libswscale/swscale.c

   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 #include <inttypes.h>
  58 #include <string.h>
  59 #include <math.h>
  60 #include <stdio.h>
  61 #include "config.h"
  62 #include <assert.h>
  63 #include "swscale.h"
  64 #include "swscale_internal.h"
  65 #include "rgb2rgb.h"
  66 #include "libavutil/intreadwrite.h"
  67 #include "libavutil/x86_cpu.h"
  68 #include "libavutil/avutil.h"
  69 #include "libavutil/bswap.h"
  70 #include "libavutil/pixdesc.h"
  71
  72 #undef MOVNTQ
  73 #undef PAVGB
  74
  75 //#undef HAVE_MMX2
  76 //#define HAVE_AMD3DNOW
  77 //#undef HAVE_MMX
  78 //#undef ARCH_X86
  79 #define DITHER1XBPP
  80
  81 #define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
  82
  83 #ifdef M_PI
  84 #define PI M_PI
  85 #else
  86 #define PI 3.14159265358979323846
  87 #endif
  88
  89 #define isPacked(x)         (       \
  90            (x)==PIX_FMT_PAL8        \
  91         || (x)==PIX_FMT_YUYV422     \
  92         || (x)==PIX_FMT_UYVY422     \
  93         || isAnyRGB(x)              \
  94     )
  95 #define usePal(x) (av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL)
  96
  97 #define RGB2YUV_SHIFT 15
  98 #define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
  99 #define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 100 #define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 101 #define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
 102 #define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 103 #define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 104 #define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
 105 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 106 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 107
 108 static const double rgb2yuv_table[8][9]={
 109     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
 110     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
 111     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
 112     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
 113     {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
 114     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
 115     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
 116     {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
 117 };
 118
 119 /*
 120 NOTES
 121 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
 122
 123 TODO
 124 more intelligent misalignment avoidance for the horizontal scaler
 125 write special vertical cubic upscale version
 126 optimize C code (YV12 / minmax)
 127 add support for packed pixel YUV input & output
 128 add support for Y8 output
 129 optimize BGR24 & BGR32
 130 add BGR4 output support
 131 write special BGR->BGR scaler
 132 */
 133
 134 #if ARCH_X86 && CONFIG_GPL
 135 DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
 136 DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
 137 DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
 138 DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
 139 DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
 140 DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
 141 DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
 142 DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
 143
 144 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
 145         0x0103010301030103LL,
 146         0x0200020002000200LL,};
 147
 148 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
 149         0x0602060206020602LL,
 150         0x0004000400040004LL,};
 151
 152 DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
 153 DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
 154 DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
 155 DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
 156 DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
 157 DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
 158
 159 DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
 160 DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
 161 DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
 162
 163 #ifdef FAST_BGR2YV12
 164 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
 165 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
 166 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
 167 #else
 168 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
 169 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
 170 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
 171 #endif /* FAST_BGR2YV12 */
 172 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
 173 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
 174 DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
 175
 176 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
 177 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
 178 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
 179 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
 180 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
 181
 182 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
 183     {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
 184     {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
 185 };
 186
 187 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
 188
 189 #endif /* ARCH_X86 && CONFIG_GPL */
 190
 191 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
 192 {  1,   3,   1,   3,   1,   3,   1,   3, },
 193 {  2,   0,   2,   0,   2,   0,   2,   0, },
 194 };
 195
 196 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
 197 {  6,   2,   6,   2,   6,   2,   6,   2, },
 198 {  0,   4,   0,   4,   0,   4,   0,   4, },
 199 };
 200
 201 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
 202 { 17,   9,  23,  15,  16,   8,  22,  14, },
 203 {  5,  29,   3,  27,   4,  28,   2,  26, },
 204 { 21,  13,  19,  11,  20,  12,  18,  10, },
 205 {  0,  24,   6,  30,   1,  25,   7,  31, },
 206 { 16,   8,  22,  14,  17,   9,  23,  15, },
 207 {  4,  28,   2,  26,   5,  29,   3,  27, },
 208 { 20,  12,  18,  10,  21,  13,  19,  11, },
 209 {  1,  25,   7,  31,   0,  24,   6,  30, },
 210 };
 211
 212 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
 213 {  0,  55,  14,  68,   3,  58,  17,  72, },
 214 { 37,  18,  50,  32,  40,  22,  54,  35, },
 215 {  9,  64,   5,  59,  13,  67,   8,  63, },
 216 { 46,  27,  41,  23,  49,  31,  44,  26, },
 217 {  2,  57,  16,  71,   1,  56,  15,  70, },
 218 { 39,  21,  52,  34,  38,  19,  51,  33, },
 219 { 11,  66,   7,  62,  10,  65,   6,  60, },
 220 { 48,  30,  43,  25,  47,  29,  42,  24, },
 221 };
 222
 223 #if 1
 224 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 225 {117,  62, 158, 103, 113,  58, 155, 100, },
 226 { 34, 199,  21, 186,  31, 196,  17, 182, },
 227 {144,  89, 131,  76, 141,  86, 127,  72, },
 228 {  0, 165,  41, 206,  10, 175,  52, 217, },
 229 {110,  55, 151,  96, 120,  65, 162, 107, },
 230 { 28, 193,  14, 179,  38, 203,  24, 189, },
 231 {138,  83, 124,  69, 148,  93, 134,  79, },
 232 {  7, 172,  48, 213,   3, 168,  45, 210, },
 233 };
 234 #elif 1
 235 // tries to correct a gamma of 1.5
 236 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 237 {  0, 143,  18, 200,   2, 156,  25, 215, },
 238 { 78,  28, 125,  64,  89,  36, 138,  74, },
 239 { 10, 180,   3, 161,  16, 195,   8, 175, },
 240 {109,  51,  93,  38, 121,  60, 105,  47, },
 241 {  1, 152,  23, 210,   0, 147,  20, 205, },
 242 { 85,  33, 134,  71,  81,  30, 130,  67, },
 243 { 14, 190,   6, 171,  12, 185,   5, 166, },
 244 {117,  57, 101,  44, 113,  54,  97,  41, },
 245 };
 246 #elif 1
 247 // tries to correct a gamma of 2.0
 248 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 249 {  0, 124,   8, 193,   0, 140,  12, 213, },
 250 { 55,  14, 104,  42,  66,  19, 119,  52, },
 251 {  3, 168,   1, 145,   6, 187,   3, 162, },
 252 { 86,  31,  70,  21,  99,  39,  82,  28, },
 253 {  0, 134,  11, 206,   0, 129,   9, 200, },
 254 { 62,  17, 114,  48,  58,  16, 109,  45, },
 255 {  5, 181,   2, 157,   4, 175,   1, 151, },
 256 { 95,  36,  78,  26,  90,  34,  74,  24, },
 257 };
 258 #else
 259 // tries to correct a gamma of 2.5
 260 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 261 {  0, 107,   3, 187,   0, 125,   6, 212, },
 262 { 39,   7,  86,  28,  49,  11, 102,  36, },
 263 {  1, 158,   0, 131,   3, 180,   1, 151, },
 264 { 68,  19,  52,  12,  81,  25,  64,  17, },
 265 {  0, 119,   5, 203,   0, 113,   4, 195, },
 266 { 45,   9,  96,  33,  42,   8,  91,  30, },
 267 {  2, 172,   1, 144,   2, 165,   0, 137, },
 268 { 77,  23,  60,  15,  72,  21,  56,  14, },
 269 };
 270 #endif
 271
 272 static av_always_inline void yuv2yuvX16inC_template(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,
 275                                                     int dstW, int chrDstW, int big_endian)
 276 {
 277     //FIXME Optimize (just quickly written not optimized..)
 278     int i;
 279
 280     for (i = 0; i < dstW; i++) {
 281         int val = 1 << 10;
 282         int j;
 283
 284         for (j = 0; j < lumFilterSize; j++)
 285             val += lumSrc[j][i] * lumFilter[j];
 286
 287         if (big_endian) {
 288             AV_WB16(&dest[i], av_clip_uint16(val >> 11));
 289         } else {
 290             AV_WL16(&dest[i], av_clip_uint16(val >> 11));
 291         }
 292     }
 293
 294     if (uDest) {
 295         for (i = 0; i < chrDstW; i++) {
 296             int u = 1 << 10;
 297             int v = 1 << 10;
 298             int j;
 299
 300             for (j = 0; j < chrFilterSize; j++) {
 301                 u += chrSrc[j][i       ] * chrFilter[j];
 302                 v += chrSrc[j][i + VOFW] * chrFilter[j];
 303             }
 304
 305             if (big_endian) {
 306                 AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
 307                 AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
 308             } else {
 309                 AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
 310                 AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
 311             }
 312         }
 313     }
 314
 315     if (CONFIG_SWSCALE_ALPHA && aDest) {
 316         for (i = 0; i < dstW; i++) {
 317             int val = 1 << 10;
 318             int j;
 319
 320             for (j = 0; j < lumFilterSize; j++)
 321                 val += alpSrc[j][i] * lumFilter[j];
 322
 323             if (big_endian) {
 324                 AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
 325             } else {
 326                 AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
 327             }
 328         }
 329     }
 330 }
 331
 332 static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 333                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 334                                  const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
 335                                  enum PixelFormat dstFormat)
 336 {
 337     if (isBE(dstFormat)) {
 338         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
 339                                chrFilter, chrSrc, chrFilterSize,
 340                                alpSrc,
 341                                dest, uDest, vDest, aDest,
 342                                dstW, chrDstW, 1);
 343     } else {
 344         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
 345                                chrFilter, chrSrc, chrFilterSize,
 346                                alpSrc,
 347                                dest, uDest, vDest, aDest,
 348                                dstW, chrDstW, 0);
 349     }
 350 }
 351
 352 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 353                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 354                                const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
 355 {
 356     //FIXME Optimize (just quickly written not optimized..)
 357     int i;
 358     for (i=0; i<dstW; i++) {
 359         int val=1<<18;
 360         int j;
 361         for (j=0; j<lumFilterSize; j++)
 362             val += lumSrc[j][i] * lumFilter[j];
 363
 364         dest[i]= av_clip_uint8(val>>19);
 365     }
 366
 367     if (uDest)
 368         for (i=0; i<chrDstW; i++) {
 369             int u=1<<18;
 370             int v=1<<18;
 371             int j;
 372             for (j=0; j<chrFilterSize; j++) {
 373                 u += chrSrc[j][i] * chrFilter[j];
 374                 v += chrSrc[j][i + VOFW] * chrFilter[j];
 375             }
 376
 377             uDest[i]= av_clip_uint8(u>>19);
 378             vDest[i]= av_clip_uint8(v>>19);
 379         }
 380
 381     if (CONFIG_SWSCALE_ALPHA && aDest)
 382         for (i=0; i<dstW; i++) {
 383             int val=1<<18;
 384             int j;
 385             for (j=0; j<lumFilterSize; j++)
 386                 val += alpSrc[j][i] * lumFilter[j];
 387
 388             aDest[i]= av_clip_uint8(val>>19);
 389         }
 390
 391 }
 392
 393 static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 394                                 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 395                                 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
 396 {
 397     //FIXME Optimize (just quickly written not optimized..)
 398     int i;
 399     for (i=0; i<dstW; i++) {
 400         int val=1<<18;
 401         int j;
 402         for (j=0; j<lumFilterSize; j++)
 403             val += lumSrc[j][i] * lumFilter[j];
 404
 405         dest[i]= av_clip_uint8(val>>19);
 406     }
 407
 408     if (!uDest)
 409         return;
 410
 411     if (dstFormat == PIX_FMT_NV12)
 412         for (i=0; i<chrDstW; i++) {
 413             int u=1<<18;
 414             int v=1<<18;
 415             int j;
 416             for (j=0; j<chrFilterSize; j++) {
 417                 u += chrSrc[j][i] * chrFilter[j];
 418                 v += chrSrc[j][i + VOFW] * chrFilter[j];
 419             }
 420
 421             uDest[2*i]= av_clip_uint8(u>>19);
 422             uDest[2*i+1]= av_clip_uint8(v>>19);
 423         }
 424     else
 425         for (i=0; i<chrDstW; i++) {
 426             int u=1<<18;
 427             int v=1<<18;
 428             int j;
 429             for (j=0; j<chrFilterSize; j++) {
 430                 u += chrSrc[j][i] * chrFilter[j];
 431                 v += chrSrc[j][i + VOFW] * chrFilter[j];
 432             }
 433
 434             uDest[2*i]= av_clip_uint8(v>>19);
 435             uDest[2*i+1]= av_clip_uint8(u>>19);
 436         }
 437 }
 438
 439 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
 440     for (i=0; i<(dstW>>1); i++) {\
 441         int j;\
 442         int Y1 = 1<<18;\
 443         int Y2 = 1<<18;\
 444         int U  = 1<<18;\
 445         int V  = 1<<18;\
 446         int av_unused A1, A2;\
 447         type av_unused *r, *b, *g;\
 448         const int i2= 2*i;\
 449         \
 450         for (j=0; j<lumFilterSize; j++) {\
 451             Y1 += lumSrc[j][i2] * lumFilter[j];\
 452             Y2 += lumSrc[j][i2+1] * lumFilter[j];\
 453         }\
 454         for (j=0; j<chrFilterSize; j++) {\
 455             U += chrSrc[j][i] * chrFilter[j];\
 456             V += chrSrc[j][i+VOFW] * chrFilter[j];\
 457         }\
 458         Y1>>=19;\
 459         Y2>>=19;\
 460         U >>=19;\
 461         V >>=19;\
 462         if (alpha) {\
 463             A1 = 1<<18;\
 464             A2 = 1<<18;\
 465             for (j=0; j<lumFilterSize; j++) {\
 466                 A1 += alpSrc[j][i2  ] * lumFilter[j];\
 467                 A2 += alpSrc[j][i2+1] * lumFilter[j];\
 468             }\
 469             A1>>=19;\
 470             A2>>=19;\
 471         }
 472
 473 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
 474         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
 475         if ((Y1|Y2|U|V)&256) {\
 476             if (Y1>255)   Y1=255; \
 477             else if (Y1<0)Y1=0;   \
 478             if (Y2>255)   Y2=255; \
 479             else if (Y2<0)Y2=0;   \
 480             if (U>255)    U=255;  \
 481             else if (U<0) U=0;    \
 482             if (V>255)    V=255;  \
 483             else if (V<0) V=0;    \
 484         }\
 485         if (alpha && ((A1|A2)&256)) {\
 486             A1=av_clip_uint8(A1);\
 487             A2=av_clip_uint8(A2);\
 488         }
 489
 490 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
 491     for (i=0; i<dstW; i++) {\
 492         int j;\
 493         int Y = 0;\
 494         int U = -128<<19;\
 495         int V = -128<<19;\
 496         int av_unused A;\
 497         int R,G,B;\
 498         \
 499         for (j=0; j<lumFilterSize; j++) {\
 500             Y += lumSrc[j][i     ] * lumFilter[j];\
 501         }\
 502         for (j=0; j<chrFilterSize; j++) {\
 503             U += chrSrc[j][i     ] * chrFilter[j];\
 504             V += chrSrc[j][i+VOFW] * chrFilter[j];\
 505         }\
 506         Y >>=10;\
 507         U >>=10;\
 508         V >>=10;\
 509         if (alpha) {\
 510             A = rnd;\
 511             for (j=0; j<lumFilterSize; j++)\
 512                 A += alpSrc[j][i     ] * lumFilter[j];\
 513             A >>=19;\
 514             if (A&256)\
 515                 A = av_clip_uint8(A);\
 516         }
 517
 518 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
 519     YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
 520         Y-= c->yuv2rgb_y_offset;\
 521         Y*= c->yuv2rgb_y_coeff;\
 522         Y+= rnd;\
 523         R= Y + V*c->yuv2rgb_v2r_coeff;\
 524         G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
 525         B= Y +                          U*c->yuv2rgb_u2b_coeff;\
 526         if ((R|G|B)&(0xC0000000)) {\
 527             if (R>=(256<<22))   R=(256<<22)-1; \
 528             else if (R<0)R=0;   \
 529             if (G>=(256<<22))   G=(256<<22)-1; \
 530             else if (G<0)G=0;   \
 531             if (B>=(256<<22))   B=(256<<22)-1; \
 532             else if (B<0)B=0;   \
 533         }
 534
 535 #define YSCALE_YUV_2_GRAY16_C \
 536     for (i=0; i<(dstW>>1); i++) {\
 537         int j;\
 538         int Y1 = 1<<18;\
 539         int Y2 = 1<<18;\
 540         int U  = 1<<18;\
 541         int V  = 1<<18;\
 542         \
 543         const int i2= 2*i;\
 544         \
 545         for (j=0; j<lumFilterSize; j++) {\
 546             Y1 += lumSrc[j][i2] * lumFilter[j];\
 547             Y2 += lumSrc[j][i2+1] * lumFilter[j];\
 548         }\
 549         Y1>>=11;\
 550         Y2>>=11;\
 551         if ((Y1|Y2|U|V)&65536) {\
 552             if (Y1>65535)   Y1=65535; \
 553             else if (Y1<0)Y1=0;   \
 554             if (Y2>65535)   Y2=65535; \
 555             else if (Y2<0)Y2=0;   \
 556         }
 557
 558 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
 559     YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
 560     r = (type *)c->table_rV[V];   \
 561     g = (type *)(c->table_gU[U] + c->table_gV[V]); \
 562     b = (type *)c->table_bU[U];
 563
 564 #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
 565     for (i=0; i<(dstW>>1); i++) { \
 566         const int i2= 2*i;       \
 567         int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
 568         int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
 569         int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
 570         int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
 571         type av_unused *r, *b, *g;                                    \
 572         int av_unused A1, A2;                                         \
 573         if (alpha) {\
 574             A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
 575             A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
 576         }
 577
 578 #define YSCALE_YUV_2_GRAY16_2_C   \
 579     for (i=0; i<(dstW>>1); i++) { \
 580         const int i2= 2*i;       \
 581         int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
 582         int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
 583
 584 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
 585     YSCALE_YUV_2_PACKED2_C(type,alpha)\
 586     r = (type *)c->table_rV[V];\
 587     g = (type *)(c->table_gU[U] + c->table_gV[V]);\
 588     b = (type *)c->table_bU[U];
 589
 590 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
 591     for (i=0; i<(dstW>>1); i++) {\
 592         const int i2= 2*i;\
 593         int Y1= buf0[i2  ]>>7;\
 594         int Y2= buf0[i2+1]>>7;\
 595         int U= (uvbuf1[i     ])>>7;\
 596         int V= (uvbuf1[i+VOFW])>>7;\
 597         type av_unused *r, *b, *g;\
 598         int av_unused A1, A2;\
 599         if (alpha) {\
 600             A1= abuf0[i2  ]>>7;\
 601             A2= abuf0[i2+1]>>7;\
 602         }
 603
 604 #define YSCALE_YUV_2_GRAY16_1_C \
 605     for (i=0; i<(dstW>>1); i++) {\
 606         const int i2= 2*i;\
 607         int Y1= buf0[i2  ]<<1;\
 608         int Y2= buf0[i2+1]<<1;
 609
 610 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
 611     YSCALE_YUV_2_PACKED1_C(type,alpha)\
 612     r = (type *)c->table_rV[V];\
 613     g = (type *)(c->table_gU[U] + c->table_gV[V]);\
 614     b = (type *)c->table_bU[U];
 615
 616 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
 617     for (i=0; i<(dstW>>1); i++) {\
 618         const int i2= 2*i;\
 619         int Y1= buf0[i2  ]>>7;\
 620         int Y2= buf0[i2+1]>>7;\
 621         int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
 622         int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
 623         type av_unused *r, *b, *g;\
 624         int av_unused A1, A2;\
 625         if (alpha) {\
 626             A1= abuf0[i2  ]>>7;\
 627             A2= abuf0[i2+1]>>7;\
 628         }
 629
 630 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
 631     YSCALE_YUV_2_PACKED1B_C(type,alpha)\
 632     r = (type *)c->table_rV[V];\
 633     g = (type *)(c->table_gU[U] + c->table_gV[V]);\
 634     b = (type *)c->table_bU[U];
 635
 636 #define YSCALE_YUV_2_MONO2_C \
 637     const uint8_t * const d128=dither_8x8_220[y&7];\
 638     uint8_t *g= c->table_gU[128] + c->table_gV[128];\
 639     for (i=0; i<dstW-7; i+=8) {\
 640         int acc;\
 641         acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
 642         acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
 643         acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
 644         acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
 645         acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
 646         acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
 647         acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
 648         acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
 649         ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
 650         dest++;\
 651     }
 652
 653 #define YSCALE_YUV_2_MONOX_C \
 654     const uint8_t * const d128=dither_8x8_220[y&7];\
 655     uint8_t *g= c->table_gU[128] + c->table_gV[128];\
 656     int acc=0;\
 657     for (i=0; i<dstW-1; i+=2) {\
 658         int j;\
 659         int Y1=1<<18;\
 660         int Y2=1<<18;\
 661 \
 662         for (j=0; j<lumFilterSize; j++) {\
 663             Y1 += lumSrc[j][i] * lumFilter[j];\
 664             Y2 += lumSrc[j][i+1] * lumFilter[j];\
 665         }\
 666         Y1>>=19;\
 667         Y2>>=19;\
 668         if ((Y1|Y2)&256) {\
 669             if (Y1>255)   Y1=255;\
 670             else if (Y1<0)Y1=0;\
 671             if (Y2>255)   Y2=255;\
 672             else if (Y2<0)Y2=0;\
 673         }\
 674         acc+= acc + g[Y1+d128[(i+0)&7]];\
 675         acc+= acc + g[Y2+d128[(i+1)&7]];\
 676         if ((i&7)==6) {\
 677             ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
 678             dest++;\
 679         }\
 680     }
 681
 682 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
 683     switch(c->dstFormat) {\
 684     case PIX_FMT_RGB48BE:\
 685     case PIX_FMT_RGB48LE:\
 686         func(uint8_t,0)\
 687             ((uint8_t*)dest)[ 0]= r[Y1];\
 688             ((uint8_t*)dest)[ 1]= r[Y1];\
 689             ((uint8_t*)dest)[ 2]= g[Y1];\
 690             ((uint8_t*)dest)[ 3]= g[Y1];\
 691             ((uint8_t*)dest)[ 4]= b[Y1];\
 692             ((uint8_t*)dest)[ 5]= b[Y1];\
 693             ((uint8_t*)dest)[ 6]= r[Y2];\
 694             ((uint8_t*)dest)[ 7]= r[Y2];\
 695             ((uint8_t*)dest)[ 8]= g[Y2];\
 696             ((uint8_t*)dest)[ 9]= g[Y2];\
 697             ((uint8_t*)dest)[10]= b[Y2];\
 698             ((uint8_t*)dest)[11]= b[Y2];\
 699             dest+=12;\
 700         }\
 701         break;\
 702     case PIX_FMT_RGBA:\
 703     case PIX_FMT_BGRA:\
 704         if (CONFIG_SMALL) {\
 705             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
 706             func(uint32_t,needAlpha)\
 707                 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
 708                 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
 709             }\
 710         } else {\
 711             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
 712                 func(uint32_t,1)\
 713                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
 714                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
 715                 }\
 716             } else {\
 717                 func(uint32_t,0)\
 718                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
 719                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
 720                 }\
 721             }\
 722         }\
 723         break;\
 724     case PIX_FMT_ARGB:\
 725     case PIX_FMT_ABGR:\
 726         if (CONFIG_SMALL) {\
 727             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
 728             func(uint32_t,needAlpha)\
 729                 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
 730                 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
 731             }\
 732         } else {\
 733             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
 734                 func(uint32_t,1)\
 735                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
 736                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
 737                 }\
 738             } else {\
 739                 func(uint32_t,0)\
 740                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
 741                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
 742                 }\
 743             }\
 744         }                \
 745         break;\
 746     case PIX_FMT_RGB24:\
 747         func(uint8_t,0)\
 748             ((uint8_t*)dest)[0]= r[Y1];\
 749             ((uint8_t*)dest)[1]= g[Y1];\
 750             ((uint8_t*)dest)[2]= b[Y1];\
 751             ((uint8_t*)dest)[3]= r[Y2];\
 752             ((uint8_t*)dest)[4]= g[Y2];\
 753             ((uint8_t*)dest)[5]= b[Y2];\
 754             dest+=6;\
 755         }\
 756         break;\
 757     case PIX_FMT_BGR24:\
 758         func(uint8_t,0)\
 759             ((uint8_t*)dest)[0]= b[Y1];\
 760             ((uint8_t*)dest)[1]= g[Y1];\
 761             ((uint8_t*)dest)[2]= r[Y1];\
 762             ((uint8_t*)dest)[3]= b[Y2];\
 763             ((uint8_t*)dest)[4]= g[Y2];\
 764             ((uint8_t*)dest)[5]= r[Y2];\
 765             dest+=6;\
 766         }\
 767         break;\
 768     case PIX_FMT_RGB565BE:\
 769     case PIX_FMT_RGB565LE:\
 770     case PIX_FMT_BGR565BE:\
 771     case PIX_FMT_BGR565LE:\
 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             func(uint16_t,0)\
 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_RGB555BE:\
 786     case PIX_FMT_RGB555LE:\
 787     case PIX_FMT_BGR555BE:\
 788     case PIX_FMT_BGR555LE:\
 789         {\
 790             const int dr1= dither_2x2_8[y&1    ][0];\
 791             const int dg1= dither_2x2_8[y&1    ][1];\
 792             const int db1= dither_2x2_8[(y&1)^1][0];\
 793             const int dr2= dither_2x2_8[y&1    ][1];\
 794             const int dg2= dither_2x2_8[y&1    ][0];\
 795             const int db2= dither_2x2_8[(y&1)^1][1];\
 796             func(uint16_t,0)\
 797                 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
 798                 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
 799             }\
 800         }\
 801         break;\
 802     case PIX_FMT_RGB8:\
 803     case PIX_FMT_BGR8:\
 804         {\
 805             const uint8_t * const d64= dither_8x8_73[y&7];\
 806             const uint8_t * const d32= dither_8x8_32[y&7];\
 807             func(uint8_t,0)\
 808                 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
 809                 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
 810             }\
 811         }\
 812         break;\
 813     case PIX_FMT_RGB4:\
 814     case PIX_FMT_BGR4:\
 815         {\
 816             const uint8_t * const d64= dither_8x8_73 [y&7];\
 817             const uint8_t * const d128=dither_8x8_220[y&7];\
 818             func(uint8_t,0)\
 819                 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
 820                                  + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
 821             }\
 822         }\
 823         break;\
 824     case PIX_FMT_RGB4_BYTE:\
 825     case PIX_FMT_BGR4_BYTE:\
 826         {\
 827             const uint8_t * const d64= dither_8x8_73 [y&7];\
 828             const uint8_t * const d128=dither_8x8_220[y&7];\
 829             func(uint8_t,0)\
 830                 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
 831                 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
 832             }\
 833         }\
 834         break;\
 835     case PIX_FMT_MONOBLACK:\
 836     case PIX_FMT_MONOWHITE:\
 837         {\
 838             func_monoblack\
 839         }\
 840         break;\
 841     case PIX_FMT_YUYV422:\
 842         func2\
 843             ((uint8_t*)dest)[2*i2+0]= Y1;\
 844             ((uint8_t*)dest)[2*i2+1]= U;\
 845             ((uint8_t*)dest)[2*i2+2]= Y2;\
 846             ((uint8_t*)dest)[2*i2+3]= V;\
 847         }                \
 848         break;\
 849     case PIX_FMT_UYVY422:\
 850         func2\
 851             ((uint8_t*)dest)[2*i2+0]= U;\
 852             ((uint8_t*)dest)[2*i2+1]= Y1;\
 853             ((uint8_t*)dest)[2*i2+2]= V;\
 854             ((uint8_t*)dest)[2*i2+3]= Y2;\
 855         }                \
 856         break;\
 857     case PIX_FMT_GRAY16BE:\
 858         func_g16\
 859             ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
 860             ((uint8_t*)dest)[2*i2+1]= Y1;\
 861             ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
 862             ((uint8_t*)dest)[2*i2+3]= Y2;\
 863         }                \
 864         break;\
 865     case PIX_FMT_GRAY16LE:\
 866         func_g16\
 867             ((uint8_t*)dest)[2*i2+0]= Y1;\
 868             ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
 869             ((uint8_t*)dest)[2*i2+2]= Y2;\
 870             ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
 871         }                \
 872         break;\
 873     }
 874
 875 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 876                                   const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 877                                   const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
 878 {
 879     int i;
 880     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)
 881 }
 882
 883 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 884                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 885                                     const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
 886 {
 887     int i;
 888     int step= c->dstFormatBpp/8;
 889     int aidx= 3;
 890
 891     switch(c->dstFormat) {
 892     case PIX_FMT_ARGB:
 893         dest++;
 894         aidx= 0;
 895     case PIX_FMT_RGB24:
 896         aidx--;
 897     case PIX_FMT_RGBA:
 898         if (CONFIG_SMALL) {
 899             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
 900             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
 901                 dest[aidx]= needAlpha ? A : 255;
 902                 dest[0]= R>>22;
 903                 dest[1]= G>>22;
 904                 dest[2]= B>>22;
 905                 dest+= step;
 906             }
 907         } else {
 908             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 909                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
 910                     dest[aidx]= A;
 911                     dest[0]= R>>22;
 912                     dest[1]= G>>22;
 913                     dest[2]= B>>22;
 914                     dest+= step;
 915                 }
 916             } else {
 917                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
 918                     dest[aidx]= 255;
 919                     dest[0]= R>>22;
 920                     dest[1]= G>>22;
 921                     dest[2]= B>>22;
 922                     dest+= step;
 923                 }
 924             }
 925         }
 926         break;
 927     case PIX_FMT_ABGR:
 928         dest++;
 929         aidx= 0;
 930     case PIX_FMT_BGR24:
 931         aidx--;
 932     case PIX_FMT_BGRA:
 933         if (CONFIG_SMALL) {
 934             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
 935             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
 936                 dest[aidx]= needAlpha ? A : 255;
 937                 dest[0]= B>>22;
 938                 dest[1]= G>>22;
 939                 dest[2]= R>>22;
 940                 dest+= step;
 941             }
 942         } else {
 943             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 944                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
 945                     dest[aidx]= A;
 946                     dest[0]= B>>22;
 947                     dest[1]= G>>22;
 948                     dest[2]= R>>22;
 949                     dest+= step;
 950                 }
 951             } else {
 952                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
 953                     dest[aidx]= 255;
 954                     dest[0]= B>>22;
 955                     dest[1]= G>>22;
 956                     dest[2]= R>>22;
 957                     dest+= step;
 958                 }
 959             }
 960         }
 961         break;
 962     default:
 963         assert(0);
 964     }
 965 }
 966
 967 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
 968 {
 969     int i;
 970     uint8_t *ptr = plane + stride*y;
 971     for (i=0; i<height; i++) {
 972         memset(ptr, val, width);
 973         ptr += stride;
 974     }
 975 }
 976
 977 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width,
 978                             uint32_t *unused)
 979 {
 980     int i;
 981     for (i = 0; i < width; i++) {
 982         int r = src[i*6+0];
 983         int g = src[i*6+2];
 984         int b = src[i*6+4];
 985
 986         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
 987     }
 988 }
 989
 990 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
 991                              const uint8_t *src1, const uint8_t *src2,
 992                              int width, uint32_t *unused)
 993 {
 994     int i;
 995     assert(src1==src2);
 996     for (i = 0; i < width; i++) {
 997         int r = src1[6*i + 0];
 998         int g = src1[6*i + 2];
 999         int b = src1[6*i + 4];
1000
1001         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1002         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1003     }
1004 }
1005
1006 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1007                                   const uint8_t *src1, const uint8_t *src2,
1008                                   int width, uint32_t *unused)
1009 {
1010     int i;
1011     assert(src1==src2);
1012     for (i = 0; i < width; i++) {
1013         int r= src1[12*i + 0] + src1[12*i + 6];
1014         int g= src1[12*i + 2] + src1[12*i + 8];
1015         int b= src1[12*i + 4] + src1[12*i + 10];
1016
1017         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1018         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1019     }
1020 }
1021
1022 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1023 static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1024 {\
1025     int i;\
1026     for (i=0; i<width; i++) {\
1027         int b= (((const type*)src)[i]>>shb)&maskb;\
1028         int g= (((const type*)src)[i]>>shg)&maskg;\
1029         int r= (((const type*)src)[i]>>shr)&maskr;\
1030 \
1031         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1032     }\
1033 }
1034
1035 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1036 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1037 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1038 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1039 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1040 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1041
1042 static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1043 {
1044     int i;
1045     for (i=0; i<width; i++) {
1046         dst[i]= src[4*i];
1047     }
1048 }
1049
1050 #define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1051 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1052 {\
1053     int i;\
1054     for (i=0; i<width; i++) {\
1055         int b= (((const type*)src)[i]&maskb)>>shb;\
1056         int g= (((const type*)src)[i]&maskg)>>shg;\
1057         int r= (((const type*)src)[i]&maskr)>>shr;\
1058 \
1059         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1060         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1061     }\
1062 }\
1063 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1064 {\
1065     int i;\
1066     for (i=0; i<width; i++) {\
1067         int pix0= ((const type*)src)[2*i+0];\
1068         int pix1= ((const type*)src)[2*i+1];\
1069         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1070         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1071         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1072         g&= maskg|(2*maskg);\
1073 \
1074         g>>=shg;\
1075 \
1076         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1077         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1078     }\
1079 }
1080
1081 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1082 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1083 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1084 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1085 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1086 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1087
1088 static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1089 {
1090     int i;
1091     for (i=0; i<width; i++) {
1092         int d= src[i];
1093
1094         dst[i]= pal[d] & 0xFF;
1095     }
1096 }
1097
1098 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1099                            const uint8_t *src1, const uint8_t *src2,
1100                            long width, uint32_t *pal)
1101 {
1102     int i;
1103     assert(src1 == src2);
1104     for (i=0; i<width; i++) {
1105         int p= pal[src1[i]];
1106
1107         dstU[i]= p>>8;
1108         dstV[i]= p>>16;
1109     }
1110 }
1111
1112 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1113 {
1114     int i, j;
1115     for (i=0; i<width/8; i++) {
1116         int d= ~src[i];
1117         for(j=0; j<8; j++)
1118             dst[8*i+j]= ((d>>(7-j))&1)*255;
1119     }
1120 }
1121
1122 static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1123 {
1124     int i, j;
1125     for (i=0; i<width/8; i++) {
1126         int d= src[i];
1127         for(j=0; j<8; j++)
1128             dst[8*i+j]= ((d>>(7-j))&1)*255;
1129     }
1130 }
1131
1132 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1133 //Plain C versions
1134 #if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1135 #define COMPILE_C
1136 #endif
1137
1138 #if ARCH_PPC
1139 #if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT
1140 #define COMPILE_ALTIVEC
1141 #endif
1142 #endif //ARCH_PPC
1143
1144 #if ARCH_X86
1145
1146 #if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1147 #define COMPILE_MMX
1148 #endif
1149
1150 #if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1151 #define COMPILE_MMX2
1152 #endif
1153
1154 #if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1155 #define COMPILE_3DNOW
1156 #endif
1157 #endif //ARCH_X86
1158
1159 #define COMPILE_TEMPLATE_MMX 0
1160 #define COMPILE_TEMPLATE_MMX2 0
1161 #define COMPILE_TEMPLATE_AMD3DNOW 0
1162 #define COMPILE_TEMPLATE_ALTIVEC 0
1163
1164 #ifdef COMPILE_C
1165 #define RENAME(a) a ## _C
1166 #include "swscale_template.c"
1167 #endif
1168
1169 #ifdef COMPILE_ALTIVEC
1170 #undef RENAME
1171 #undef COMPILE_TEMPLATE_ALTIVEC
1172 #define COMPILE_TEMPLATE_ALTIVEC 1
1173 #define RENAME(a) a ## _altivec
1174 #include "swscale_template.c"
1175 #endif
1176
1177 #if ARCH_X86
1178
1179 //MMX versions
1180 #ifdef COMPILE_MMX
1181 #undef RENAME
1182 #undef COMPILE_TEMPLATE_MMX
1183 #undef COMPILE_TEMPLATE_MMX2
1184 #undef COMPILE_TEMPLATE_AMD3DNOW
1185 #define COMPILE_TEMPLATE_MMX 1
1186 #define COMPILE_TEMPLATE_MMX2 0
1187 #define COMPILE_TEMPLATE_AMD3DNOW 0
1188 #define RENAME(a) a ## _MMX
1189 #include "swscale_template.c"
1190 #endif
1191
1192 //MMX2 versions
1193 #ifdef COMPILE_MMX2
1194 #undef RENAME
1195 #undef COMPILE_TEMPLATE_MMX
1196 #undef COMPILE_TEMPLATE_MMX2
1197 #undef COMPILE_TEMPLATE_AMD3DNOW
1198 #define COMPILE_TEMPLATE_MMX 1
1199 #define COMPILE_TEMPLATE_MMX2 1
1200 #define COMPILE_TEMPLATE_AMD3DNOW 0
1201 #define RENAME(a) a ## _MMX2
1202 #include "swscale_template.c"
1203 #endif
1204
1205 //3DNOW versions
1206 #ifdef COMPILE_3DNOW
1207 #undef RENAME
1208 #undef COMPILE_TEMPLATE_MMX
1209 #undef COMPILE_TEMPLATE_MMX2
1210 #undef COMPILE_TEMPLATE_AMD3DNOW
1211 #define COMPILE_TEMPLATE_MMX 1
1212 #define COMPILE_TEMPLATE_MMX2 0
1213 #define COMPILE_TEMPLATE_AMD3DNOW 1
1214 #define RENAME(a) a ## _3DNow
1215 #include "swscale_template.c"
1216 #endif
1217
1218 #endif //ARCH_X86
1219
1220 SwsFunc ff_getSwsFunc(SwsContext *c)
1221 {
1222 #if CONFIG_RUNTIME_CPUDETECT
1223     int flags = c->flags;
1224
1225 #if ARCH_X86 && CONFIG_GPL
1226     // ordered per speed fastest first
1227     if (flags & SWS_CPU_CAPS_MMX2) {
1228         sws_init_swScale_MMX2(c);
1229         return swScale_MMX2;
1230     } else if (flags & SWS_CPU_CAPS_3DNOW) {
1231         sws_init_swScale_3DNow(c);
1232         return swScale_3DNow;
1233     } else if (flags & SWS_CPU_CAPS_MMX) {
1234         sws_init_swScale_MMX(c);
1235         return swScale_MMX;
1236     } else {
1237         sws_init_swScale_C(c);
1238         return swScale_C;
1239     }
1240
1241 #else
1242 #if ARCH_PPC
1243     if (flags & SWS_CPU_CAPS_ALTIVEC) {
1244         sws_init_swScale_altivec(c);
1245         return swScale_altivec;
1246     } else {
1247         sws_init_swScale_C(c);
1248         return swScale_C;
1249     }
1250 #endif
1251     sws_init_swScale_C(c);
1252     return swScale_C;
1253 #endif /* ARCH_X86 && CONFIG_GPL */
1254 #else //CONFIG_RUNTIME_CPUDETECT
1255 #if   COMPILE_TEMPLATE_MMX2
1256     sws_init_swScale_MMX2(c);
1257     return swScale_MMX2;
1258 #elif COMPILE_TEMPLATE_AMD3DNOW
1259     sws_init_swScale_3DNow(c);
1260     return swScale_3DNow;
1261 #elif COMPILE_TEMPLATE_MMX
1262     sws_init_swScale_MMX(c);
1263     return swScale_MMX;
1264 #elif COMPILE_TEMPLATE_ALTIVEC
1265     sws_init_swScale_altivec(c);
1266     return swScale_altivec;
1267 #else
1268     sws_init_swScale_C(c);
1269     return swScale_C;
1270 #endif
1271 #endif //!CONFIG_RUNTIME_CPUDETECT
1272 }
1273
1274 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1275                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1276 {
1277     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1278     /* Copy Y plane */
1279     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1280         memcpy(dst, src[0], srcSliceH*dstStride[0]);
1281     else {
1282         int i;
1283         const uint8_t *srcPtr= src[0];
1284         uint8_t *dstPtr= dst;
1285         for (i=0; i<srcSliceH; i++) {
1286             memcpy(dstPtr, srcPtr, c->srcW);
1287             srcPtr+= srcStride[0];
1288             dstPtr+= dstStride[0];
1289         }
1290     }
1291     dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1292     if (c->dstFormat == PIX_FMT_NV12)
1293         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1294     else
1295         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1296
1297     return srcSliceH;
1298 }
1299
1300 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1301                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1302 {
1303     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1304
1305     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1306
1307     return srcSliceH;
1308 }
1309
1310 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1311                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1312 {
1313     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1314
1315     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1316
1317     return srcSliceH;
1318 }
1319
1320 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1321                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1322 {
1323     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1324
1325     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1326
1327     return srcSliceH;
1328 }
1329
1330 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1331                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1332 {
1333     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1334
1335     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1336
1337     return srcSliceH;
1338 }
1339
1340 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1341                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1342 {
1343     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1344     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1345     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1346
1347     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1348
1349     if (dstParam[3])
1350         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1351
1352     return srcSliceH;
1353 }
1354
1355 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1356                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1357 {
1358     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1359     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1360     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1361
1362     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1363
1364     return srcSliceH;
1365 }
1366
1367 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1368                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1369 {
1370     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1371     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1372     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1373
1374     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1375
1376     if (dstParam[3])
1377         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1378
1379     return srcSliceH;
1380 }
1381
1382 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1383                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1384 {
1385     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1386     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1387     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1388
1389     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1390
1391     return srcSliceH;
1392 }
1393
1394 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1395                            int srcSliceH, uint8_t* dst[], int dstStride[])
1396 {
1397     const enum PixelFormat srcFormat= c->srcFormat;
1398     const enum PixelFormat dstFormat= c->dstFormat;
1399     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1400                  const uint8_t *palette)=NULL;
1401     int i;
1402     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1403     const uint8_t *srcPtr= src[0];
1404
1405     if (usePal(srcFormat)) {
1406         switch (dstFormat) {
1407         case PIX_FMT_RGB32  : conv = palette8topacked32; break;
1408         case PIX_FMT_BGR32  : conv = palette8topacked32; break;
1409         case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
1410         case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
1411         case PIX_FMT_RGB24  : conv = palette8topacked24; break;
1412         case PIX_FMT_BGR24  : conv = palette8topacked24; break;
1413         }
1414     }
1415
1416     if (!conv)
1417         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1418                sws_format_name(srcFormat), sws_format_name(dstFormat));
1419     else {
1420         for (i=0; i<srcSliceH; i++) {
1421             conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1422             srcPtr+= srcStride[0];
1423             dstPtr+= dstStride[0];
1424         }
1425     }
1426
1427     return srcSliceH;
1428 }
1429
1430 #define isRGBA32(x) (            \
1431            (x) == PIX_FMT_ARGB   \
1432         || (x) == PIX_FMT_RGBA   \
1433         || (x) == PIX_FMT_BGRA   \
1434         || (x) == PIX_FMT_ABGR   \
1435         )
1436
1437 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1438 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1439                            int srcSliceH, uint8_t* dst[], int dstStride[])
1440 {
1441     const enum PixelFormat srcFormat= c->srcFormat;
1442     const enum PixelFormat dstFormat= c->dstFormat;
1443     const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1444     const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1445     const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1446     const int dstId= c->dstFormatBpp >> 2;
1447     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1448
1449 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1450
1451     if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1452         if (     CONV_IS(ABGR, RGBA)
1453               || CONV_IS(ARGB, BGRA)
1454               || CONV_IS(BGRA, ARGB)
1455               || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1456         else if (CONV_IS(ABGR, ARGB)
1457               || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1458         else if (CONV_IS(ABGR, BGRA)
1459               || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1460         else if (CONV_IS(BGRA, RGBA)
1461               || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1462         else if (CONV_IS(BGRA, ABGR)
1463               || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1464     } else
1465     /* BGR -> BGR */
1466     if (  (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1467        || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1468         switch(srcId | (dstId<<4)) {
1469         case 0x34: conv= rgb16to15; break;
1470         case 0x36: conv= rgb24to15; break;
1471         case 0x38: conv= rgb32to15; break;
1472         case 0x43: conv= rgb15to16; break;
1473         case 0x46: conv= rgb24to16; break;
1474         case 0x48: conv= rgb32to16; break;
1475         case 0x63: conv= rgb15to24; break;
1476         case 0x64: conv= rgb16to24; break;
1477         case 0x68: conv= rgb32to24; break;
1478         case 0x83: conv= rgb15to32; break;
1479         case 0x84: conv= rgb16to32; break;
1480         case 0x86: conv= rgb24to32; break;
1481         }
1482     } else if (  (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1483              || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1484         switch(srcId | (dstId<<4)) {
1485         case 0x33: conv= rgb15tobgr15; break;
1486         case 0x34: conv= rgb16tobgr15; break;
1487         case 0x36: conv= rgb24tobgr15; break;
1488         case 0x38: conv= rgb32tobgr15; break;
1489         case 0x43: conv= rgb15tobgr16; break;
1490         case 0x44: conv= rgb16tobgr16; break;
1491         case 0x46: conv= rgb24tobgr16; break;
1492         case 0x48: conv= rgb32tobgr16; break;
1493         case 0x63: conv= rgb15tobgr24; break;
1494         case 0x64: conv= rgb16tobgr24; break;
1495         case 0x66: conv= rgb24tobgr24; break;
1496         case 0x68: conv= rgb32tobgr24; break;
1497         case 0x83: conv= rgb15tobgr32; break;
1498         case 0x84: conv= rgb16tobgr32; break;
1499         case 0x86: conv= rgb24tobgr32; break;
1500         }
1501     }
1502
1503     if (!conv) {
1504         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1505                sws_format_name(srcFormat), sws_format_name(dstFormat));
1506     } else {
1507         const uint8_t *srcPtr= src[0];
1508               uint8_t *dstPtr= dst[0];
1509         if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1510             srcPtr += ALT32_CORR;
1511
1512         if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1513             dstPtr += ALT32_CORR;
1514
1515         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1516             conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1517         else {
1518             int i;
1519             dstPtr += dstStride[0]*srcSliceY;
1520
1521             for (i=0; i<srcSliceH; i++) {
1522                 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1523                 srcPtr+= srcStride[0];
1524                 dstPtr+= dstStride[0];
1525             }
1526         }
1527     }
1528     return srcSliceH;
1529 }
1530
1531 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1532                               int srcSliceH, uint8_t* dst[], int dstStride[])
1533 {
1534     rgb24toyv12(
1535         src[0],
1536         dst[0]+ srcSliceY    *dstStride[0],
1537         dst[1]+(srcSliceY>>1)*dstStride[1],
1538         dst[2]+(srcSliceY>>1)*dstStride[2],
1539         c->srcW, srcSliceH,
1540         dstStride[0], dstStride[1], srcStride[0]);
1541     if (dst[3])
1542         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1543     return srcSliceH;
1544 }
1545
1546 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1547                              int srcSliceH, uint8_t* dst[], int dstStride[])
1548 {
1549     int i;
1550
1551     /* copy Y */
1552     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
1553         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1554     else {
1555         const uint8_t *srcPtr= src[0];
1556         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1557
1558         for (i=0; i<srcSliceH; i++) {
1559             memcpy(dstPtr, srcPtr, c->srcW);
1560             srcPtr+= srcStride[0];
1561             dstPtr+= dstStride[0];
1562         }
1563     }
1564
1565     if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {
1566         planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1567                  srcSliceH >> 2, srcStride[1], dstStride[1]);
1568         planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1569                  srcSliceH >> 2, srcStride[2], dstStride[2]);
1570     } else {
1571         planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1572                  srcSliceH >> 2, srcStride[1], dstStride[2]);
1573         planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1574                  srcSliceH >> 2, srcStride[2], dstStride[1]);
1575     }
1576     if (dst[3])
1577         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1578     return srcSliceH;
1579 }
1580
1581 /* unscaled copy like stuff (assumes nearly identical formats) */
1582 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1583                              int srcSliceH, uint8_t* dst[], int dstStride[])
1584 {
1585     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1586         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1587     else {
1588         int i;
1589         const uint8_t *srcPtr= src[0];
1590         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1591         int length=0;
1592
1593         /* universal length finder */
1594         while(length+c->srcW <= FFABS(dstStride[0])
1595            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1596         assert(length!=0);
1597
1598         for (i=0; i<srcSliceH; i++) {
1599             memcpy(dstPtr, srcPtr, length);
1600             srcPtr+= srcStride[0];
1601             dstPtr+= dstStride[0];
1602         }
1603     }
1604     return srcSliceH;
1605 }
1606
1607 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1608                              int srcSliceH, uint8_t* dst[], int dstStride[])
1609 {
1610     int plane, i, j;
1611     for (plane=0; plane<4; plane++) {
1612         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
1613         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1614         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1615         const uint8_t *srcPtr= src[plane];
1616         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1617
1618         if (!dst[plane]) continue;
1619         // ignore palette for GRAY8
1620         if (plane == 1 && !dst[2]) continue;
1621         if (!src[plane] || (plane == 1 && !src[2])) {
1622             if(is16BPS(c->dstFormat))
1623                 length*=2;
1624             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1625         } else {
1626             if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1627                 if (!isBE(c->srcFormat)) srcPtr++;
1628                 for (i=0; i<height; i++) {
1629                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1630                     srcPtr+= srcStride[plane];
1631                     dstPtr+= dstStride[plane];
1632                 }
1633             } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1634                 for (i=0; i<height; i++) {
1635                     for (j=0; j<length; j++) {
1636                         dstPtr[ j<<1   ] = srcPtr[j];
1637                         dstPtr[(j<<1)+1] = srcPtr[j];
1638                     }
1639                     srcPtr+= srcStride[plane];
1640                     dstPtr+= dstStride[plane];
1641                 }
1642             } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1643                   && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1644
1645                 for (i=0; i<height; i++) {
1646                     for (j=0; j<length; j++)
1647                         ((uint16_t*)dstPtr)[j] = bswap_16(((const uint16_t*)srcPtr)[j]);
1648                     srcPtr+= srcStride[plane];
1649                     dstPtr+= dstStride[plane];
1650                 }
1651             } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
1652                 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
1653             else {
1654                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1655                     length*=2;
1656                 for (i=0; i<height; i++) {
1657                     memcpy(dstPtr, srcPtr, length);
1658                     srcPtr+= srcStride[plane];
1659                     dstPtr+= dstStride[plane];
1660                 }
1661             }
1662         }
1663     }
1664     return srcSliceH;
1665 }
1666
1667 int ff_hardcodedcpuflags(void)
1668 {
1669     int flags = 0;
1670 #if   COMPILE_TEMPLATE_MMX2
1671     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1672 #elif COMPILE_TEMPLATE_AMD3DNOW
1673     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1674 #elif COMPILE_TEMPLATE_MMX
1675     flags |= SWS_CPU_CAPS_MMX;
1676 #elif COMPILE_TEMPLATE_ALTIVEC
1677     flags |= SWS_CPU_CAPS_ALTIVEC;
1678 #elif ARCH_BFIN
1679     flags |= SWS_CPU_CAPS_BFIN;
1680 #endif
1681     return flags;
1682 }
1683
1684 void ff_get_unscaled_swscale(SwsContext *c)
1685 {
1686     const enum PixelFormat srcFormat = c->srcFormat;
1687     const enum PixelFormat dstFormat = c->dstFormat;
1688     const int flags = c->flags;
1689     const int dstH = c->dstH;
1690     int needsDither;
1691
1692     needsDither= isAnyRGB(dstFormat)
1693         &&  c->dstFormatBpp < 24
1694         && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1695
1696     /* yv12_to_nv12 */
1697     if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1698         c->swScale= planarToNv12Wrapper;
1699     }
1700     /* yuv2bgr */
1701     if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1702         && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1703         c->swScale= ff_yuv2rgb_get_func_ptr(c);
1704     }
1705
1706     if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1707         c->swScale= yvu9ToYv12Wrapper;
1708     }
1709
1710     /* bgr24toYV12 */
1711     if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1712         c->swScale= bgr24ToYv12Wrapper;
1713
1714     /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1715     if (   isAnyRGB(srcFormat)
1716         && isAnyRGB(dstFormat)
1717         && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
1718         && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
1719         && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
1720         && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
1721         && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1722         && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1723         && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1724         && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1725         && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
1726         && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
1727         && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1728         c->swScale= rgbToRgbWrapper;
1729
1730     if ((usePal(srcFormat) && (
1731         dstFormat == PIX_FMT_RGB32   ||
1732         dstFormat == PIX_FMT_RGB32_1 ||
1733         dstFormat == PIX_FMT_RGB24   ||
1734         dstFormat == PIX_FMT_BGR32   ||
1735         dstFormat == PIX_FMT_BGR32_1 ||
1736         dstFormat == PIX_FMT_BGR24)))
1737         c->swScale= palToRgbWrapper;
1738
1739     if (srcFormat == PIX_FMT_YUV422P) {
1740         if (dstFormat == PIX_FMT_YUYV422)
1741             c->swScale= yuv422pToYuy2Wrapper;
1742         else if (dstFormat == PIX_FMT_UYVY422)
1743             c->swScale= yuv422pToUyvyWrapper;
1744     }
1745
1746     /* LQ converters if -sws 0 or -sws 4*/
1747     if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1748         /* yv12_to_yuy2 */
1749         if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1750             if (dstFormat == PIX_FMT_YUYV422)
1751                 c->swScale= planarToYuy2Wrapper;
1752             else if (dstFormat == PIX_FMT_UYVY422)
1753                 c->swScale= planarToUyvyWrapper;
1754         }
1755     }
1756     if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1757         c->swScale= yuyvToYuv420Wrapper;
1758     if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1759         c->swScale= uyvyToYuv420Wrapper;
1760     if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1761         c->swScale= yuyvToYuv422Wrapper;
1762     if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1763         c->swScale= uyvyToYuv422Wrapper;
1764
1765 #ifdef COMPILE_ALTIVEC
1766     if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1767         !(c->flags & SWS_BITEXACT) &&
1768         srcFormat == PIX_FMT_YUV420P) {
1769         // unscaled YV12 -> packed YUV, we want speed
1770         if (dstFormat == PIX_FMT_YUYV422)
1771             c->swScale= yv12toyuy2_unscaled_altivec;
1772         else if (dstFormat == PIX_FMT_UYVY422)
1773             c->swScale= yv12touyvy_unscaled_altivec;
1774     }
1775 #endif
1776
1777     /* simple copy */
1778     if (  srcFormat == dstFormat
1779         || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1780         || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1781         || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1782         || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1783         || (isGray(dstFormat) && isGray(srcFormat))
1784         || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1785             && c->chrDstHSubSample == c->chrSrcHSubSample
1786             && c->chrDstVSubSample == c->chrSrcVSubSample
1787             && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1788             && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1789     {
1790         if (isPacked(c->srcFormat))
1791             c->swScale= packedCopyWrapper;
1792         else /* Planar YUV or gray */
1793             c->swScale= planarCopyWrapper;
1794     }
1795 #if ARCH_BFIN
1796     if (flags & SWS_CPU_CAPS_BFIN)
1797         ff_bfin_get_unscaled_swscale (c);
1798 #endif
1799 }
1800
1801 static void reset_ptr(const uint8_t* src[], int format)
1802 {
1803     if(!isALPHA(format))
1804         src[3]=NULL;
1805     if(!isPlanarYUV(format)) {
1806         src[3]=src[2]=NULL;
1807
1808         if (!usePal(format))
1809             src[1]= NULL;
1810     }
1811 }
1812
1813 /**
1814  * swscale wrapper, so we don't need to export the SwsContext.
1815  * Assumes planar YUV to be in YUV order instead of YVU.
1816  */
1817 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1818               int srcSliceH, uint8_t* const dst[], const int dstStride[])
1819 {
1820     int i;
1821     const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1822     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1823
1824     // do not mess up sliceDir if we have a "trailing" 0-size slice
1825     if (srcSliceH == 0)
1826         return 0;
1827
1828     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1829         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1830         return 0;
1831     }
1832     if (c->sliceDir == 0) {
1833         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1834     }
1835
1836     if (usePal(c->srcFormat)) {
1837         for (i=0; i<256; i++) {
1838             int p, r, g, b,y,u,v;
1839             if(c->srcFormat == PIX_FMT_PAL8) {
1840                 p=((const uint32_t*)(src[1]))[i];
1841                 r= (p>>16)&0xFF;
1842                 g= (p>> 8)&0xFF;
1843                 b=  p     &0xFF;
1844             } else if(c->srcFormat == PIX_FMT_RGB8) {
1845                 r= (i>>5    )*36;
1846                 g= ((i>>2)&7)*36;
1847                 b= (i&3     )*85;
1848             } else if(c->srcFormat == PIX_FMT_BGR8) {
1849                 b= (i>>6    )*85;
1850                 g= ((i>>3)&7)*36;
1851                 r= (i&7     )*36;
1852             } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
1853                 r= (i>>3    )*255;
1854                 g= ((i>>1)&3)*85;
1855                 b= (i&1     )*255;
1856             } else if(c->srcFormat == PIX_FMT_GRAY8) {
1857                 r = g = b = i;
1858             } else {
1859                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
1860                 b= (i>>3    )*255;
1861                 g= ((i>>1)&3)*85;
1862                 r= (i&1     )*255;
1863             }
1864             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1865             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1866             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1867             c->pal_yuv[i]= y + (u<<8) + (v<<16);
1868
1869             switch(c->dstFormat) {
1870             case PIX_FMT_BGR32:
1871 #if !HAVE_BIGENDIAN
1872             case PIX_FMT_RGB24:
1873 #endif
1874                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);
1875                 break;
1876             case PIX_FMT_BGR32_1:
1877 #if HAVE_BIGENDIAN
1878             case PIX_FMT_BGR24:
1879 #endif
1880                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
1881                 break;
1882             case PIX_FMT_RGB32_1:
1883 #if HAVE_BIGENDIAN
1884             case PIX_FMT_RGB24:
1885 #endif
1886                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
1887                 break;
1888             case PIX_FMT_RGB32:
1889 #if !HAVE_BIGENDIAN
1890             case PIX_FMT_BGR24:
1891 #endif
1892             default:
1893                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);
1894             }
1895         }
1896     }
1897
1898     // copy strides, so they can safely be modified
1899     if (c->sliceDir == 1) {
1900         // slices go from top to bottom
1901         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
1902         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
1903
1904         reset_ptr(src2, c->srcFormat);
1905         reset_ptr((const uint8_t**)dst2, c->dstFormat);
1906
1907         /* reset slice direction at end of frame */
1908         if (srcSliceY + srcSliceH == c->srcH)
1909             c->sliceDir = 0;
1910
1911         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
1912     } else {
1913         // slices go from bottom to top => we flip the image internally
1914         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
1915         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
1916
1917         src2[0] += (srcSliceH-1)*srcStride[0];
1918         if (!usePal(c->srcFormat))
1919             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
1920         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
1921         src2[3] += (srcSliceH-1)*srcStride[3];
1922         dst2[0] += ( c->dstH                      -1)*dstStride[0];
1923         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
1924         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
1925         dst2[3] += ( c->dstH                      -1)*dstStride[3];
1926
1927         reset_ptr(src2, c->srcFormat);
1928         reset_ptr((const uint8_t**)dst2, c->dstFormat);
1929
1930         /* reset slice direction at end of frame */
1931         if (!srcSliceY)
1932             c->sliceDir = 0;
1933
1934         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
1935     }
1936 }
1937
1938 #if LIBSWSCALE_VERSION_MAJOR < 1
1939 int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
1940                       int srcSliceH, uint8_t* dst[], int dstStride[])
1941 {
1942     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
1943 }
1944 #endif