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.
28425 Commits
32 Branches
1.0GB
Tree: c38d9780eb
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'c38d9780eb'
${ noResults }
FFmpeg/libswscale/swscale.c

2034 lines
71KB
Raw Blame History 

  1. /*

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

  3.  *

  4.  * This file is part of Libav.

  5.  *

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

  7.  * modify it under the terms of the GNU Lesser General Public

  8.  * License as published by the Free Software Foundation; either

  9.  * version 2.1 of the License, or (at your option) any later version.

  10.  *

  11.  * Libav 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 GNU

  14.  * Lesser General Public License for more details.

  15.  *

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

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

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

  19.  */

  20. 

  21. /*

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

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

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

  25. 

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

  27.   YV12 -> {BGR,RGB}{1,4,8,12,15,16,24,32}

  28.   x -> x

  29.   YUV9 -> YV12

  30.   YUV9/YV12 -> Y800

  31.   Y800 -> YUV9/YV12

  32.   BGR24 -> BGR32 & RGB24 -> RGB32

  33.   BGR32 -> BGR24 & RGB32 -> RGB24

  34.   BGR15 -> BGR16

  35. */

  36. 

  37. /*

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

  39.  YV12 -> BGR12/BGR16

  40.  YV12 -> YV12

  41.  BGR15 -> BGR16

  42.  BGR16 -> BGR16

  43.  YVU9 -> YV12

  44. 

  45. untested special converters

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

  47.   YV12/I420 -> YV12/I420

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

  49.   BGR24 -> BGR32 & RGB24 -> RGB32

  50.   BGR32 -> BGR24 & RGB32 -> RGB24

  51.   BGR24 -> YV12

  52. */

  53. 

  54. #include <inttypes.h>

  55. #include <string.h>

  56. #include <math.h>

  57. #include <stdio.h>

  58. #include "config.h"

  59. #include <assert.h>

  60. #include "swscale.h"

  61. #include "swscale_internal.h"

  62. #include "rgb2rgb.h"

  63. #include "libavutil/intreadwrite.h"

  64. #include "libavutil/x86_cpu.h"

  65. #include "libavutil/avutil.h"

  66. #include "libavutil/mathematics.h"

  67. #include "libavutil/bswap.h"

  68. #include "libavutil/pixdesc.h"

  69. 

  70. #undef MOVNTQ

  71. #undef PAVGB

  72. 

  73. //#undef HAVE_MMX2

  74. //#define HAVE_AMD3DNOW

  75. //#undef HAVE_MMX

  76. //#undef ARCH_X86

  77. #define DITHER1XBPP

  78. 

  79. #define isPacked(x)         (       \

  80.            (x)==PIX_FMT_PAL8        \

  81.         || (x)==PIX_FMT_YUYV422     \

  82.         || (x)==PIX_FMT_UYVY422     \

  83.         || (x)==PIX_FMT_Y400A       \

  84.         || isAnyRGB(x)              \

  85.     )

  86. 

  87. #define RGB2YUV_SHIFT 15

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

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

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

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

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

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

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

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

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

  97. 

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

  99.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709

  100.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709

  101.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  102.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  103.     {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC

  104.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  105.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  106.     {0.701 , 0.087 , 0.212 , -0.384, 0.5, -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M

  107. };

  108. 

  109. /*

  110. NOTES

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

  112. 

  113. TODO

  114. more intelligent misalignment avoidance for the horizontal scaler

  115. write special vertical cubic upscale version

  116. optimize C code (YV12 / minmax)

  117. add support for packed pixel YUV input & output

  118. add support for Y8 output

  119. optimize BGR24 & BGR32

  120. add BGR4 output support

  121. write special BGR->BGR scaler

  122. */

  123. 

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

  125. {  1,   3,   1,   3,   1,   3,   1,   3, },

  126. {  2,   0,   2,   0,   2,   0,   2,   0, },

  127. };

  128. 

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

  130. {  6,   2,   6,   2,   6,   2,   6,   2, },

  131. {  0,   4,   0,   4,   0,   4,   0,   4, },

  132. };

  133. 

  134. DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={

  135. {  8,   4,  11,   7,   8,   4,  11,   7, },

  136. {  2,  14,   1,  13,   2,  14,   1,  13, },

  137. { 10,   6,   9,   5,  10,   6,   9,   5, },

  138. {  0,  12,   3,  15,   0,  12,   3,  15, },

  139. };

  140. 

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

  142. { 17,   9,  23,  15,  16,   8,  22,  14, },

  143. {  5,  29,   3,  27,   4,  28,   2,  26, },

  144. { 21,  13,  19,  11,  20,  12,  18,  10, },

  145. {  0,  24,   6,  30,   1,  25,   7,  31, },

  146. { 16,   8,  22,  14,  17,   9,  23,  15, },

  147. {  4,  28,   2,  26,   5,  29,   3,  27, },

  148. { 20,  12,  18,  10,  21,  13,  19,  11, },

  149. {  1,  25,   7,  31,   0,  24,   6,  30, },

  150. };

  151. 

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

  153. {  0,  55,  14,  68,   3,  58,  17,  72, },

  154. { 37,  18,  50,  32,  40,  22,  54,  35, },

  155. {  9,  64,   5,  59,  13,  67,   8,  63, },

  156. { 46,  27,  41,  23,  49,  31,  44,  26, },

  157. {  2,  57,  16,  71,   1,  56,  15,  70, },

  158. { 39,  21,  52,  34,  38,  19,  51,  33, },

  159. { 11,  66,   7,  62,  10,  65,   6,  60, },

  160. { 48,  30,  43,  25,  47,  29,  42,  24, },

  161. };

  162. 

  163. #if 1

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

  165. {117,  62, 158, 103, 113,  58, 155, 100, },

  166. { 34, 199,  21, 186,  31, 196,  17, 182, },

  167. {144,  89, 131,  76, 141,  86, 127,  72, },

  168. {  0, 165,  41, 206,  10, 175,  52, 217, },

  169. {110,  55, 151,  96, 120,  65, 162, 107, },

  170. { 28, 193,  14, 179,  38, 203,  24, 189, },

  171. {138,  83, 124,  69, 148,  93, 134,  79, },

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

  173. };

  174. #elif 1

  175. // tries to correct a gamma of 1.5

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

  177. {  0, 143,  18, 200,   2, 156,  25, 215, },

  178. { 78,  28, 125,  64,  89,  36, 138,  74, },

  179. { 10, 180,   3, 161,  16, 195,   8, 175, },

  180. {109,  51,  93,  38, 121,  60, 105,  47, },

  181. {  1, 152,  23, 210,   0, 147,  20, 205, },

  182. { 85,  33, 134,  71,  81,  30, 130,  67, },

  183. { 14, 190,   6, 171,  12, 185,   5, 166, },

  184. {117,  57, 101,  44, 113,  54,  97,  41, },

  185. };

  186. #elif 1

  187. // tries to correct a gamma of 2.0

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

  189. {  0, 124,   8, 193,   0, 140,  12, 213, },

  190. { 55,  14, 104,  42,  66,  19, 119,  52, },

  191. {  3, 168,   1, 145,   6, 187,   3, 162, },

  192. { 86,  31,  70,  21,  99,  39,  82,  28, },

  193. {  0, 134,  11, 206,   0, 129,   9, 200, },

  194. { 62,  17, 114,  48,  58,  16, 109,  45, },

  195. {  5, 181,   2, 157,   4, 175,   1, 151, },

  196. { 95,  36,  78,  26,  90,  34,  74,  24, },

  197. };

  198. #else

  199. // tries to correct a gamma of 2.5

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

  201. {  0, 107,   3, 187,   0, 125,   6, 212, },

  202. { 39,   7,  86,  28,  49,  11, 102,  36, },

  203. {  1, 158,   0, 131,   3, 180,   1, 151, },

  204. { 68,  19,  52,  12,  81,  25,  64,  17, },

  205. {  0, 119,   5, 203,   0, 113,   4, 195, },

  206. { 45,   9,  96,  33,  42,   8,  91,  30, },

  207. {  2, 172,   1, 144,   2, 165,   0, 137, },

  208. { 77,  23,  60,  15,  72,  21,  56,  14, },

  209. };

  210. #endif

  211. 

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

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

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

  215.                                                     int dstW, int chrDstW, int big_endian)

  216. {

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

  218.     int i;

  219. 

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

  221.         int val = 1 << 10;

  222.         int j;

  223. 

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

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

  226. 

  227.         if (big_endian) {

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

  229.         } else {

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

  231.         }

  232.     }

  233. 

  234.     if (uDest) {

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

  236.             int u = 1 << 10;

  237.             int v = 1 << 10;

  238.             int j;

  239. 

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

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

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

  243.             }

  244. 

  245.             if (big_endian) {

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

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

  248.             } else {

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

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

  251.             }

  252.         }

  253.     }

  254. 

  255.     if (CONFIG_SWSCALE_ALPHA && aDest) {

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

  257.             int val = 1 << 10;

  258.             int j;

  259. 

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

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

  262. 

  263.             if (big_endian) {

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

  265.             } else {

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

  267.             }

  268.         }

  269.     }

  270. }

  271. 

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

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

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

  275.                                  enum PixelFormat dstFormat)

  276. {

  277.     if (isBE(dstFormat)) {

  278.         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,

  279.                                chrFilter, chrSrc, chrFilterSize,

  280.                                alpSrc,

  281.                                dest, uDest, vDest, aDest,

  282.                                dstW, chrDstW, 1);

  283.     } else {

  284.         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,

  285.                                chrFilter, chrSrc, chrFilterSize,

  286.                                alpSrc,

  287.                                dest, uDest, vDest, aDest,

  288.                                dstW, chrDstW, 0);

  289.     }

  290. }

  291. 

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

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

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

  295. {

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

  297.     int i;

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

  299.         int val=1<<18;

  300.         int j;

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

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

  303. 

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

  305.     }

  306. 

  307.     if (uDest)

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

  309.             int u=1<<18;

  310.             int v=1<<18;

  311.             int j;

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

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

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

  315.             }

  316. 

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

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

  319.         }

  320. 

  321.     if (CONFIG_SWSCALE_ALPHA && aDest)

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

  323.             int val=1<<18;

  324.             int j;

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

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

  327. 

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

  329.         }

  330. 

  331. }

  332. 

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

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

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

  336. {

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

  338.     int i;

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

  340.         int val=1<<18;

  341.         int j;

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

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

  344. 

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

  346.     }

  347. 

  348.     if (!uDest)

  349.         return;

  350. 

  351.     if (dstFormat == PIX_FMT_NV12)

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

  353.             int u=1<<18;

  354.             int v=1<<18;

  355.             int j;

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

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

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

  359.             }

  360. 

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

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

  363.         }

  364.     else

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

  366.             int u=1<<18;

  367.             int v=1<<18;

  368.             int j;

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

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

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

  372.             }

  373. 

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

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

  376.         }

  377. }

  378. 

  379. #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \

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

  381.         int j;\

  382.         int Y1 = 1<<18;\

  383.         int Y2 = 1<<18;\

  384.         int U  = 1<<18;\

  385.         int V  = 1<<18;\

  386.         int av_unused A1, A2;\

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

  388.         const int i2= 2*i;\

  389.         \

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

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

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

  393.         }\

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

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

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

  397.         }\

  398.         Y1>>=19;\

  399.         Y2>>=19;\

  400.         U >>=19;\

  401.         V >>=19;\

  402.         if (alpha) {\

  403.             A1 = 1<<18;\

  404.             A2 = 1<<18;\

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

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

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

  408.             }\

  409.             A1>>=19;\

  410.             A2>>=19;\

  411.         }

  412. 

  413. #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \

  414.         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\

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

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

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

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

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

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

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

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

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

  424.         }\

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

  426.             A1=av_clip_uint8(A1);\

  427.             A2=av_clip_uint8(A2);\

  428.         }

  429. 

  430. #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \

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

  432.         int j;\

  433.         int Y = 0;\

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

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

  436.         int av_unused A;\

  437.         int R,G,B;\

  438.         \

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

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

  441.         }\

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

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

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

  445.         }\

  446.         Y >>=10;\

  447.         U >>=10;\

  448.         V >>=10;\

  449.         if (alpha) {\

  450.             A = rnd;\

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

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

  453.             A >>=19;\

  454.             if (A&256)\

  455.                 A = av_clip_uint8(A);\

  456.         }

  457. 

  458. #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \

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

  460.         Y-= c->yuv2rgb_y_offset;\

  461.         Y*= c->yuv2rgb_y_coeff;\

  462.         Y+= rnd;\

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

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

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

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

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

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

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

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

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

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

  473.         }

  474. 

  475. #define YSCALE_YUV_2_GRAY16_C \

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

  477.         int j;\

  478.         int Y1 = 1<<18;\

  479.         int Y2 = 1<<18;\

  480.         int U  = 1<<18;\

  481.         int V  = 1<<18;\

  482.         \

  483.         const int i2= 2*i;\

  484.         \

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

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

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

  488.         }\

  489.         Y1>>=11;\

  490.         Y2>>=11;\

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

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

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

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

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

  496.         }

  497. 

  498. #define YSCALE_YUV_2_RGBX_C(type,alpha) \

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

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

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

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

  503. 

  504. #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \

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

  506.         const int i2= 2*i;       \

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

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

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

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

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

  512.         int av_unused A1, A2;                                         \

  513.         if (alpha) {\

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

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

  516.         }

  517. 

  518. #define YSCALE_YUV_2_GRAY16_2_C   \

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

  520.         const int i2= 2*i;       \

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

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

  523. 

  524. #define YSCALE_YUV_2_RGB2_C(type,alpha) \

  525.     YSCALE_YUV_2_PACKED2_C(type,alpha)\

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

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

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

  529. 

  530. #define YSCALE_YUV_2_PACKED1_C(type,alpha) \

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

  532.         const int i2= 2*i;\

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

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

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

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

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

  538.         int av_unused A1, A2;\

  539.         if (alpha) {\

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

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

  542.         }

  543. 

  544. #define YSCALE_YUV_2_GRAY16_1_C \

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

  546.         const int i2= 2*i;\

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

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

  549. 

  550. #define YSCALE_YUV_2_RGB1_C(type,alpha) \

  551.     YSCALE_YUV_2_PACKED1_C(type,alpha)\

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

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

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

  555. 

  556. #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \

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

  558.         const int i2= 2*i;\

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

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

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

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

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

  564.         int av_unused A1, A2;\

  565.         if (alpha) {\

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

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

  568.         }

  569. 

  570. #define YSCALE_YUV_2_RGB1B_C(type,alpha) \

  571.     YSCALE_YUV_2_PACKED1B_C(type,alpha)\

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

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

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

  575. 

  576. #define YSCALE_YUV_2_MONO2_C \

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

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

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

  580.         int acc;\

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

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

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

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

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

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

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

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

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

  590.         dest++;\

  591.     }

  592. 

  593. #define YSCALE_YUV_2_MONOX_C \

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

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

  596.     int acc=0;\

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

  598.         int j;\

  599.         int Y1=1<<18;\

  600.         int Y2=1<<18;\

  601. \

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

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

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

  605.         }\

  606.         Y1>>=19;\

  607.         Y2>>=19;\

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

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

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

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

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

  613.         }\

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

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

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

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

  618.             dest++;\

  619.         }\

  620.     }

  621. 

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

  623.     switch(c->dstFormat) {\

  624.     case PIX_FMT_RGB48BE:\

  625.     case PIX_FMT_RGB48LE:\

  626.         func(uint8_t,0)\

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

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

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

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

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

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

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

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

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

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

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

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

  639.             dest+=12;\

  640.         }\

  641.         break;\

  642.     case PIX_FMT_BGR48BE:\

  643.     case PIX_FMT_BGR48LE:\

  644.         func(uint8_t,0)\

  645.             ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\

  646.             ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\

  647.             ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\

  648.             ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\

  649.             ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\

  650.             ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\

  651.             dest+=12;\

  652.         }\

  653.         break;\

  654.     case PIX_FMT_RGBA:\

  655.     case PIX_FMT_BGRA:\

  656.         if (CONFIG_SMALL) {\

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

  658.             func(uint32_t,needAlpha)\

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

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

  661.             }\

  662.         } else {\

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

  664.                 func(uint32_t,1)\

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

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

  667.                 }\

  668.             } else {\

  669.                 func(uint32_t,0)\

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

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

  672.                 }\

  673.             }\

  674.         }\

  675.         break;\

  676.     case PIX_FMT_ARGB:\

  677.     case PIX_FMT_ABGR:\

  678.         if (CONFIG_SMALL) {\

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

  680.             func(uint32_t,needAlpha)\

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

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

  683.             }\

  684.         } else {\

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

  686.                 func(uint32_t,1)\

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

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

  689.                 }\

  690.             } else {\

  691.                 func(uint32_t,0)\

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

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

  694.                 }\

  695.             }\

  696.         }                \

  697.         break;\

  698.     case PIX_FMT_RGB24:\

  699.         func(uint8_t,0)\

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

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

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

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

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

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

  706.             dest+=6;\

  707.         }\

  708.         break;\

  709.     case PIX_FMT_BGR24:\

  710.         func(uint8_t,0)\

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

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

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

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

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

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

  717.             dest+=6;\

  718.         }\

  719.         break;\

  720.     case PIX_FMT_RGB565BE:\

  721.     case PIX_FMT_RGB565LE:\

  722.     case PIX_FMT_BGR565BE:\

  723.     case PIX_FMT_BGR565LE:\

  724.         {\

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

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

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

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

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

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

  731.             func(uint16_t,0)\

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

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

  734.             }\

  735.         }\

  736.         break;\

  737.     case PIX_FMT_RGB555BE:\

  738.     case PIX_FMT_RGB555LE:\

  739.     case PIX_FMT_BGR555BE:\

  740.     case PIX_FMT_BGR555LE:\

  741.         {\

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

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

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

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

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

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

  748.             func(uint16_t,0)\

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

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

  751.             }\

  752.         }\

  753.         break;\

  754.     case PIX_FMT_RGB444BE:\

  755.     case PIX_FMT_RGB444LE:\

  756.     case PIX_FMT_BGR444BE:\

  757.     case PIX_FMT_BGR444LE:\

  758.         {\

  759.             const int dr1= dither_4x4_16[y&3    ][0];\

  760.             const int dg1= dither_4x4_16[y&3    ][1];\

  761.             const int db1= dither_4x4_16[(y&3)^3][0];\

  762.             const int dr2= dither_4x4_16[y&3    ][1];\

  763.             const int dg2= dither_4x4_16[y&3    ][0];\

  764.             const int db2= dither_4x4_16[(y&3)^3][1];\

  765.             func(uint16_t,0)\

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

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

  768.             }\

  769.         }\

  770.         break;\

  771.     case PIX_FMT_RGB8:\

  772.     case PIX_FMT_BGR8:\

  773.         {\

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

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

  776.             func(uint8_t,0)\

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

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

  779.             }\

  780.         }\

  781.         break;\

  782.     case PIX_FMT_RGB4:\

  783.     case PIX_FMT_BGR4:\

  784.         {\

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

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

  787.             func(uint8_t,0)\

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

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

  790.             }\

  791.         }\

  792.         break;\

  793.     case PIX_FMT_RGB4_BYTE:\

  794.     case PIX_FMT_BGR4_BYTE:\

  795.         {\

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

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

  798.             func(uint8_t,0)\

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

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

  801.             }\

  802.         }\

  803.         break;\

  804.     case PIX_FMT_MONOBLACK:\

  805.     case PIX_FMT_MONOWHITE:\

  806.         {\

  807.             func_monoblack\

  808.         }\

  809.         break;\

  810.     case PIX_FMT_YUYV422:\

  811.         func2\

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

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

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

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

  816.         }                \

  817.         break;\

  818.     case PIX_FMT_UYVY422:\

  819.         func2\

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

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

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

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

  824.         }                \

  825.         break;\

  826.     case PIX_FMT_GRAY16BE:\

  827.         func_g16\

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

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

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

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

  832.         }                \

  833.         break;\

  834.     case PIX_FMT_GRAY16LE:\

  835.         func_g16\

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

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

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

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

  840.         }                \

  841.         break;\

  842.     }

  843. 

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

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

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

  847. {

  848.     int i;

  849.     YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void,0), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)

  850. }

  851. 

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

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

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

  855. {

  856.     int i;

  857.     int step= c->dstFormatBpp/8;

  858.     int aidx= 3;

  859. 

  860.     switch(c->dstFormat) {

  861.     case PIX_FMT_ARGB:

  862.         dest++;

  863.         aidx= 0;

  864.     case PIX_FMT_RGB24:

  865.         aidx--;

  866.     case PIX_FMT_RGBA:

  867.         if (CONFIG_SMALL) {

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

  869.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

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

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

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

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

  874.                 dest+= step;

  875.             }

  876.         } else {

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

  878.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  879.                     dest[aidx]= A;

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

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

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

  883.                     dest+= step;

  884.                 }

  885.             } else {

  886.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  887.                     dest[aidx]= 255;

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

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

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

  891.                     dest+= step;

  892.                 }

  893.             }

  894.         }

  895.         break;

  896.     case PIX_FMT_ABGR:

  897.         dest++;

  898.         aidx= 0;

  899.     case PIX_FMT_BGR24:

  900.         aidx--;

  901.     case PIX_FMT_BGRA:

  902.         if (CONFIG_SMALL) {

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

  904.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

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

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

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

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

  909.                 dest+= step;

  910.             }

  911.         } else {

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

  913.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  914.                     dest[aidx]= A;

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

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

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

  918.                     dest+= step;

  919.                 }

  920.             } else {

  921.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  922.                     dest[aidx]= 255;

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

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

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

  926.                     dest+= step;

  927.                 }

  928.             }

  929.         }

  930.         break;

  931.     default:

  932.         assert(0);

  933.     }

  934. }

  935. 

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

  937. {

  938.     int i;

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

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

  941.         memset(ptr, val, width);

  942.         ptr += stride;

  943.     }

  944. }

  945. 

  946. static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,

  947.                             uint32_t *unused)

  948. {

  949.     int i;

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

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

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

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

  954. 

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

  956.     }

  957. }

  958. 

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

  960.                              const uint8_t *src1, const uint8_t *src2,

  961.                              long width, uint32_t *unused)

  962. {

  963.     int i;

  964.     assert(src1==src2);

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

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

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

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

  969. 

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

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

  972.     }

  973. }

  974. 

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

  976.                                   const uint8_t *src1, const uint8_t *src2,

  977.                                   long width, uint32_t *unused)

  978. {

  979.     int i;

  980.     assert(src1==src2);

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

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

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

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

  985. 

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

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

  988.     }

  989. }

  990. 

  991. static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, long width,

  992.                             uint32_t *unused)

  993. {

  994.     int i;

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

  996.         int b = src[i*6+0];

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

  998.         int r = src[i*6+4];

  999. 

  1000.         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1001.     }

  1002. }

  1003. 

  1004. static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,

  1005.                              const uint8_t *src1, const uint8_t *src2,

  1006.                              long width, uint32_t *unused)

  1007. {

  1008.     int i;

  1009.     for (i = 0; i < width; i++) {

  1010.         int b = src1[6*i + 0];

  1011.         int g = src1[6*i + 2];

  1012.         int r = src1[6*i + 4];

  1013. 

  1014.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1015.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1016.     }

  1017. }

  1018. 

  1019. static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1020.                                   const uint8_t *src1, const uint8_t *src2,

  1021.                                   long width, uint32_t *unused)

  1022. {

  1023.     int i;

  1024.     for (i = 0; i < width; i++) {

  1025.         int b= src1[12*i + 0] + src1[12*i + 6];

  1026.         int g= src1[12*i + 2] + src1[12*i + 8];

  1027.         int r= src1[12*i + 4] + src1[12*i + 10];

  1028. 

  1029.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1030.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1031.     }

  1032. }

  1033. 

  1034. #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\

  1035. static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\

  1036. {\

  1037.     int i;\

  1038.     for (i=0; i<width; i++) {\

  1039.         int b= (((const type*)src)[i]>>shb)&maskb;\

  1040.         int g= (((const type*)src)[i]>>shg)&maskg;\

  1041.         int r= (((const type*)src)[i]>>shr)&maskr;\

  1042. \

  1043.         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\

  1044.     }\

  1045. }

  1046. 

  1047. BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1048. BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)

  1049. BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1050. BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)

  1051. BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)

  1052. BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)

  1053. BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)

  1054. BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)

  1055. 

  1056. static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1057. {

  1058.     int i;

  1059.     for (i=0; i<width; i++) {

  1060.         dst[i]= src[4*i];

  1061.     }

  1062. }

  1063. 

  1064. #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \

  1065. static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1066. {\

  1067.     int i;\

  1068.     for (i=0; i<width; i++) {\

  1069.         int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\

  1070.         int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\

  1071.         int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\

  1072. \

  1073.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\

  1074.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\

  1075.     }\

  1076. }\

  1077. static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1078. {\

  1079.     int i;\

  1080.     for (i=0; i<width; i++) {\

  1081.         int pix0= ((const type*)src)[2*i+0]>>shp;\

  1082.         int pix1= ((const type*)src)[2*i+1]>>shp;\

  1083.         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\

  1084.         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\

  1085.         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\

  1086.         g&= maskg|(2*maskg);\

  1087. \

  1088.         g>>=shg;\

  1089. \

  1090.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\

  1091.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\

  1092.     }\

  1093. }

  1094. 

  1095. BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1096. BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1097. BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1098. BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1099. BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)

  1100. BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)

  1101. BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)

  1102. BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)

  1103. 

  1104. static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)

  1105. {

  1106.     int i;

  1107.     for (i=0; i<width; i++) {

  1108.         int d= src[i];

  1109. 

  1110.         dst[i]= pal[d] & 0xFF;

  1111.     }

  1112. }

  1113. 

  1114. static inline void palToUV(uint8_t *dstU, uint8_t *dstV,

  1115.                            const uint8_t *src1, const uint8_t *src2,

  1116.                            long width, uint32_t *pal)

  1117. {

  1118.     int i;

  1119.     assert(src1 == src2);

  1120.     for (i=0; i<width; i++) {

  1121.         int p= pal[src1[i]];

  1122. 

  1123.         dstU[i]= p>>8;

  1124.         dstV[i]= p>>16;

  1125.     }

  1126. }

  1127. 

  1128. static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1129. {

  1130.     int i, j;

  1131.     for (i=0; i<width/8; i++) {

  1132.         int d= ~src[i];

  1133.         for(j=0; j<8; j++)

  1134.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1135.     }

  1136. }

  1137. 

  1138. static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1139. {

  1140.     int i, j;

  1141.     for (i=0; i<width/8; i++) {

  1142.         int d= src[i];

  1143.         for(j=0; j<8; j++)

  1144.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1145.     }

  1146. }

  1147. 

  1148. //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one

  1149. //Plain C versions

  1150. #if CONFIG_RUNTIME_CPUDETECT

  1151. #  define COMPILE_C 1

  1152. #  if   ARCH_X86

  1153. #    define COMPILE_MMX     HAVE_MMX

  1154. #    define COMPILE_MMX2    HAVE_MMX2

  1155. #    define COMPILE_3DNOW   HAVE_AMD3DNOW

  1156. #  elif ARCH_PPC

  1157. #    define COMPILE_ALTIVEC HAVE_ALTIVEC

  1158. #  endif

  1159. #else /* CONFIG_RUNTIME_CPUDETECT */

  1160. #  if   ARCH_X86

  1161. #    if   HAVE_MMX2

  1162. #      define COMPILE_MMX2  1

  1163. #    elif HAVE_AMD3DNOW

  1164. #      define COMPILE_3DNOW 1

  1165. #    elif HAVE_MMX

  1166. #      define COMPILE_MMX   1

  1167. #    else

  1168. #      define COMPILE_C     1

  1169. #    endif

  1170. #  elif ARCH_PPC && HAVE_ALTIVEC

  1171. #    define COMPILE_ALTIVEC 1

  1172. #  else

  1173. #    define COMPILE_C       1

  1174. #  endif

  1175. #endif

  1176. 

  1177. #ifndef COMPILE_C

  1178. #  define COMPILE_C 0

  1179. #endif

  1180. #ifndef COMPILE_MMX

  1181. #  define COMPILE_MMX 0

  1182. #endif

  1183. #ifndef COMPILE_MMX2

  1184. #  define COMPILE_MMX2 0

  1185. #endif

  1186. #ifndef COMPILE_3DNOW

  1187. #  define COMPILE_3DNOW 0

  1188. #endif

  1189. #ifndef COMPILE_ALTIVEC

  1190. #  define COMPILE_ALTIVEC 0

  1191. #endif

  1192. 

  1193. #define COMPILE_TEMPLATE_MMX 0

  1194. #define COMPILE_TEMPLATE_MMX2 0

  1195. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1196. #define COMPILE_TEMPLATE_ALTIVEC 0

  1197. 

  1198. #include "swscale_template.c"

  1199. 

  1200. #if COMPILE_ALTIVEC

  1201. #undef RENAME

  1202. #undef COMPILE_TEMPLATE_ALTIVEC

  1203. #define COMPILE_TEMPLATE_ALTIVEC 1

  1204. #define RENAME(a) a ## _altivec

  1205. #include "ppc/swscale_template.c"

  1206. #endif

  1207. 

  1208. #if ARCH_X86

  1209. 

  1210. //MMX versions

  1211. #if COMPILE_MMX

  1212. #undef RENAME

  1213. #undef COMPILE_TEMPLATE_MMX

  1214. #undef COMPILE_TEMPLATE_MMX2

  1215. #undef COMPILE_TEMPLATE_AMD3DNOW

  1216. #define COMPILE_TEMPLATE_MMX 1

  1217. #define COMPILE_TEMPLATE_MMX2 0

  1218. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1219. #define RENAME(a) a ## _MMX

  1220. #include "x86/swscale_template.c"

  1221. #endif

  1222. 

  1223. //MMX2 versions

  1224. #if COMPILE_MMX2

  1225. #undef RENAME

  1226. #undef COMPILE_TEMPLATE_MMX

  1227. #undef COMPILE_TEMPLATE_MMX2

  1228. #undef COMPILE_TEMPLATE_AMD3DNOW

  1229. #define COMPILE_TEMPLATE_MMX 1

  1230. #define COMPILE_TEMPLATE_MMX2 1

  1231. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1232. #define RENAME(a) a ## _MMX2

  1233. #include "x86/swscale_template.c"

  1234. #endif

  1235. 

  1236. //3DNOW versions

  1237. #if COMPILE_3DNOW

  1238. #undef RENAME

  1239. #undef COMPILE_TEMPLATE_MMX

  1240. #undef COMPILE_TEMPLATE_MMX2

  1241. #undef COMPILE_TEMPLATE_AMD3DNOW

  1242. #define COMPILE_TEMPLATE_MMX 1

  1243. #define COMPILE_TEMPLATE_MMX2 0

  1244. #define COMPILE_TEMPLATE_AMD3DNOW 1

  1245. #define RENAME(a) a ## _3DNow

  1246. #include "x86/swscale_template.c"

  1247. #endif

  1248. 

  1249. #endif //ARCH_X86

  1250. 

  1251. SwsFunc ff_getSwsFunc(SwsContext *c)

  1252. {

  1253.     sws_init_swScale_c(c);

  1254. 

  1255. #if CONFIG_RUNTIME_CPUDETECT

  1256. #if ARCH_X86

  1257.     // ordered per speed fastest first

  1258.     if (c->flags & SWS_CPU_CAPS_MMX2) {

  1259.         sws_init_swScale_MMX2(c);

  1260.         return swScale_MMX2;

  1261.     } else if (c->flags & SWS_CPU_CAPS_3DNOW) {

  1262.         sws_init_swScale_3DNow(c);

  1263.         return swScale_3DNow;

  1264.     } else if (c->flags & SWS_CPU_CAPS_MMX) {

  1265.         sws_init_swScale_MMX(c);

  1266.         return swScale_MMX;

  1267.     }

  1268. 

  1269. #else

  1270. #if COMPILE_ALTIVEC

  1271.     if (c->flags & SWS_CPU_CAPS_ALTIVEC) {

  1272.         sws_init_swScale_altivec(c);

  1273.         return swScale_altivec;

  1274.     }

  1275. #endif

  1276. #endif /* ARCH_X86 */

  1277. #else //CONFIG_RUNTIME_CPUDETECT

  1278. #if   COMPILE_TEMPLATE_MMX2

  1279.     sws_init_swScale_MMX2(c);

  1280.     return swScale_MMX2;

  1281. #elif COMPILE_TEMPLATE_AMD3DNOW

  1282.     sws_init_swScale_3DNow(c);

  1283.     return swScale_3DNow;

  1284. #elif COMPILE_TEMPLATE_MMX

  1285.     sws_init_swScale_MMX(c);

  1286.     return swScale_MMX;

  1287. #elif COMPILE_TEMPLATE_ALTIVEC

  1288.     sws_init_swScale_altivec(c);

  1289.     return swScale_altivec;

  1290. #endif

  1291. #endif //!CONFIG_RUNTIME_CPUDETECT

  1292. 

  1293.     return swScale_c;

  1294. }

  1295. 

  1296. static void copyPlane(const uint8_t *src, int srcStride,

  1297.                       int srcSliceY, int srcSliceH, int width,

  1298.                       uint8_t *dst, int dstStride)

  1299. {

  1300.     dst += dstStride * srcSliceY;

  1301.     if (dstStride == srcStride && srcStride > 0) {

  1302.         memcpy(dst, src, srcSliceH * dstStride);

  1303.     } else {

  1304.         int i;

  1305.         for (i=0; i<srcSliceH; i++) {

  1306.             memcpy(dst, src, width);

  1307.             src += srcStride;

  1308.             dst += dstStride;

  1309.         }

  1310.     }

  1311. }

  1312. 

  1313. static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1314.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1315. {

  1316.     uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1317. 

  1318.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  1319.               dstParam[0], dstStride[0]);

  1320. 

  1321.     if (c->dstFormat == PIX_FMT_NV12)

  1322.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  1323.     else

  1324.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  1325. 

  1326.     return srcSliceH;

  1327. }

  1328. 

  1329. static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1330.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1331. {

  1332.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1333. 

  1334.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1335. 

  1336.     return srcSliceH;

  1337. }

  1338. 

  1339. static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1340.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1341. {

  1342.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1343. 

  1344.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1345. 

  1346.     return srcSliceH;

  1347. }

  1348. 

  1349. static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1350.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1351. {

  1352.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1353. 

  1354.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1355. 

  1356.     return srcSliceH;

  1357. }

  1358. 

  1359. static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1360.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1361. {

  1362.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1363. 

  1364.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1365. 

  1366.     return srcSliceH;

  1367. }

  1368. 

  1369. static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1370.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1371. {

  1372.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1373.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1374.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1375. 

  1376.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1377. 

  1378.     if (dstParam[3])

  1379.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1380. 

  1381.     return srcSliceH;

  1382. }

  1383. 

  1384. static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1385.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1386. {

  1387.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1388.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1389.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1390. 

  1391.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1392. 

  1393.     return srcSliceH;

  1394. }

  1395. 

  1396. static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1397.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1398. {

  1399.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1400.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1401.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1402. 

  1403.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1404. 

  1405.     if (dstParam[3])

  1406.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1407. 

  1408.     return srcSliceH;

  1409. }

  1410. 

  1411. static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1412.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1413. {

  1414.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1415.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1416.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1417. 

  1418.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1419. 

  1420.     return srcSliceH;

  1421. }

  1422. 

  1423. static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1424. {

  1425.     long i;

  1426.     for (i=0; i<num_pixels; i++)

  1427.         ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);

  1428. }

  1429. 

  1430. static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1431. {

  1432.     long i;

  1433. 

  1434.     for (i=0; i<num_pixels; i++)

  1435.         ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];

  1436. }

  1437. 

  1438. static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1439. {

  1440.     long i;

  1441. 

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

  1443.         //FIXME slow?

  1444.         dst[0]= palette[src[i<<1]*4+0];

  1445.         dst[1]= palette[src[i<<1]*4+1];

  1446.         dst[2]= palette[src[i<<1]*4+2];

  1447.         dst+= 3;

  1448.     }

  1449. }

  1450. 

  1451. static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1452.                            int srcSliceH, uint8_t* dst[], int dstStride[])

  1453. {

  1454.     const enum PixelFormat srcFormat= c->srcFormat;

  1455.     const enum PixelFormat dstFormat= c->dstFormat;

  1456.     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,

  1457.                  const uint8_t *palette)=NULL;

  1458.     int i;

  1459.     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1460.     const uint8_t *srcPtr= src[0];

  1461. 

  1462.     if (srcFormat == PIX_FMT_Y400A) {

  1463.         switch (dstFormat) {

  1464.         case PIX_FMT_RGB32  : conv = gray8aToPacked32; break;

  1465.         case PIX_FMT_BGR32  : conv = gray8aToPacked32; break;

  1466.         case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;

  1467.         case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;

  1468.         case PIX_FMT_RGB24  : conv = gray8aToPacked24; break;

  1469.         case PIX_FMT_BGR24  : conv = gray8aToPacked24; break;

  1470.         }

  1471.     } else if (usePal(srcFormat)) {

  1472.         switch (dstFormat) {

  1473.         case PIX_FMT_RGB32  : conv = sws_convertPalette8ToPacked32; break;

  1474.         case PIX_FMT_BGR32  : conv = sws_convertPalette8ToPacked32; break;

  1475.         case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;

  1476.         case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;

  1477.         case PIX_FMT_RGB24  : conv = sws_convertPalette8ToPacked24; break;

  1478.         case PIX_FMT_BGR24  : conv = sws_convertPalette8ToPacked24; break;

  1479.         }

  1480.     }

  1481. 

  1482.     if (!conv)

  1483.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1484.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1485.     else {

  1486.         for (i=0; i<srcSliceH; i++) {

  1487.             conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  1488.             srcPtr+= srcStride[0];

  1489.             dstPtr+= dstStride[0];

  1490.         }

  1491.     }

  1492. 

  1493.     return srcSliceH;

  1494. }

  1495. 

  1496. #define isRGBA32(x) (            \

  1497.            (x) == PIX_FMT_ARGB   \

  1498.         || (x) == PIX_FMT_RGBA   \

  1499.         || (x) == PIX_FMT_BGRA   \

  1500.         || (x) == PIX_FMT_ABGR   \

  1501.         )

  1502. 

  1503. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

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

  1505.                            int srcSliceH, uint8_t* dst[], int dstStride[])

  1506. {

  1507.     const enum PixelFormat srcFormat= c->srcFormat;

  1508.     const enum PixelFormat dstFormat= c->dstFormat;

  1509.     const int srcBpp= (c->srcFormatBpp + 7) >> 3;

  1510.     const int dstBpp= (c->dstFormatBpp + 7) >> 3;

  1511.     const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1512.     const int dstId= c->dstFormatBpp >> 2;

  1513.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  1514. 

  1515. #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)

  1516. 

  1517.     if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {

  1518.         if (     CONV_IS(ABGR, RGBA)

  1519.               || CONV_IS(ARGB, BGRA)

  1520.               || CONV_IS(BGRA, ARGB)

  1521.               || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;

  1522.         else if (CONV_IS(ABGR, ARGB)

  1523.               || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;

  1524.         else if (CONV_IS(ABGR, BGRA)

  1525.               || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;

  1526.         else if (CONV_IS(BGRA, RGBA)

  1527.               || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;

  1528.         else if (CONV_IS(BGRA, ABGR)

  1529.               || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;

  1530.     } else

  1531.     /* BGR -> BGR */

  1532.     if (  (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))

  1533.        || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {

  1534.         switch(srcId | (dstId<<4)) {

  1535.         case 0x34: conv= rgb16to15; break;

  1536.         case 0x36: conv= rgb24to15; break;

  1537.         case 0x38: conv= rgb32to15; break;

  1538.         case 0x43: conv= rgb15to16; break;

  1539.         case 0x46: conv= rgb24to16; break;

  1540.         case 0x48: conv= rgb32to16; break;

  1541.         case 0x63: conv= rgb15to24; break;

  1542.         case 0x64: conv= rgb16to24; break;

  1543.         case 0x68: conv= rgb32to24; break;

  1544.         case 0x83: conv= rgb15to32; break;

  1545.         case 0x84: conv= rgb16to32; break;

  1546.         case 0x86: conv= rgb24to32; break;

  1547.         }

  1548.     } else if (  (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))

  1549.              || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {

  1550.         switch(srcId | (dstId<<4)) {

  1551.         case 0x33: conv= rgb15tobgr15; break;

  1552.         case 0x34: conv= rgb16tobgr15; break;

  1553.         case 0x36: conv= rgb24tobgr15; break;

  1554.         case 0x38: conv= rgb32tobgr15; break;

  1555.         case 0x43: conv= rgb15tobgr16; break;

  1556.         case 0x44: conv= rgb16tobgr16; break;

  1557.         case 0x46: conv= rgb24tobgr16; break;

  1558.         case 0x48: conv= rgb32tobgr16; break;

  1559.         case 0x63: conv= rgb15tobgr24; break;

  1560.         case 0x64: conv= rgb16tobgr24; break;

  1561.         case 0x66: conv= rgb24tobgr24; break;

  1562.         case 0x68: conv= rgb32tobgr24; break;

  1563.         case 0x83: conv= rgb15tobgr32; break;

  1564.         case 0x84: conv= rgb16tobgr32; break;

  1565.         case 0x86: conv= rgb24tobgr32; break;

  1566.         }

  1567.     }

  1568. 

  1569.     if (!conv) {

  1570.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1571.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1572.     } else {

  1573.         const uint8_t *srcPtr= src[0];

  1574.               uint8_t *dstPtr= dst[0];

  1575.         if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))

  1576.             srcPtr += ALT32_CORR;

  1577. 

  1578.         if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))

  1579.             dstPtr += ALT32_CORR;

  1580. 

  1581.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  1582.             conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  1583.         else {

  1584.             int i;

  1585.             dstPtr += dstStride[0]*srcSliceY;

  1586. 

  1587.             for (i=0; i<srcSliceH; i++) {

  1588.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  1589.                 srcPtr+= srcStride[0];

  1590.                 dstPtr+= dstStride[0];

  1591.             }

  1592.         }

  1593.     }

  1594.     return srcSliceH;

  1595. }

  1596. 

  1597. static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1598.                               int srcSliceH, uint8_t* dst[], int dstStride[])

  1599. {

  1600.     rgb24toyv12(

  1601.         src[0],

  1602.         dst[0]+ srcSliceY    *dstStride[0],

  1603.         dst[1]+(srcSliceY>>1)*dstStride[1],

  1604.         dst[2]+(srcSliceY>>1)*dstStride[2],

  1605.         c->srcW, srcSliceH,

  1606.         dstStride[0], dstStride[1], srcStride[0]);

  1607.     if (dst[3])

  1608.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1609.     return srcSliceH;

  1610. }

  1611. 

  1612. static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1613.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1614. {

  1615.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  1616.               dst[0], dstStride[0]);

  1617. 

  1618.     planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  1619.              srcSliceH >> 2, srcStride[1], dstStride[1]);

  1620.     planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  1621.              srcSliceH >> 2, srcStride[2], dstStride[2]);

  1622.     if (dst[3])

  1623.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1624.     return srcSliceH;

  1625. }

  1626. 

  1627. /* unscaled copy like stuff (assumes nearly identical formats) */

  1628. static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1629.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1630. {

  1631.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1632.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  1633.     else {

  1634.         int i;

  1635.         const uint8_t *srcPtr= src[0];

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

  1637.         int length=0;

  1638. 

  1639.         /* universal length finder */

  1640.         while(length+c->srcW <= FFABS(dstStride[0])

  1641.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  1642.         assert(length!=0);

  1643. 

  1644.         for (i=0; i<srcSliceH; i++) {

  1645.             memcpy(dstPtr, srcPtr, length);

  1646.             srcPtr+= srcStride[0];

  1647.             dstPtr+= dstStride[0];

  1648.         }

  1649.     }

  1650.     return srcSliceH;

  1651. }

  1652. 

  1653. static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1654.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1655. {

  1656.     int plane, i, j;

  1657.     for (plane=0; plane<4; plane++) {

  1658.         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  1659.         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  1660.         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  1661.         const uint8_t *srcPtr= src[plane];

  1662.         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  1663. 

  1664.         if (!dst[plane]) continue;

  1665.         // ignore palette for GRAY8

  1666.         if (plane == 1 && !dst[2]) continue;

  1667.         if (!src[plane] || (plane == 1 && !src[2])) {

  1668.             if(is16BPS(c->dstFormat))

  1669.                 length*=2;

  1670.             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);

  1671.         } else {

  1672.             if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {

  1673.                 if (!isBE(c->srcFormat)) srcPtr++;

  1674.                 for (i=0; i<height; i++) {

  1675.                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  1676.                     srcPtr+= srcStride[plane];

  1677.                     dstPtr+= dstStride[plane];

  1678.                 }

  1679.             } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {

  1680.                 for (i=0; i<height; i++) {

  1681.                     for (j=0; j<length; j++) {

  1682.                         dstPtr[ j<<1   ] = srcPtr[j];

  1683.                         dstPtr[(j<<1)+1] = srcPtr[j];

  1684.                     }

  1685.                     srcPtr+= srcStride[plane];

  1686.                     dstPtr+= dstStride[plane];

  1687.                 }

  1688.             } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)

  1689.                   && isBE(c->srcFormat) != isBE(c->dstFormat)) {

  1690. 

  1691.                 for (i=0; i<height; i++) {

  1692.                     for (j=0; j<length; j++)

  1693.                         ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);

  1694.                     srcPtr+= srcStride[plane];

  1695.                     dstPtr+= dstStride[plane];

  1696.                 }

  1697.             } else if (dstStride[plane] == srcStride[plane] &&

  1698.                        srcStride[plane] > 0 && srcStride[plane] == length) {

  1699.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane],

  1700.                        height*dstStride[plane]);

  1701.             } else {

  1702.                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))

  1703.                     length*=2;

  1704.                 for (i=0; i<height; i++) {

  1705.                     memcpy(dstPtr, srcPtr, length);

  1706.                     srcPtr+= srcStride[plane];

  1707.                     dstPtr+= dstStride[plane];

  1708.                 }

  1709.             }

  1710.         }

  1711.     }

  1712.     return srcSliceH;

  1713. }

  1714. 

  1715. int ff_hardcodedcpuflags(void)

  1716. {

  1717.     int flags = 0;

  1718. #if   COMPILE_TEMPLATE_MMX2

  1719.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  1720. #elif COMPILE_TEMPLATE_AMD3DNOW

  1721.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  1722. #elif COMPILE_TEMPLATE_MMX

  1723.     flags |= SWS_CPU_CAPS_MMX;

  1724. #elif COMPILE_TEMPLATE_ALTIVEC

  1725.     flags |= SWS_CPU_CAPS_ALTIVEC;

  1726. #elif ARCH_BFIN

  1727.     flags |= SWS_CPU_CAPS_BFIN;

  1728. #endif

  1729.     return flags;

  1730. }

  1731. 

  1732. void ff_get_unscaled_swscale(SwsContext *c)

  1733. {

  1734.     const enum PixelFormat srcFormat = c->srcFormat;

  1735.     const enum PixelFormat dstFormat = c->dstFormat;

  1736.     const int flags = c->flags;

  1737.     const int dstH = c->dstH;

  1738.     int needsDither;

  1739. 

  1740.     needsDither= isAnyRGB(dstFormat)

  1741.         &&  c->dstFormatBpp < 24

  1742.         && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));

  1743. 

  1744.     /* yv12_to_nv12 */

  1745.     if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {

  1746.         c->swScale= planarToNv12Wrapper;

  1747.     }

  1748.     /* yuv2bgr */

  1749.     if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)

  1750.         && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {

  1751.         c->swScale= ff_yuv2rgb_get_func_ptr(c);

  1752.     }

  1753. 

  1754.     if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {

  1755.         c->swScale= yvu9ToYv12Wrapper;

  1756.     }

  1757. 

  1758.     /* bgr24toYV12 */

  1759.     if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))

  1760.         c->swScale= bgr24ToYv12Wrapper;

  1761. 

  1762.     /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  1763.     if (   isAnyRGB(srcFormat)

  1764.         && isAnyRGB(dstFormat)

  1765.         && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  1766.         && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  1767.         && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  1768.         && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  1769.         && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  1770.         && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  1771.         && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  1772.         && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE

  1773.         && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE

  1774.         && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE

  1775.         && srcFormat != PIX_FMT_BGR48LE   && dstFormat != PIX_FMT_BGR48LE

  1776.         && srcFormat != PIX_FMT_BGR48BE   && dstFormat != PIX_FMT_BGR48BE

  1777.         && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  1778.         c->swScale= rgbToRgbWrapper;

  1779. 

  1780.     if ((usePal(srcFormat) && (

  1781.         dstFormat == PIX_FMT_RGB32   ||

  1782.         dstFormat == PIX_FMT_RGB32_1 ||

  1783.         dstFormat == PIX_FMT_RGB24   ||

  1784.         dstFormat == PIX_FMT_BGR32   ||

  1785.         dstFormat == PIX_FMT_BGR32_1 ||

  1786.         dstFormat == PIX_FMT_BGR24)))

  1787.         c->swScale= palToRgbWrapper;

  1788. 

  1789.     if (srcFormat == PIX_FMT_YUV422P) {

  1790.         if (dstFormat == PIX_FMT_YUYV422)

  1791.             c->swScale= yuv422pToYuy2Wrapper;

  1792.         else if (dstFormat == PIX_FMT_UYVY422)

  1793.             c->swScale= yuv422pToUyvyWrapper;

  1794.     }

  1795. 

  1796.     /* LQ converters if -sws 0 or -sws 4*/

  1797.     if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {

  1798.         /* yv12_to_yuy2 */

  1799.         if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {

  1800.             if (dstFormat == PIX_FMT_YUYV422)

  1801.                 c->swScale= planarToYuy2Wrapper;

  1802.             else if (dstFormat == PIX_FMT_UYVY422)

  1803.                 c->swScale= planarToUyvyWrapper;

  1804.         }

  1805.     }

  1806.     if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  1807.         c->swScale= yuyvToYuv420Wrapper;

  1808.     if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  1809.         c->swScale= uyvyToYuv420Wrapper;

  1810.     if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)

  1811.         c->swScale= yuyvToYuv422Wrapper;

  1812.     if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)

  1813.         c->swScale= uyvyToYuv422Wrapper;

  1814. 

  1815. #if COMPILE_ALTIVEC

  1816.     if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  1817.         !(c->flags & SWS_BITEXACT) &&

  1818.         srcFormat == PIX_FMT_YUV420P) {

  1819.         // unscaled YV12 -> packed YUV, we want speed

  1820.         if (dstFormat == PIX_FMT_YUYV422)

  1821.             c->swScale= yv12toyuy2_unscaled_altivec;

  1822.         else if (dstFormat == PIX_FMT_UYVY422)

  1823.             c->swScale= yv12touyvy_unscaled_altivec;

  1824.     }

  1825. #endif

  1826. 

  1827.     /* simple copy */

  1828.     if (  srcFormat == dstFormat

  1829.         || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)

  1830.         || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)

  1831.         || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  1832.         || (isPlanarYUV(dstFormat) && isGray(srcFormat))

  1833.         || (isGray(dstFormat) && isGray(srcFormat))

  1834.         || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)

  1835.             && c->chrDstHSubSample == c->chrSrcHSubSample

  1836.             && c->chrDstVSubSample == c->chrSrcVSubSample

  1837.             && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21

  1838.             && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))

  1839.     {

  1840.         if (isPacked(c->srcFormat))

  1841.             c->swScale= packedCopyWrapper;

  1842.         else /* Planar YUV or gray */

  1843.             c->swScale= planarCopyWrapper;

  1844.     }

  1845. #if ARCH_BFIN

  1846.     if (flags & SWS_CPU_CAPS_BFIN)

  1847.         ff_bfin_get_unscaled_swscale (c);

  1848. #endif

  1849. }

  1850. 

  1851. static void reset_ptr(const uint8_t* src[], int format)

  1852. {

  1853.     if(!isALPHA(format))

  1854.         src[3]=NULL;

  1855.     if(!isPlanarYUV(format)) {

  1856.         src[3]=src[2]=NULL;

  1857. 

  1858.         if (!usePal(format))

  1859.             src[1]= NULL;

  1860.     }

  1861. }

  1862. 

  1863. static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,

  1864.                                 const int linesizes[4])

  1865. {

  1866.     const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];

  1867.     int i;

  1868. 

  1869.     for (i = 0; i < 4; i++) {

  1870.         int plane = desc->comp[i].plane;

  1871.         if (!data[plane] || !linesizes[plane])

  1872.             return 0;

  1873.     }

  1874. 

  1875.     return 1;

  1876. }

  1877. 

  1878. /**

  1879.  * swscale wrapper, so we don't need to export the SwsContext.

  1880.  * Assumes planar YUV to be in YUV order instead of YVU.

  1881.  */

  1882. int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,

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

  1884. {

  1885.     int i;

  1886.     const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};

  1887.     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};

  1888. 

  1889.     // do not mess up sliceDir if we have a "trailing" 0-size slice

  1890.     if (srcSliceH == 0)

  1891.         return 0;

  1892. 

  1893.     if (!check_image_pointers(src, c->srcFormat, srcStride)) {

  1894.         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");

  1895.         return 0;

  1896.     }

  1897.     if (!check_image_pointers(dst, c->dstFormat, dstStride)) {

  1898.         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");

  1899.         return 0;

  1900.     }

  1901. 

  1902.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  1903.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  1904.         return 0;

  1905.     }

  1906.     if (c->sliceDir == 0) {

  1907.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  1908.     }

  1909. 

  1910.     if (usePal(c->srcFormat)) {

  1911.         for (i=0; i<256; i++) {

  1912.             int p, r, g, b,y,u,v;

  1913.             if(c->srcFormat == PIX_FMT_PAL8) {

  1914.                 p=((const uint32_t*)(src[1]))[i];

  1915.                 r= (p>>16)&0xFF;

  1916.                 g= (p>> 8)&0xFF;

  1917.                 b=  p     &0xFF;

  1918.             } else if(c->srcFormat == PIX_FMT_RGB8) {

  1919.                 r= (i>>5    )*36;

  1920.                 g= ((i>>2)&7)*36;

  1921.                 b= (i&3     )*85;

  1922.             } else if(c->srcFormat == PIX_FMT_BGR8) {

  1923.                 b= (i>>6    )*85;

  1924.                 g= ((i>>3)&7)*36;

  1925.                 r= (i&7     )*36;

  1926.             } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {

  1927.                 r= (i>>3    )*255;

  1928.                 g= ((i>>1)&3)*85;

  1929.                 b= (i&1     )*255;

  1930.             } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {

  1931.                 r = g = b = i;

  1932.             } else {

  1933.                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);

  1934.                 b= (i>>3    )*255;

  1935.                 g= ((i>>1)&3)*85;

  1936.                 r= (i&1     )*255;

  1937.             }

  1938.             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  1939.             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  1940.             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  1941.             c->pal_yuv[i]= y + (u<<8) + (v<<16);

  1942. 

  1943.             switch(c->dstFormat) {

  1944.             case PIX_FMT_BGR32:

  1945. #if !HAVE_BIGENDIAN

  1946.             case PIX_FMT_RGB24:

  1947. #endif

  1948.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);

  1949.                 break;

  1950.             case PIX_FMT_BGR32_1:

  1951. #if HAVE_BIGENDIAN

  1952.             case PIX_FMT_BGR24:

  1953. #endif

  1954.                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;

  1955.                 break;

  1956.             case PIX_FMT_RGB32_1:

  1957. #if HAVE_BIGENDIAN

  1958.             case PIX_FMT_RGB24:

  1959. #endif

  1960.                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;

  1961.                 break;

  1962.             case PIX_FMT_RGB32:

  1963. #if !HAVE_BIGENDIAN

  1964.             case PIX_FMT_BGR24:

  1965. #endif

  1966.             default:

  1967.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);

  1968.             }

  1969.         }

  1970.     }

  1971. 

  1972.     // copy strides, so they can safely be modified

  1973.     if (c->sliceDir == 1) {

  1974.         // slices go from top to bottom

  1975.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};

  1976.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};

  1977. 

  1978.         reset_ptr(src2, c->srcFormat);

  1979.         reset_ptr((const uint8_t**)dst2, c->dstFormat);

  1980. 

  1981.         /* reset slice direction at end of frame */

  1982.         if (srcSliceY + srcSliceH == c->srcH)

  1983.             c->sliceDir = 0;

  1984. 

  1985.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);

  1986.     } else {

  1987.         // slices go from bottom to top => we flip the image internally

  1988.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};

  1989.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};

  1990. 

  1991.         src2[0] += (srcSliceH-1)*srcStride[0];

  1992.         if (!usePal(c->srcFormat))

  1993.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  1994.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  1995.         src2[3] += (srcSliceH-1)*srcStride[3];

  1996.         dst2[0] += ( c->dstH                      -1)*dstStride[0];

  1997.         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];

  1998.         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];

  1999.         dst2[3] += ( c->dstH                      -1)*dstStride[3];

  2000. 

  2001.         reset_ptr(src2, c->srcFormat);

  2002.         reset_ptr((const uint8_t**)dst2, c->dstFormat);

  2003. 

  2004.         /* reset slice direction at end of frame */

  2005.         if (!srcSliceY)

  2006.             c->sliceDir = 0;

  2007. 

  2008.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  2009.     }

  2010. }

  2011. 

  2012. /* Convert the palette to the same packed 32-bit format as the palette */

  2013. void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  2014. {

  2015.     long i;

  2016. 

  2017.     for (i=0; i<num_pixels; i++)

  2018.         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];

  2019. }

  2020. 

  2021. /* Palette format: ABCD -> dst format: ABC */

  2022. void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  2023. {

  2024.     long i;

  2025. 

  2026.     for (i=0; i<num_pixels; i++) {

  2027.         //FIXME slow?

  2028.         dst[0]= palette[src[i]*4+0];

  2029.         dst[1]= palette[src[i]*4+1];

  2030.         dst[2]= palette[src[i]*4+2];

  2031.         dst+= 3;

  2032.     }

  2033. }