];

  1982.     int64_t cgu = -inv_table[2];

  1983.     int64_t cgv = -inv_table[3];

  1984.     int64_t cy  = 1<<16;

  1985.     int64_t oy  = 0;

  1986. 

  1987.     memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);

  1988.     memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);

  1989. 

  1990.     c->brightness= brightness;

  1991.     c->contrast  = contrast;

  1992.     c->saturation= saturation;

  1993.     c->srcRange  = srcRange;

  1994.     c->dstRange  = dstRange;

  1995.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return 0;

  1996. 

  1997.     c->uOffset=   0x0400040004000400LL;

  1998.     c->vOffset=   0x0400040004000400LL;

  1999. 

  2000.     if (!srcRange){

  2001.         cy= (cy*255) / 219;

  2002.         oy= 16<<16;

  2003.     }else{

  2004.         crv= (crv*224) / 255;

  2005.         cbu= (cbu*224) / 255;

  2006.         cgu= (cgu*224) / 255;

  2007.         cgv= (cgv*224) / 255;

  2008.     }

  2009. 

  2010.     cy = (cy *contrast             )>>16;

  2011.     crv= (crv*contrast * saturation)>>32;

  2012.     cbu= (cbu*contrast * saturation)>>32;

  2013.     cgu= (cgu*contrast * saturation)>>32;

  2014.     cgv= (cgv*contrast * saturation)>>32;

  2015. 

  2016.     oy -= 256*brightness;

  2017. 

  2018.     c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;

  2019.     c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;

  2020.     c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;

  2021.     c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;

  2022.     c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;

  2023.     c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;

  2024. 

  2025.     yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);

  2026.     //FIXME factorize

  2027. 

  2028. #ifdef COMPILE_ALTIVEC

  2029.     if (c->flags & SWS_CPU_CAPS_ALTIVEC)

  2030.         yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);

  2031. #endif

  2032.     return 0;

  2033. }

  2034. 

  2035. /**

  2036.  * @return -1 if not supported

  2037.  */

  2038. int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){

  2039.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  2040. 

  2041.     *inv_table = c->srcColorspaceTable;

  2042.     *table     = c->dstColorspaceTable;

  2043.     *srcRange  = c->srcRange;

  2044.     *dstRange  = c->dstRange;

  2045.     *brightness= c->brightness;

  2046.     *contrast  = c->contrast;

  2047.     *saturation= c->saturation;

  2048. 

  2049.     return 0;

  2050. }

  2051. 

  2052. static int handle_jpeg(int *format)

  2053. {

  2054.     switch (*format) {

  2055.         case PIX_FMT_YUVJ420P:

  2056.             *format = PIX_FMT_YUV420P;

  2057.             return 1;

  2058.         case PIX_FMT_YUVJ422P:

  2059.             *format = PIX_FMT_YUV422P;

  2060.             return 1;

  2061.         case PIX_FMT_YUVJ444P:

  2062.             *format = PIX_FMT_YUV444P;

  2063.             return 1;

  2064.         case PIX_FMT_YUVJ440P:

  2065.             *format = PIX_FMT_YUV440P;

  2066.             return 1;

  2067.         default:

  2068.             return 0;

  2069.     }

  2070. }

  2071. 

  2072. SwsContext *sws_getContext(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat, int flags,

  2073.                            SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){

  2074. 

  2075.     SwsContext *c;

  2076.     int i;

  2077.     int usesVFilter, usesHFilter;

  2078.     int unscaled, needsDither;

  2079.     int srcRange, dstRange;

  2080.     SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};

  2081. #if defined(ARCH_X86)

  2082.     if (flags & SWS_CPU_CAPS_MMX)

  2083.         asm volatile("emms\n\t"::: "memory");

  2084. #endif

  2085. 

  2086. #if !defined(RUNTIME_CPUDETECT) || !defined (CONFIG_GPL) //ensure that the flags match the compiled variant if cpudetect is off

  2087.     flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);

  2088. #ifdef HAVE_MMX2

  2089.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  2090. #elif defined (HAVE_3DNOW)

  2091.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  2092. #elif defined (HAVE_MMX)

  2093.     flags |= SWS_CPU_CAPS_MMX;

  2094. #elif defined (HAVE_ALTIVEC)

  2095.     flags |= SWS_CPU_CAPS_ALTIVEC;

  2096. #elif defined (ARCH_BFIN)

  2097.     flags |= SWS_CPU_CAPS_BFIN;

  2098. #endif

  2099. #endif /* RUNTIME_CPUDETECT */

  2100.     if (clip_table[512] != 255) globalInit();

  2101.     if (!rgb15to16) sws_rgb2rgb_init(flags);

  2102. 

  2103.     unscaled = (srcW == dstW && srcH == dstH);

  2104.     needsDither= (isBGR(dstFormat) || isRGB(dstFormat))

  2105.         && (fmt_depth(dstFormat))<24

  2106.         && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

  2107. 

  2108.     srcRange = handle_jpeg(&srcFormat);

  2109.     dstRange = handle_jpeg(&dstFormat);

  2110. 

  2111.     if (!isSupportedIn(srcFormat))

  2112.     {

  2113.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));

  2114.         return NULL;

  2115.     }

  2116.     if (!isSupportedOut(dstFormat))

  2117.     {

  2118.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));

  2119.         return NULL;

  2120.     }

  2121. 

  2122.     i= flags & ( SWS_POINT

  2123.                 |SWS_AREA

  2124.                 |SWS_BILINEAR

  2125.                 |SWS_FAST_BILINEAR

  2126.                 |SWS_BICUBIC

  2127.                 |SWS_X

  2128.                 |SWS_GAUSS

  2129.                 |SWS_LANCZOS

  2130.                 |SWS_SINC

  2131.                 |SWS_SPLINE

  2132.                 |SWS_BICUBLIN);

  2133.     if(!i || (i & (i-1)))

  2134.     {

  2135.         av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be choosen\n");

  2136.         return NULL;

  2137.     }

  2138. 

  2139. 

  2140.     /* sanity check */

  2141.     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

  2142.     {

  2143.         av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",

  2144.                srcW, srcH, dstW, dstH);

  2145.         return NULL;

  2146.     }

  2147.     if(srcW > VOFW || dstW > VOFW){

  2148.         av_log(NULL, AV_LOG_ERROR, "swScaler: Compile time max width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");

  2149.         return NULL;

  2150.     }

  2151. 

  2152.     if (!dstFilter) dstFilter= &dummyFilter;

  2153.     if (!srcFilter) srcFilter= &dummyFilter;

  2154. 

  2155.     c= av_mallocz(sizeof(SwsContext));

  2156. 

  2157.     c->av_class = &sws_context_class;

  2158.     c->srcW= srcW;

  2159.     c->srcH= srcH;

  2160.     c->dstW= dstW;

  2161.     c->dstH= dstH;

  2162.     c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;

  2163.     c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;

  2164.     c->flags= flags;

  2165.     c->dstFormat= dstFormat;

  2166.     c->srcFormat= srcFormat;

  2167.     c->vRounder= 4* 0x0001000100010001ULL;

  2168. 

  2169.     usesHFilter= usesVFilter= 0;

  2170.     if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;

  2171.     if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;

  2172.     if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;

  2173.     if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;

  2174.     if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;

  2175.     if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;

  2176.     if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;

  2177.     if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;

  2178. 

  2179.     getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);

  2180.     getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

  2181. 

  2182.     // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation

  2183.     if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

  2184. 

  2185.     // drop some chroma lines if the user wants it

  2186.     c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;

  2187.     c->chrSrcVSubSample+= c->vChrDrop;

  2188. 

  2189.     // drop every 2. pixel for chroma calculation unless user wants full chroma

  2190.     if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)

  2191.       && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8

  2192.       && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4

  2193.       && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE)

  2194.         c->chrSrcHSubSample=1;

  2195. 

  2196.     if (param){

  2197.         c->param[0] = param[0];

  2198.         c->param[1] = param[1];

  2199.     }else{

  2200.         c->param[0] =

  2201.         c->param[1] = SWS_PARAM_DEFAULT;

  2202.     }

  2203. 

  2204.     c->chrIntHSubSample= c->chrDstHSubSample;

  2205.     c->chrIntVSubSample= c->chrSrcVSubSample;

  2206. 

  2207.     // Note the -((-x)>>y) is so that we always round toward +inf.

  2208.     c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);

  2209.     c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);

  2210.     c->chrDstW= -((-dstW) >> c->chrDstHSubSample);

  2211.     c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

  2212. 

  2213.     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);

  2214. 

  2215.     /* unscaled special Cases */

  2216.     if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))

  2217.     {

  2218.         /* yv12_to_nv12 */

  2219.         if (srcFormat == PIX_FMT_YUV420P && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))

  2220.         {

  2221.             c->swScale= PlanarToNV12Wrapper;

  2222.         }

  2223. #ifdef CONFIG_GPL

  2224.         /* yuv2bgr */

  2225.         if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat)))

  2226.         {

  2227.             c->swScale= yuv2rgb_get_func_ptr(c);

  2228.         }

  2229. #endif

  2230. 

  2231.         if (srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P)

  2232.         {

  2233.             c->swScale= yvu9toyv12Wrapper;

  2234.         }

  2235. 

  2236.         /* bgr24toYV12 */

  2237.         if (srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P)

  2238.             c->swScale= bgr24toyv12Wrapper;

  2239. 

  2240.         /* rgb/bgr -> rgb/bgr (no dither needed forms) */

  2241.         if (  (isBGR(srcFormat) || isRGB(srcFormat))

  2242.            && (isBGR(dstFormat) || isRGB(dstFormat))

  2243.            && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  2244.            && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  2245.            && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  2246.            && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  2247.            && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  2248.            && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  2249.            && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  2250.                                              && dstFormat != PIX_FMT_RGB32_1

  2251.                                              && dstFormat != PIX_FMT_BGR32_1

  2252.            && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2253.              c->swScale= rgb2rgbWrapper;

  2254. 

  2255.         if (srcFormat == PIX_FMT_YUV422P)

  2256.         {

  2257.             if (dstFormat == PIX_FMT_YUYV422)

  2258.                 c->swScale= YUV422PToYuy2Wrapper;

  2259.             else if (dstFormat == PIX_FMT_UYVY422)

  2260.                 c->swScale= YUV422PToUyvyWrapper;

  2261.         }

  2262. 

  2263.         /* LQ converters if -sws 0 or -sws 4*/

  2264.         if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){

  2265.             /* yv12_to_yuy2 */

  2266.             if (srcFormat == PIX_FMT_YUV420P)

  2267.             {

  2268.                 if (dstFormat == PIX_FMT_YUYV422)

  2269.                     c->swScale= PlanarToYuy2Wrapper;

  2270.                 else if (dstFormat == PIX_FMT_UYVY422)

  2271.                     c->swScale= PlanarToUyvyWrapper;

  2272.             }

  2273.         }

  2274. 

  2275. #ifdef COMPILE_ALTIVEC

  2276.         if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  2277.             srcFormat == PIX_FMT_YUV420P) {

  2278.           // unscaled YV12 -> packed YUV, we want speed

  2279.           if (dstFormat == PIX_FMT_YUYV422)

  2280.               c->swScale= yv12toyuy2_unscaled_altivec;

  2281.           else if (dstFormat == PIX_FMT_UYVY422)

  2282.               c->swScale= yv12touyvy_unscaled_altivec;

  2283.         }

  2284. #endif

  2285. 

  2286.         /* simple copy */

  2287.         if (  srcFormat == dstFormat

  2288.             || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  2289.             || (isPlanarYUV(dstFormat) && isGray(srcFormat)))

  2290.         {

  2291.             if (isPacked(c->srcFormat))

  2292.                 c->swScale= packedCopy;

  2293.             else /* Planar YUV or gray */

  2294.                 c->swScale= planarCopy;

  2295.         }

  2296. 

  2297.         /* gray16{le,be} conversions */

  2298.         if (isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))

  2299.         {

  2300.             c->swScale= gray16togray;

  2301.         }

  2302.         if ((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))

  2303.         {

  2304.             c->swScale= graytogray16;

  2305.         }

  2306.         if (srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))

  2307.         {

  2308.             c->swScale= gray16swap;

  2309.         }

  2310. 

  2311. #ifdef ARCH_BFIN

  2312.         if (flags & SWS_CPU_CAPS_BFIN)

  2313.             ff_bfin_get_unscaled_swscale (c);

  2314. #endif

  2315. 

  2316.         if (c->swScale){

  2317.             if (flags&SWS_PRINT_INFO)

  2318.                 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",

  2319.                                 sws_format_name(srcFormat), sws_format_name(dstFormat));

  2320.             return c;

  2321.         }

  2322.     }

  2323. 

  2324.     if (flags & SWS_CPU_CAPS_MMX2)

  2325.     {

  2326.         c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;

  2327.         if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))

  2328.         {

  2329.             if (flags&SWS_PRINT_INFO)

  2330.                 av_log(c, AV_LOG_INFO, "output Width is not a multiple of 32 -> no MMX2 scaler\n");

  2331.         }

  2332.         if (usesHFilter) c->canMMX2BeUsed=0;

  2333.     }

  2334.     else

  2335.         c->canMMX2BeUsed=0;

  2336. 

  2337.     c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;

  2338.     c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

  2339. 

  2340.     // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst

  2341.     // but only for the FAST_BILINEAR mode otherwise do correct scaling

  2342.     // n-2 is the last chrominance sample available

  2343.     // this is not perfect, but no one should notice the difference, the more correct variant

  2344.     // would be like the vertical one, but that would require some special code for the

  2345.     // first and last pixel

  2346.     if (flags&SWS_FAST_BILINEAR)

  2347.     {

  2348.         if (c->canMMX2BeUsed)

  2349.         {

  2350.             c->lumXInc+= 20;

  2351.             c->chrXInc+= 20;

  2352.         }

  2353.         //we don't use the x86asm scaler if mmx is available

  2354.         else if (flags & SWS_CPU_CAPS_MMX)

  2355.         {

  2356.             c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;

  2357.             c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;

  2358.         }

  2359.     }

  2360. 

  2361.     /* precalculate horizontal scaler filter coefficients */

  2362.     {

  2363.         const int filterAlign=

  2364.             (flags & SWS_CPU_CAPS_MMX) ? 4 :

  2365.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2366.             1;

  2367. 

  2368.         initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,

  2369.                    srcW      ,       dstW, filterAlign, 1<<14,

  2370.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2371.                    srcFilter->lumH, dstFilter->lumH, c->param);

  2372.         initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,

  2373.                    c->chrSrcW, c->chrDstW, filterAlign, 1<<14,

  2374.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2375.                    srcFilter->chrH, dstFilter->chrH, c->param);

  2376. 

  2377. #define MAX_FUNNY_CODE_SIZE 10000

  2378. #if defined(COMPILE_MMX2)

  2379. // can't downscale !!!

  2380.         if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))

  2381.         {

  2382. #ifdef MAP_ANONYMOUS

  2383.             c->funnyYCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2384.             c->funnyUVCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2385. #else

  2386.             c->funnyYCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2387.             c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2388. #endif

  2389. 

  2390.             c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));

  2391.             c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));

  2392.             c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));

  2393.             c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));

  2394. 

  2395.             initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);

  2396.             initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);

  2397.         }

  2398. #endif /* defined(COMPILE_MMX2) */

  2399.     } // Init Horizontal stuff

  2400. 

  2401. 

  2402. 

  2403.     /* precalculate vertical scaler filter coefficients */

  2404.     {

  2405.         const int filterAlign=

  2406.             (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :

  2407.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2408.             1;

  2409. 

  2410.         initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,

  2411.                    srcH      ,        dstH, filterAlign, (1<<12)-4,

  2412.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2413.                    srcFilter->lumV, dstFilter->lumV, c->param);

  2414.         initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,

  2415.                    c->chrSrcH, c->chrDstH, filterAlign, (1<<12)-4,

  2416.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2417.                    srcFilter->chrV, dstFilter->chrV, c->param);

  2418. 

  2419. #ifdef HAVE_ALTIVEC

  2420.         c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);

  2421.         c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);

  2422. 

  2423.         for (i=0;i<c->vLumFilterSize*c->dstH;i++) {

  2424.             int j;

  2425.             short *p = (short *)&c->vYCoeffsBank[i];

  2426.             for (j=0;j<8;j++)

  2427.                 p[j] = c->vLumFilter[i];

  2428.         }

  2429. 

  2430.         for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {

  2431.             int j;

  2432.             short *p = (short *)&c->vCCoeffsBank[i];

  2433.             for (j=0;j<8;j++)

  2434.                 p[j] = c->vChrFilter[i];

  2435.         }

  2436. #endif

  2437.     }

  2438. 

  2439.     // Calculate Buffer Sizes so that they won't run out while handling these damn slices

  2440.     c->vLumBufSize= c->vLumFilterSize;

  2441.     c->vChrBufSize= c->vChrFilterSize;

  2442.     for (i=0; i<dstH; i++)

  2443.     {

  2444.         int chrI= i*c->chrDstH / dstH;

  2445.         int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,

  2446.                            ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

  2447. 

  2448.         nextSlice>>= c->chrSrcVSubSample;

  2449.         nextSlice<<= c->chrSrcVSubSample;

  2450.         if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)

  2451.             c->vLumBufSize= nextSlice - c->vLumFilterPos[i];

  2452.         if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))

  2453.             c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];

  2454.     }

  2455. 

  2456.     // allocate pixbufs (we use dynamic allocation because otherwise we would need to

  2457.     c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));

  2458.     c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));

  2459.     //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)

  2460.     /* align at 16 bytes for AltiVec */

  2461.     for (i=0; i<c->vLumBufSize; i++)

  2462.         c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);

  2463.     for (i=0; i<c->vChrBufSize; i++)

  2464.         c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);

  2465. 

  2466.     //try to avoid drawing green stuff between the right end and the stride end

  2467.     for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);

  2468. 

  2469.     assert(2*VOFW == VOF);

  2470. 

  2471.     assert(c->chrDstH <= dstH);

  2472. 

  2473.     if (flags&SWS_PRINT_INFO)

  2474.     {

  2475. #ifdef DITHER1XBPP

  2476.         const char *dither= " dithered";

  2477. #else

  2478.         const char *dither= "";

  2479. #endif

  2480.         if (flags&SWS_FAST_BILINEAR)

  2481.             av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");

  2482.         else if (flags&SWS_BILINEAR)

  2483.             av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");

  2484.         else if (flags&SWS_BICUBIC)

  2485.             av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");

  2486.         else if (flags&SWS_X)

  2487.             av_log(c, AV_LOG_INFO, "Experimental scaler, ");

  2488.         else if (flags&SWS_POINT)

  2489.             av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");

  2490.         else if (flags&SWS_AREA)

  2491.             av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");

  2492.         else if (flags&SWS_BICUBLIN)

  2493.             av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");

  2494.         else if (flags&SWS_GAUSS)

  2495.             av_log(c, AV_LOG_INFO, "Gaussian scaler, ");

  2496.         else if (flags&SWS_SINC)

  2497.             av_log(c, AV_LOG_INFO, "Sinc scaler, ");

  2498.         else if (flags&SWS_LANCZOS)

  2499.             av_log(c, AV_LOG_INFO, "Lanczos scaler, ");

  2500.         else if (flags&SWS_SPLINE)

  2501.             av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");

  2502.         else

  2503.             av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");

  2504. 

  2505.         if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)

  2506.             av_log(c, AV_LOG_INFO, "from %s to%s %s ",

  2507.                    sws_format_name(srcFormat), dither, sws_format_name(dstFormat));

  2508.         else

  2509.             av_log(c, AV_LOG_INFO, "from %s to %s ",

  2510.                    sws_format_name(srcFormat), sws_format_name(dstFormat));

  2511. 

  2512.         if (flags & SWS_CPU_CAPS_MMX2)

  2513.             av_log(c, AV_LOG_INFO, "using MMX2\n");

  2514.         else if (flags & SWS_CPU_CAPS_3DNOW)

  2515.             av_log(c, AV_LOG_INFO, "using 3DNOW\n");

  2516.         else if (flags & SWS_CPU_CAPS_MMX)

  2517.             av_log(c, AV_LOG_INFO, "using MMX\n");

  2518.         else if (flags & SWS_CPU_CAPS_ALTIVEC)

  2519.             av_log(c, AV_LOG_INFO, "using AltiVec\n");

  2520.         else

  2521.             av_log(c, AV_LOG_INFO, "using C\n");

  2522.     }

  2523. 

  2524.     if (flags & SWS_PRINT_INFO)

  2525.     {

  2526.         if (flags & SWS_CPU_CAPS_MMX)

  2527.         {

  2528.             if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))

  2529.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");

  2530.             else

  2531.             {

  2532.                 if (c->hLumFilterSize==4)

  2533.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");

  2534.                 else if (c->hLumFilterSize==8)

  2535.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");

  2536.                 else

  2537.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");

  2538. 

  2539.                 if (c->hChrFilterSize==4)

  2540.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");

  2541.                 else if (c->hChrFilterSize==8)

  2542.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");

  2543.                 else

  2544.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");

  2545.             }

  2546.         }

  2547.         else

  2548.         {

  2549. #if defined(ARCH_X86)

  2550.             av_log(c, AV_LOG_VERBOSE, "using X86-Asm scaler for horizontal scaling\n");

  2551. #else

  2552.             if (flags & SWS_FAST_BILINEAR)

  2553.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");

  2554.             else

  2555.                 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");

  2556. #endif

  2557.         }

  2558.         if (isPlanarYUV(dstFormat))

  2559.         {

  2560.             if (c->vLumFilterSize==1)

  2561.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2562.             else

  2563.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2564.         }

  2565.         else

  2566.         {

  2567.             if (c->vLumFilterSize==1 && c->vChrFilterSize==2)

  2568.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"

  2569.                        "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2570.             else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)

  2571.                 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2572.             else

  2573.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2574.         }

  2575. 

  2576.         if (dstFormat==PIX_FMT_BGR24)

  2577.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 Converter\n",

  2578.                    (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));

  2579.         else if (dstFormat==PIX_FMT_RGB32)

  2580.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2581.         else if (dstFormat==PIX_FMT_BGR565)

  2582.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2583.         else if (dstFormat==PIX_FMT_BGR555)

  2584.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2585. 

  2586.         av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);

  2587.     }

  2588.     if (flags & SWS_PRINT_INFO)

  2589.     {

  2590.         av_log(c, AV_LOG_DEBUG, "Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2591.                c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);

  2592.         av_log(c, AV_LOG_DEBUG, "Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2593.                c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);

  2594.     }

  2595. 

  2596.     c->swScale= getSwsFunc(flags);

  2597.     return c;

  2598. }

  2599. 

  2600. /**

  2601.  * swscale wrapper, so we don't need to export the SwsContext.

  2602.  * assumes planar YUV to be in YUV order instead of YVU

  2603.  */

  2604. int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2605.               int srcSliceH, uint8_t* dst[], int dstStride[]){

  2606.     int i;

  2607.     uint8_t* src2[4]= {src[0], src[1], src[2]};

  2608.     uint32_t pal[256];

  2609.     int use_pal=   c->srcFormat == PIX_FMT_PAL8

  2610.                 || c->srcFormat == PIX_FMT_BGR4_BYTE

  2611.                 || c->srcFormat == PIX_FMT_RGB4_BYTE

  2612.                 || c->srcFormat == PIX_FMT_BGR8

  2613.                 || c->srcFormat == PIX_FMT_RGB8;

  2614. 

  2615.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  2616.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  2617.         return 0;

  2618.     }

  2619.     if (c->sliceDir == 0) {

  2620.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  2621.     }

  2622. 

  2623.     if (use_pal){

  2624.         for (i=0; i<256; i++){

  2625.             int p, r, g, b,y,u,v;

  2626.             if(c->srcFormat == PIX_FMT_PAL8){

  2627.                 p=((uint32_t*)(src[1]))[i];

  2628.                 r= (p>>16)&0xFF;

  2629.                 g= (p>> 8)&0xFF;

  2630.                 b=  p     &0xFF;

  2631.             }else if(c->srcFormat == PIX_FMT_RGB8){

  2632.                 r= (i>>5    )*36;

  2633.                 g= ((i>>2)&7)*36;

  2634.                 b= (i&3     )*85;

  2635.             }else if(c->srcFormat == PIX_FMT_BGR8){

  2636.                 b= (i>>6    )*85;

  2637.                 g= ((i>>3)&7)*36;

  2638.                 r= (i&7     )*36;

  2639.             }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){

  2640.                 r= (i>>3    )*255;

  2641.                 g= ((i>>1)&3)*85;

  2642.                 b= (i&1     )*255;

  2643.             }else if(c->srcFormat == PIX_FMT_BGR4_BYTE){

  2644.                 b= (i>>3    )*255;

  2645.                 g= ((i>>1)&3)*85;

  2646.                 r= (i&1     )*255;

  2647.             }

  2648.             y= av_clip_uint8(((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16 );

  2649.             u= av_clip_uint8(((RU*r + GU*g + BU*b)>>RGB2YUV_SHIFT) + 128);

  2650.             v= av_clip_uint8(((RV*r + GV*g + BV*b)>>RGB2YUV_SHIFT) + 128);

  2651.             pal[i]= y + (u<<8) + (v<<16);

  2652.         }

  2653.         src2[1]= (uint8_t*)pal;

  2654.     }

  2655. 

  2656.     // copy strides, so they can safely be modified

  2657.     if (c->sliceDir == 1) {

  2658.         // slices go from top to bottom

  2659.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2]};

  2660.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2]};

  2661.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);

  2662.     } else {

  2663.         // slices go from bottom to top => we flip the image internally

  2664.         uint8_t* dst2[4]= {dst[0] + (c->dstH-1)*dstStride[0],

  2665.                            dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],

  2666.                            dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2]};

  2667.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2]};

  2668.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2]};

  2669. 

  2670.         src2[0] += (srcSliceH-1)*srcStride[0];

  2671.         if (!use_pal)

  2672.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  2673.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  2674. 

  2675.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  2676.     }

  2677. }

  2678. 

  2679. /**

  2680.  * swscale wrapper, so we don't need to export the SwsContext

  2681.  */

  2682. int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2683.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  2684.     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);

  2685. }

  2686. 

  2687. SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,

  2688.                                 float lumaSharpen, float chromaSharpen,

  2689.                                 float chromaHShift, float chromaVShift,

  2690.                                 int verbose)

  2691. {

  2692.     SwsFilter *filter= av_malloc(sizeof(SwsFilter));

  2693. 

  2694.     if (lumaGBlur!=0.0){

  2695.         filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);

  2696.         filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);

  2697.     }else{

  2698.         filter->lumH= sws_getIdentityVec();

  2699.         filter->lumV= sws_getIdentityVec();

  2700.     }

  2701. 

  2702.     if (chromaGBlur!=0.0){

  2703.         filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);

  2704.         filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);

  2705.     }else{

  2706.         filter->chrH= sws_getIdentityVec();

  2707.         filter->chrV= sws_getIdentityVec();

  2708.     }

  2709. 

  2710.     if (chromaSharpen!=0.0){

  2711.         SwsVector *id= sws_getIdentityVec();

  2712.         sws_scaleVec(filter->chrH, -chromaSharpen);

  2713.         sws_scaleVec(filter->chrV, -chromaSharpen);

  2714.         sws_addVec(filter->chrH, id);

  2715.         sws_addVec(filter->chrV, id);

  2716.         sws_freeVec(id);

  2717.     }

  2718. 

  2719.     if (lumaSharpen!=0.0){

  2720.         SwsVector *id= sws_getIdentityVec();

  2721.         sws_scaleVec(filter->lumH, -lumaSharpen);

  2722.         sws_scaleVec(filter->lumV, -lumaSharpen);

  2723.         sws_addVec(filter->lumH, id);

  2724.         sws_addVec(filter->lumV, id);

  2725.         sws_freeVec(id);

  2726.     }

  2727. 

  2728.     if (chromaHShift != 0.0)

  2729.         sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));

  2730. 

  2731.     if (chromaVShift != 0.0)

  2732.         sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));

  2733. 

  2734.     sws_normalizeVec(filter->chrH, 1.0);

  2735.     sws_normalizeVec(filter->chrV, 1.0);

  2736.     sws_normalizeVec(filter->lumH, 1.0);

  2737.     sws_normalizeVec(filter->lumV, 1.0);

  2738. 

  2739.     if (verbose) sws_printVec(filter->chrH);

  2740.     if (verbose) sws_printVec(filter->lumH);

  2741. 

  2742.     return filter;

  2743. }

  2744. 

  2745. /**

  2746.  * returns a normalized gaussian curve used to filter stuff

  2747.  * quality=3 is high quality, lowwer is lowwer quality

  2748.  */

  2749. SwsVector *sws_getGaussianVec(double variance, double quality){

  2750.     const int length= (int)(variance*quality + 0.5) | 1;

  2751.     int i;

  2752.     double *coeff= av_malloc(length*sizeof(double));

  2753.     double middle= (length-1)*0.5;

  2754.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2755. 

  2756.     vec->coeff= coeff;

  2757.     vec->length= length;

  2758. 

  2759.     for (i=0; i<length; i++)

  2760.     {

  2761.         double dist= i-middle;

  2762.         coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);

  2763.     }

  2764. 

  2765.     sws_normalizeVec(vec, 1.0);

  2766. 

  2767.     return vec;

  2768. }

  2769. 

  2770. SwsVector *sws_getConstVec(double c, int length){

  2771.     int i;

  2772.     double *coeff= av_malloc(length*sizeof(double));

  2773.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2774. 

  2775.     vec->coeff= coeff;

  2776.     vec->length= length;

  2777. 

  2778.     for (i=0; i<length; i++)

  2779.         coeff[i]= c;

  2780. 

  2781.     return vec;

  2782. }

  2783. 

  2784. 

  2785. SwsVector *sws_getIdentityVec(void){

  2786.     return sws_getConstVec(1.0, 1);

  2787. }

  2788. 

  2789. double sws_dcVec(SwsVector *a){

  2790.     int i;

  2791.     double sum=0;

  2792. 

  2793.     for (i=0; i<a->length; i++)

  2794.         sum+= a->coeff[i];

  2795. 

  2796.     return sum;

  2797. }

  2798. 

  2799. void sws_scaleVec(SwsVector *a, double scalar){

  2800.     int i;

  2801. 

  2802.     for (i=0; i<a->length; i++)

  2803.         a->coeff[i]*= scalar;

  2804. }

  2805. 

  2806. void sws_normalizeVec(SwsVector *a, double height){

  2807.     sws_scaleVec(a, height/sws_dcVec(a));

  2808. }

  2809. 

  2810. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){

  2811.     int length= a->length + b->length - 1;

  2812.     double *coeff= av_malloc(length*sizeof(double));

  2813.     int i, j;

  2814.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2815. 

  2816.     vec->coeff= coeff;

  2817.     vec->length= length;

  2818. 

  2819.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2820. 

  2821.     for (i=0; i<a->length; i++)

  2822.     {

  2823.         for (j=0; j<b->length; j++)

  2824.         {

  2825.             coeff[i+j]+= a->coeff[i]*b->coeff[j];

  2826.         }

  2827.     }

  2828. 

  2829.     return vec;

  2830. }

  2831. 

  2832. static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){

  2833.     int length= FFMAX(a->length, b->length);

  2834.     double *coeff= av_malloc(length*sizeof(double));

  2835.     int i;

  2836.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2837. 

  2838.     vec->coeff= coeff;

  2839.     vec->length= length;

  2840. 

  2841.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2842. 

  2843.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  2844.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];

  2845. 

  2846.     return vec;

  2847. }

  2848. 

  2849. static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){

  2850.     int length= FFMAX(a->length, b->length);

  2851.     double *coeff= av_malloc(length*sizeof(double));

  2852.     int i;

  2853.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2854. 

  2855.     vec->coeff= coeff;

  2856.     vec->length= length;

  2857. 

  2858.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2859. 

  2860.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  2861.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];

  2862. 

  2863.     return vec;

  2864. }

  2865. 

  2866. /* shift left / or right if "shift" is negative */

  2867. static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){

  2868.     int length= a->length + FFABS(shift)*2;

  2869.     double *coeff= av_malloc(length*sizeof(double));

  2870.     int i;

  2871.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2872. 

  2873.     vec->coeff= coeff;

  2874.     vec->length= length;

  2875. 

  2876.     for (i=0; i<length; i++) coeff[i]= 0.0;

  2877. 

  2878.     for (i=0; i<a->length; i++)

  2879.     {

  2880.         coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];

  2881.     }

  2882. 

  2883.     return vec;

  2884. }

  2885. 

  2886. void sws_shiftVec(SwsVector *a, int shift){

  2887.     SwsVector *shifted= sws_getShiftedVec(a, shift);

  2888.     av_free(a->coeff);

  2889.     a->coeff= shifted->coeff;

  2890.     a->length= shifted->length;

  2891.     av_free(shifted);

  2892. }

  2893. 

  2894. void sws_addVec(SwsVector *a, SwsVector *b){

  2895.     SwsVector *sum= sws_sumVec(a, b);

  2896.     av_free(a->coeff);

  2897.     a->coeff= sum->coeff;

  2898.     a->length= sum->length;

  2899.     av_free(sum);

  2900. }

  2901. 

  2902. void sws_subVec(SwsVector *a, SwsVector *b){

  2903.     SwsVector *diff= sws_diffVec(a, b);

  2904.     av_free(a->coeff);

  2905.     a->coeff= diff->coeff;

  2906.     a->length= diff->length;

  2907.     av_free(diff);

  2908. }

  2909. 

  2910. void sws_convVec(SwsVector *a, SwsVector *b){

  2911.     SwsVector *conv= sws_getConvVec(a, b);

  2912.     av_free(a->coeff);

  2913.     a->coeff= conv->coeff;

  2914.     a->length= conv->length;

  2915.     av_free(conv);

  2916. }

  2917. 

  2918. SwsVector *sws_cloneVec(SwsVector *a){

  2919.     double *coeff= av_malloc(a->length*sizeof(double));

  2920.     int i;

  2921.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  2922. 

  2923.     vec->coeff= coeff;

  2924.     vec->length= a->length;

  2925. 

  2926.     for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];

  2927. 

  2928.     return vec;

  2929. }

  2930. 

  2931. void sws_printVec(SwsVector *a){

  2932.     int i;

  2933.     double max=0;

  2934.     double min=0;

  2935.     double range;

  2936. 

  2937.     for (i=0; i<a->length; i++)

  2938.         if (a->coeff[i]>max) max= a->coeff[i];

  2939. 

  2940.     for (i=0; i<a->length; i++)

  2941.         if (a->coeff[i]<min) min= a->coeff[i];

  2942. 

  2943.     range= max - min;

  2944. 

  2945.     for (i=0; i<a->length; i++)

  2946.     {

  2947.         int x= (int)((a->coeff[i]-min)*60.0/range +0.5);

  2948.         av_log(NULL, AV_LOG_DEBUG, "%1.3f ", a->coeff[i]);

  2949.         for (;x>0; x--) av_log(NULL, AV_LOG_DEBUG, " ");

  2950.         av_log(NULL, AV_LOG_DEBUG, "|\n");

  2951.     }

  2952. }

  2953. 

  2954. void sws_freeVec(SwsVector *a){

  2955.     if (!a) return;

  2956.     av_freep(&a->coeff);

  2957.     a->length=0;

  2958.     av_free(a);

  2959. }

  2960. 

  2961. void sws_freeFilter(SwsFilter *filter){

  2962.     if (!filter) return;

  2963. 

  2964.     if (filter->lumH) sws_freeVec(filter->lumH);

  2965.     if (filter->lumV) sws_freeVec(filter->lumV);

  2966.     if (filter->chrH) sws_freeVec(filter->chrH);

  2967.     if (filter->chrV) sws_freeVec(filter->chrV);

  2968.     av_free(filter);

  2969. }

  2970. 

  2971. 

  2972. void sws_freeContext(SwsContext *c){

  2973.     int i;

  2974.     if (!c) return;

  2975. 

  2976.     if (c->lumPixBuf)

  2977.     {

  2978.         for (i=0; i<c->vLumBufSize; i++)

  2979.             av_freep(&c->lumPixBuf[i]);

  2980.         av_freep(&c->lumPixBuf);

  2981.     }

  2982. 

  2983.     if (c->chrPixBuf)

  2984.     {

  2985.         for (i=0; i<c->vChrBufSize; i++)

  2986.             av_freep(&c->chrPixBuf[i]);

  2987.         av_freep(&c->chrPixBuf);

  2988.     }

  2989. 

  2990.     av_freep(&c->vLumFilter);

  2991.     av_freep(&c->vChrFilter);

  2992.     av_freep(&c->hLumFilter);

  2993.     av_freep(&c->hChrFilter);

  2994. #ifdef HAVE_ALTIVEC

  2995.     av_freep(&c->vYCoeffsBank);

  2996.     av_freep(&c->vCCoeffsBank);

  2997. #endif

  2998. 

  2999.     av_freep(&c->vLumFilterPos);

  3000.     av_freep(&c->vChrFilterPos);

  3001.     av_freep(&c->hLumFilterPos);

  3002.     av_freep(&c->hChrFilterPos);

  3003. 

  3004. #if defined(ARCH_X86) && defined(CONFIG_GPL)

  3005. #ifdef MAP_ANONYMOUS

  3006.     if (c->funnyYCode) munmap(c->funnyYCode, MAX_FUNNY_CODE_SIZE);

  3007.     if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);

  3008. #else

  3009.     av_free(c->funnyYCode);

  3010.     av_free(c->funnyUVCode);

  3011. #endif

  3012.     c->funnyYCode=NULL;

  3013.     c->funnyUVCode=NULL;

  3014. #endif /* defined(ARCH_X86) */

  3015. 

  3016.     av_freep(&c->lumMmx2Filter);

  3017.     av_freep(&c->chrMmx2Filter);

  3018.     av_freep(&c->lumMmx2FilterPos);

  3019.     av_freep(&c->chrMmx2FilterPos);

  3020.     av_freep(&c->yuvTable);

  3021. 

  3022.     av_free(c);

  3023. }

  3024. 

  3025. /**

  3026.  * Checks if context is valid or reallocs a new one instead.

  3027.  * If context is NULL, just calls sws_getContext() to get a new one.

  3028.  * Otherwise, checks if the parameters are the same already saved in context.

  3029.  * If that is the case, returns the current context.

  3030.  * Otherwise, frees context and gets a new one.

  3031.  *

  3032.  * Be warned that srcFilter, dstFilter are not checked, they are

  3033.  * asumed to remain valid.

  3034.  */

  3035. struct SwsContext *sws_getCachedContext(struct SwsContext *context,

  3036.                                         int srcW, int srcH, int srcFormat,

  3037.                                         int dstW, int dstH, int dstFormat, int flags,

  3038.                                         SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)

  3039. {

  3040.     static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};

  3041. 

  3042.     if (!param)

  3043.         param = default_param;

  3044. 

  3045.     if (context) {

  3046.         if (context->srcW != srcW || context->srcH != srcH ||

  3047.             context->srcFormat != srcFormat ||

  3048.             context->dstW != dstW || context->dstH != dstH ||

  3049.             context->dstFormat != dstFormat || context->flags != flags ||

  3050.             context->param[0] != param[0] || context->param[1] != param[1])

  3051.         {

  3052.             sws_freeContext(context);

  3053.             context = NULL;

  3054.         }

  3055.     }

  3056.     if (!context) {

  3057.         return sws_getContext(srcW, srcH, srcFormat,

  3058.                               dstW, dstH, dstFormat, flags,

  3059.                               srcFilter, dstFilter, param);

  3060.     }

  3061.     return context;

  3062. }