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}{1,4,8,12,15,16,24,32}
  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
  96 #define RGB2YUV_SHIFT 15
  97 #define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
  98 #define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
  99 #define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 100 #define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
 101 #define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 102 #define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 103 #define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
 104 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 105 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 106
 107 static const double rgb2yuv_table[8][9]={
 108     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
 109     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
 110     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
 111     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
 112     {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
 113     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
 114     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
 115     {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
 116 };
 117
 118 /*
 119 NOTES
 120 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
 121
 122 TODO
 123 more intelligent misalignment avoidance for the horizontal scaler
 124 write special vertical cubic upscale version
 125 optimize C code (YV12 / minmax)
 126 add support for packed pixel YUV input & output
 127 add support for Y8 output
 128 optimize BGR24 & BGR32
 129 add BGR4 output support
 130 write special BGR->BGR scaler
 131 */
 132
 133 #if ARCH_X86 && CONFIG_GPL
 134 DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
 135 DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
 136 DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
 137 DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
 138 DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
 139 DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
 140 DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
 141 DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
 142
 143 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
 144         0x0103010301030103LL,
 145         0x0200020002000200LL,};
 146
 147 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
 148         0x0602060206020602LL,
 149         0x0004000400040004LL,};
 150
 151 DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
 152 DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
 153 DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
 154 DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
 155 DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
 156 DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
 157
 158 DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
 159 DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
 160 DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
 161
 162 #ifdef FAST_BGR2YV12
 163 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
 164 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
 165 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
 166 #else
 167 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
 168 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
 169 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
 170 #endif /* FAST_BGR2YV12 */
 171 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
 172 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
 173 DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
 174
 175 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
 176 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
 177 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
 178 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
 179 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
 180
 181 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
 182     {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
 183     {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
 184 };
 185
 186 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
 187
 188 #endif /* ARCH_X86 && CONFIG_GPL */
 189
 190 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
 191 {  1,   3,   1,   3,   1,   3,   1,   3, },
 192 {  2,   0,   2,   0,   2,   0,   2,   0, },
 193 };
 194
 195 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
 196 {  6,   2,   6,   2,   6,   2,   6,   2, },
 197 {  0,   4,   0,   4,   0,   4,   0,   4, },
 198 };
 199
 200 DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={
 201 {  8,   4,  11,   7,   8,   4,  11,   7, },
 202 {  2,  14,   1,  13,   2,  14,   1,  13, },
 203 { 10,   6,   9,   5,  10,   6,   9,   5, },
 204 {  0,  12,   3,  15,   0,  12,   3,  15, },
 205 };
 206
 207 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
 208 { 17,   9,  23,  15,  16,   8,  22,  14, },
 209 {  5,  29,   3,  27,   4,  28,   2,  26, },
 210 { 21,  13,  19,  11,  20,  12,  18,  10, },
 211 {  0,  24,   6,  30,   1,  25,   7,  31, },
 212 { 16,   8,  22,  14,  17,   9,  23,  15, },
 213 {  4,  28,   2,  26,   5,  29,   3,  27, },
 214 { 20,  12,  18,  10,  21,  13,  19,  11, },
 215 {  1,  25,   7,  31,   0,  24,   6,  30, },
 216 };
 217
 218 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
 219 {  0,  55,  14,  68,   3,  58,  17,  72, },
 220 { 37,  18,  50,  32,  40,  22,  54,  35, },
 221 {  9,  64,   5,  59,  13,  67,   8,  63, },
 222 { 46,  27,  41,  23,  49,  31,  44,  26, },
 223 {  2,  57,  16,  71,   1,  56,  15,  70, },
 224 { 39,  21,  52,  34,  38,  19,  51,  33, },
 225 { 11,  66,   7,  62,  10,  65,   6,  60, },
 226 { 48,  30,  43,  25,  47,  29,  42,  24, },
 227 };
 228
 229 #if 1
 230 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 231 {117,  62, 158, 103, 113,  58, 155, 100, },
 232 { 34, 199,  21, 186,  31, 196,  17, 182, },
 233 {144,  89, 131,  76, 141,  86, 127,  72, },
 234 {  0, 165,  41, 206,  10, 175,  52, 217, },
 235 {110,  55, 151,  96, 120,  65, 162, 107, },
 236 { 28, 193,  14, 179,  38, 203,  24, 189, },
 237 {138,  83, 124,  69, 148,  93, 134,  79, },
 238 {  7, 172,  48, 213,   3, 168,  45, 210, },
 239 };
 240 #elif 1
 241 // tries to correct a gamma of 1.5
 242 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 243 {  0, 143,  18, 200,   2, 156,  25, 215, },
 244 { 78,  28, 125,  64,  89,  36, 138,  74, },
 245 { 10, 180,   3, 161,  16, 195,   8, 175, },
 246 {109,  51,  93,  38, 121,  60, 105,  47, },
 247 {  1, 152,  23, 210,   0, 147,  20, 205, },
 248 { 85,  33, 134,  71,  81,  30, 130,  67, },
 249 { 14, 190,   6, 171,  12, 185,   5, 166, },
 250 {117,  57, 101,  44, 113,  54,  97,  41, },
 251 };
 252 #elif 1
 253 // tries to correct a gamma of 2.0
 254 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 255 {  0, 124,   8, 193,   0, 140,  12, 213, },
 256 { 55,  14, 104,  42,  66,  19, 119,  52, },
 257 {  3, 168,   1, 145,   6, 187,   3, 162, },
 258 { 86,  31,  70,  21,  99,  39,  82,  28, },
 259 {  0, 134,  11, 206,   0, 129,   9, 200, },
 260 { 62,  17, 114,  48,  58,  16, 109,  45, },
 261 {  5, 181,   2, 157,   4, 175,   1, 151, },
 262 { 95,  36,  78,  26,  90,  34,  74,  24, },
 263 };
 264 #else
 265 // tries to correct a gamma of 2.5
 266 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 267 {  0, 107,   3, 187,   0, 125,   6, 212, },
 268 { 39,   7,  86,  28,  49,  11, 102,  36, },
 269 {  1, 158,   0, 131,   3, 180,   1, 151, },
 270 { 68,  19,  52,  12,  81,  25,  64,  17, },
 271 {  0, 119,   5, 203,   0, 113,   4, 195, },
 272 { 45,   9,  96,  33,  42,   8,  91,  30, },
 273 {  2, 172,   1, 144,   2, 165,   0, 137, },
 274 { 77,  23,  60,  15,  72,  21,  56,  14, },
 275 };
 276 #endif
 277
 278 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 279                                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 280                                                     const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
 281                                                     int dstW, int chrDstW, int big_endian)
 282 {
 283     //FIXME Optimize (just quickly written not optimized..)
 284     int i;
 285
 286     for (i = 0; i < dstW; i++) {
 287         int val = 1 << 10;
 288         int j;
 289
 290         for (j = 0; j < lumFilterSize; j++)
 291             val += lumSrc[j][i] * lumFilter[j];
 292
 293         if (big_endian) {
 294             AV_WB16(&dest[i], av_clip_uint16(val >> 11));
 295         } else {
 296             AV_WL16(&dest[i], av_clip_uint16(val >> 11));
 297         }
 298     }
 299
 300     if (uDest) {
 301         for (i = 0; i < chrDstW; i++) {
 302             int u = 1 << 10;
 303             int v = 1 << 10;
 304             int j;
 305
 306             for (j = 0; j < chrFilterSize; j++) {
 307                 u += chrSrc[j][i       ] * chrFilter[j];
 308                 v += chrSrc[j][i + VOFW] * chrFilter[j];
 309             }
 310
 311             if (big_endian) {
 312                 AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
 313                 AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
 314             } else {
 315                 AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
 316                 AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
 317             }
 318         }
 319     }
 320
 321     if (CONFIG_SWSCALE_ALPHA && aDest) {
 322         for (i = 0; i < dstW; i++) {
 323             int val = 1 << 10;
 324             int j;
 325
 326             for (j = 0; j < lumFilterSize; j++)
 327                 val += alpSrc[j][i] * lumFilter[j];
 328
 329             if (big_endian) {
 330                 AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
 331             } else {
 332                 AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
 333             }
 334         }
 335     }
 336 }
 337
 338 static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 339                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 340                                  const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
 341                                  enum PixelFormat dstFormat)
 342 {
 343     if (isBE(dstFormat)) {
 344         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
 345                                chrFilter, chrSrc, chrFilterSize,
 346                                alpSrc,
 347                                dest, uDest, vDest, aDest,
 348                                dstW, chrDstW, 1);
 349     } else {
 350         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
 351                                chrFilter, chrSrc, chrFilterSize,
 352                                alpSrc,
 353                                dest, uDest, vDest, aDest,
 354                                dstW, chrDstW, 0);
 355     }
 356 }
 357
 358 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 359                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 360                                const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
 361 {
 362     //FIXME Optimize (just quickly written not optimized..)
 363     int i;
 364     for (i=0; i<dstW; i++) {
 365         int val=1<<18;
 366         int j;
 367         for (j=0; j<lumFilterSize; j++)
 368             val += lumSrc[j][i] * lumFilter[j];
 369
 370         dest[i]= av_clip_uint8(val>>19);
 371     }
 372
 373     if (uDest)
 374         for (i=0; i<chrDstW; i++) {
 375             int u=1<<18;
 376             int v=1<<18;
 377             int j;
 378             for (j=0; j<chrFilterSize; j++) {
 379                 u += chrSrc[j][i] * chrFilter[j];
 380                 v += chrSrc[j][i + VOFW] * chrFilter[j];
 381             }
 382
 383             uDest[i]= av_clip_uint8(u>>19);
 384             vDest[i]= av_clip_uint8(v>>19);
 385         }
 386
 387     if (CONFIG_SWSCALE_ALPHA && aDest)
 388         for (i=0; i<dstW; i++) {
 389             int val=1<<18;
 390             int j;
 391             for (j=0; j<lumFilterSize; j++)
 392                 val += alpSrc[j][i] * lumFilter[j];
 393
 394             aDest[i]= av_clip_uint8(val>>19);
 395         }
 396
 397 }
 398
 399 static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 400                                 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 401                                 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
 402 {
 403     //FIXME Optimize (just quickly written not optimized..)
 404     int i;
 405     for (i=0; i<dstW; i++) {
 406         int val=1<<18;
 407         int j;
 408         for (j=0; j<lumFilterSize; j++)
 409             val += lumSrc[j][i] * lumFilter[j];
 410
 411         dest[i]= av_clip_uint8(val>>19);
 412     }
 413
 414     if (!uDest)
 415         return;
 416
 417     if (dstFormat == PIX_FMT_NV12)
 418         for (i=0; i<chrDstW; i++) {
 419             int u=1<<18;
 420             int v=1<<18;
 421             int j;
 422             for (j=0; j<chrFilterSize; j++) {
 423                 u += chrSrc[j][i] * chrFilter[j];
 424                 v += chrSrc[j][i + VOFW] * chrFilter[j];
 425             }
 426
 427             uDest[2*i]= av_clip_uint8(u>>19);
 428             uDest[2*i+1]= av_clip_uint8(v>>19);
 429         }
 430     else
 431         for (i=0; i<chrDstW; i++) {
 432             int u=1<<18;
 433             int v=1<<18;
 434             int j;
 435             for (j=0; j<chrFilterSize; j++) {
 436                 u += chrSrc[j][i] * chrFilter[j];
 437                 v += chrSrc[j][i + VOFW] * chrFilter[j];
 438             }
 439
 440             uDest[2*i]= av_clip_uint8(v>>19);
 441             uDest[2*i+1]= av_clip_uint8(u>>19);
 442         }
 443 }
 444
 445 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
 446     for (i=0; i<(dstW>>1); i++) {\
 447         int j;\
 448         int Y1 = 1<<18;\
 449         int Y2 = 1<<18;\
 450         int U  = 1<<18;\
 451         int V  = 1<<18;\
 452         int av_unused A1, A2;\
 453         type av_unused *r, *b, *g;\
 454         const int i2= 2*i;\
 455         \
 456         for (j=0; j<lumFilterSize; j++) {\
 457             Y1 += lumSrc[j][i2] * lumFilter[j];\
 458             Y2 += lumSrc[j][i2+1] * lumFilter[j];\
 459         }\
 460         for (j=0; j<chrFilterSize; j++) {\
 461             U += chrSrc[j][i] * chrFilter[j];\
 462             V += chrSrc[j][i+VOFW] * chrFilter[j];\
 463         }\
 464         Y1>>=19;\
 465         Y2>>=19;\
 466         U >>=19;\
 467         V >>=19;\
 468         if (alpha) {\
 469             A1 = 1<<18;\
 470             A2 = 1<<18;\
 471             for (j=0; j<lumFilterSize; j++) {\
 472                 A1 += alpSrc[j][i2  ] * lumFilter[j];\
 473                 A2 += alpSrc[j][i2+1] * lumFilter[j];\
 474             }\
 475             A1>>=19;\
 476             A2>>=19;\
 477         }
 478
 479 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
 480         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
 481         if ((Y1|Y2|U|V)&256) {\
 482             if (Y1>255)   Y1=255; \
 483             else if (Y1<0)Y1=0;   \
 484             if (Y2>255)   Y2=255; \
 485             else if (Y2<0)Y2=0;   \
 486             if (U>255)    U=255;  \
 487             else if (U<0) U=0;    \
 488             if (V>255)    V=255;  \
 489             else if (V<0) V=0;    \
 490         }\
 491         if (alpha && ((A1|A2)&256)) {\
 492             A1=av_clip_uint8(A1);\
 493             A2=av_clip_uint8(A2);\
 494         }
 495
 496 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
 497     for (i=0; i<dstW; i++) {\
 498         int j;\
 499         int Y = 0;\
 500         int U = -128<<19;\
 501         int V = -128<<19;\
 502         int av_unused A;\
 503         int R,G,B;\
 504         \
 505         for (j=0; j<lumFilterSize; j++) {\
 506             Y += lumSrc[j][i     ] * lumFilter[j];\
 507         }\
 508         for (j=0; j<chrFilterSize; j++) {\
 509             U += chrSrc[j][i     ] * chrFilter[j];\
 510             V += chrSrc[j][i+VOFW] * chrFilter[j];\
 511         }\
 512         Y >>=10;\
 513         U >>=10;\
 514         V >>=10;\
 515         if (alpha) {\
 516             A = rnd;\
 517             for (j=0; j<lumFilterSize; j++)\
 518                 A += alpSrc[j][i     ] * lumFilter[j];\
 519             A >>=19;\
 520             if (A&256)\
 521                 A = av_clip_uint8(A);\
 522         }
 523
 524 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
 525     YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
 526         Y-= c->yuv2rgb_y_offset;\
 527         Y*= c->yuv2rgb_y_coeff;\
 528         Y+= rnd;\
 529         R= Y + V*c->yuv2rgb_v2r_coeff;\
 530         G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
 531         B= Y +                          U*c->yuv2rgb_u2b_coeff;\
 532         if ((R|G|B)&(0xC0000000)) {\
 533             if (R>=(256<<22))   R=(256<<22)-1; \
 534             else if (R<0)R=0;   \
 535             if (G>=(256<<22))   G=(256<<22)-1; \
 536             else if (G<0)G=0;   \
 537             if (B>=(256<<22))   B=(256<<22)-1; \
 538             else if (B<0)B=0;   \
 539         }
 540
 541 #define YSCALE_YUV_2_GRAY16_C \
 542     for (i=0; i<(dstW>>1); i++) {\
 543         int j;\
 544         int Y1 = 1<<18;\
 545         int Y2 = 1<<18;\
 546         int U  = 1<<18;\
 547         int V  = 1<<18;\
 548         \
 549         const int i2= 2*i;\
 550         \
 551         for (j=0; j<lumFilterSize; j++) {\
 552             Y1 += lumSrc[j][i2] * lumFilter[j];\
 553             Y2 += lumSrc[j][i2+1] * lumFilter[j];\
 554         }\
 555         Y1>>=11;\
 556         Y2>>=11;\
 557         if ((Y1|Y2|U|V)&65536) {\
 558             if (Y1>65535)   Y1=65535; \
 559             else if (Y1<0)Y1=0;   \
 560             if (Y2>65535)   Y2=65535; \
 561             else if (Y2<0)Y2=0;   \
 562         }
 563
 564 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
 565     YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
 566     r = (type *)c->table_rV[V];   \
 567     g = (type *)(c->table_gU[U] + c->table_gV[V]); \
 568     b = (type *)c->table_bU[U];
 569
 570 #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
 571     for (i=0; i<(dstW>>1); i++) { \
 572         const int i2= 2*i;       \
 573         int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
 574         int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
 575         int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
 576         int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
 577         type av_unused *r, *b, *g;                                    \
 578         int av_unused A1, A2;                                         \
 579         if (alpha) {\
 580             A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
 581             A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
 582         }
 583
 584 #define YSCALE_YUV_2_GRAY16_2_C   \
 585     for (i=0; i<(dstW>>1); i++) { \
 586         const int i2= 2*i;       \
 587         int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
 588         int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
 589
 590 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
 591     YSCALE_YUV_2_PACKED2_C(type,alpha)\
 592     r = (type *)c->table_rV[V];\
 593     g = (type *)(c->table_gU[U] + c->table_gV[V]);\
 594     b = (type *)c->table_bU[U];
 595
 596 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
 597     for (i=0; i<(dstW>>1); i++) {\
 598         const int i2= 2*i;\
 599         int Y1= buf0[i2  ]>>7;\
 600         int Y2= buf0[i2+1]>>7;\
 601         int U= (uvbuf1[i     ])>>7;\
 602         int V= (uvbuf1[i+VOFW])>>7;\
 603         type av_unused *r, *b, *g;\
 604         int av_unused A1, A2;\
 605         if (alpha) {\
 606             A1= abuf0[i2  ]>>7;\
 607             A2= abuf0[i2+1]>>7;\
 608         }
 609
 610 #define YSCALE_YUV_2_GRAY16_1_C \
 611     for (i=0; i<(dstW>>1); i++) {\
 612         const int i2= 2*i;\
 613         int Y1= buf0[i2  ]<<1;\
 614         int Y2= buf0[i2+1]<<1;
 615
 616 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
 617     YSCALE_YUV_2_PACKED1_C(type,alpha)\
 618     r = (type *)c->table_rV[V];\
 619     g = (type *)(c->table_gU[U] + c->table_gV[V]);\
 620     b = (type *)c->table_bU[U];
 621
 622 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
 623     for (i=0; i<(dstW>>1); i++) {\
 624         const int i2= 2*i;\
 625         int Y1= buf0[i2  ]>>7;\
 626         int Y2= buf0[i2+1]>>7;\
 627         int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
 628         int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
 629         type av_unused *r, *b, *g;\
 630         int av_unused A1, A2;\
 631         if (alpha) {\
 632             A1= abuf0[i2  ]>>7;\
 633             A2= abuf0[i2+1]>>7;\
 634         }
 635
 636 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
 637     YSCALE_YUV_2_PACKED1B_C(type,alpha)\
 638     r = (type *)c->table_rV[V];\
 639     g = (type *)(c->table_gU[U] + c->table_gV[V]);\
 640     b = (type *)c->table_bU[U];
 641
 642 #define YSCALE_YUV_2_MONO2_C \
 643     const uint8_t * const d128=dither_8x8_220[y&7];\
 644     uint8_t *g= c->table_gU[128] + c->table_gV[128];\
 645     for (i=0; i<dstW-7; i+=8) {\
 646         int acc;\
 647         acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
 648         acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
 649         acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
 650         acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
 651         acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
 652         acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
 653         acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
 654         acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
 655         ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
 656         dest++;\
 657     }
 658
 659 #define YSCALE_YUV_2_MONOX_C \
 660     const uint8_t * const d128=dither_8x8_220[y&7];\
 661     uint8_t *g= c->table_gU[128] + c->table_gV[128];\
 662     int acc=0;\
 663     for (i=0; i<dstW-1; i+=2) {\
 664         int j;\
 665         int Y1=1<<18;\
 666         int Y2=1<<18;\
 667 \
 668         for (j=0; j<lumFilterSize; j++) {\
 669             Y1 += lumSrc[j][i] * lumFilter[j];\
 670             Y2 += lumSrc[j][i+1] * lumFilter[j];\
 671         }\
 672         Y1>>=19;\
 673         Y2>>=19;\
 674         if ((Y1|Y2)&256) {\
 675             if (Y1>255)   Y1=255;\
 676             else if (Y1<0)Y1=0;\
 677             if (Y2>255)   Y2=255;\
 678             else if (Y2<0)Y2=0;\
 679         }\
 680         acc+= acc + g[Y1+d128[(i+0)&7]];\
 681         acc+= acc + g[Y2+d128[(i+1)&7]];\
 682         if ((i&7)==6) {\
 683             ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
 684             dest++;\
 685         }\
 686     }
 687
 688 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
 689     switch(c->dstFormat) {\
 690     case PIX_FMT_RGB48BE:\
 691     case PIX_FMT_RGB48LE:\
 692         func(uint8_t,0)\
 693             ((uint8_t*)dest)[ 0]= r[Y1];\
 694             ((uint8_t*)dest)[ 1]= r[Y1];\
 695             ((uint8_t*)dest)[ 2]= g[Y1];\
 696             ((uint8_t*)dest)[ 3]= g[Y1];\
 697             ((uint8_t*)dest)[ 4]= b[Y1];\
 698             ((uint8_t*)dest)[ 5]= b[Y1];\
 699             ((uint8_t*)dest)[ 6]= r[Y2];\
 700             ((uint8_t*)dest)[ 7]= r[Y2];\
 701             ((uint8_t*)dest)[ 8]= g[Y2];\
 702             ((uint8_t*)dest)[ 9]= g[Y2];\
 703             ((uint8_t*)dest)[10]= b[Y2];\
 704             ((uint8_t*)dest)[11]= b[Y2];\
 705             dest+=12;\
 706         }\
 707         break;\
 708     case PIX_FMT_RGBA:\
 709     case PIX_FMT_BGRA:\
 710         if (CONFIG_SMALL) {\
 711             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
 712             func(uint32_t,needAlpha)\
 713                 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
 714                 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
 715             }\
 716         } else {\
 717             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
 718                 func(uint32_t,1)\
 719                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
 720                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
 721                 }\
 722             } else {\
 723                 func(uint32_t,0)\
 724                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
 725                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
 726                 }\
 727             }\
 728         }\
 729         break;\
 730     case PIX_FMT_ARGB:\
 731     case PIX_FMT_ABGR:\
 732         if (CONFIG_SMALL) {\
 733             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
 734             func(uint32_t,needAlpha)\
 735                 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
 736                 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
 737             }\
 738         } else {\
 739             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
 740                 func(uint32_t,1)\
 741                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
 742                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
 743                 }\
 744             } else {\
 745                 func(uint32_t,0)\
 746                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
 747                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
 748                 }\
 749             }\
 750         }                \
 751         break;\
 752     case PIX_FMT_RGB24:\
 753         func(uint8_t,0)\
 754             ((uint8_t*)dest)[0]= r[Y1];\
 755             ((uint8_t*)dest)[1]= g[Y1];\
 756             ((uint8_t*)dest)[2]= b[Y1];\
 757             ((uint8_t*)dest)[3]= r[Y2];\
 758             ((uint8_t*)dest)[4]= g[Y2];\
 759             ((uint8_t*)dest)[5]= b[Y2];\
 760             dest+=6;\
 761         }\
 762         break;\
 763     case PIX_FMT_BGR24:\
 764         func(uint8_t,0)\
 765             ((uint8_t*)dest)[0]= b[Y1];\
 766             ((uint8_t*)dest)[1]= g[Y1];\
 767             ((uint8_t*)dest)[2]= r[Y1];\
 768             ((uint8_t*)dest)[3]= b[Y2];\
 769             ((uint8_t*)dest)[4]= g[Y2];\
 770             ((uint8_t*)dest)[5]= r[Y2];\
 771             dest+=6;\
 772         }\
 773         break;\
 774     case PIX_FMT_RGB565BE:\
 775     case PIX_FMT_RGB565LE:\
 776     case PIX_FMT_BGR565BE:\
 777     case PIX_FMT_BGR565LE:\
 778         {\
 779             const int dr1= dither_2x2_8[y&1    ][0];\
 780             const int dg1= dither_2x2_4[y&1    ][0];\
 781             const int db1= dither_2x2_8[(y&1)^1][0];\
 782             const int dr2= dither_2x2_8[y&1    ][1];\
 783             const int dg2= dither_2x2_4[y&1    ][1];\
 784             const int db2= dither_2x2_8[(y&1)^1][1];\
 785             func(uint16_t,0)\
 786                 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
 787                 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
 788             }\
 789         }\
 790         break;\
 791     case PIX_FMT_RGB555BE:\
 792     case PIX_FMT_RGB555LE:\
 793     case PIX_FMT_BGR555BE:\
 794     case PIX_FMT_BGR555LE:\
 795         {\
 796             const int dr1= dither_2x2_8[y&1    ][0];\
 797             const int dg1= dither_2x2_8[y&1    ][1];\
 798             const int db1= dither_2x2_8[(y&1)^1][0];\
 799             const int dr2= dither_2x2_8[y&1    ][1];\
 800             const int dg2= dither_2x2_8[y&1    ][0];\
 801             const int db2= dither_2x2_8[(y&1)^1][1];\
 802             func(uint16_t,0)\
 803                 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
 804                 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
 805             }\
 806         }\
 807         break;\
 808     case PIX_FMT_RGB8:\
 809     case PIX_FMT_BGR8:\
 810         {\
 811             const uint8_t * const d64= dither_8x8_73[y&7];\
 812             const uint8_t * const d32= dither_8x8_32[y&7];\
 813             func(uint8_t,0)\
 814                 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
 815                 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
 816             }\
 817         }\
 818         break;\
 819     case PIX_FMT_RGB4:\
 820     case PIX_FMT_BGR4:\
 821         {\
 822             const uint8_t * const d64= dither_8x8_73 [y&7];\
 823             const uint8_t * const d128=dither_8x8_220[y&7];\
 824             func(uint8_t,0)\
 825                 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
 826                                  + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
 827             }\
 828         }\
 829         break;\
 830     case PIX_FMT_RGB4_BYTE:\
 831     case PIX_FMT_BGR4_BYTE:\
 832         {\
 833             const uint8_t * const d64= dither_8x8_73 [y&7];\
 834             const uint8_t * const d128=dither_8x8_220[y&7];\
 835             func(uint8_t,0)\
 836                 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
 837                 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
 838             }\
 839         }\
 840         break;\
 841     case PIX_FMT_MONOBLACK:\
 842     case PIX_FMT_MONOWHITE:\
 843         {\
 844             func_monoblack\
 845         }\
 846         break;\
 847     case PIX_FMT_YUYV422:\
 848         func2\
 849             ((uint8_t*)dest)[2*i2+0]= Y1;\
 850             ((uint8_t*)dest)[2*i2+1]= U;\
 851             ((uint8_t*)dest)[2*i2+2]= Y2;\
 852             ((uint8_t*)dest)[2*i2+3]= V;\
 853         }                \
 854         break;\
 855     case PIX_FMT_UYVY422:\
 856         func2\
 857             ((uint8_t*)dest)[2*i2+0]= U;\
 858             ((uint8_t*)dest)[2*i2+1]= Y1;\
 859             ((uint8_t*)dest)[2*i2+2]= V;\
 860             ((uint8_t*)dest)[2*i2+3]= Y2;\
 861         }                \
 862         break;\
 863     case PIX_FMT_GRAY16BE:\
 864         func_g16\
 865             ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
 866             ((uint8_t*)dest)[2*i2+1]= Y1;\
 867             ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
 868             ((uint8_t*)dest)[2*i2+3]= Y2;\
 869         }                \
 870         break;\
 871     case PIX_FMT_GRAY16LE:\
 872         func_g16\
 873             ((uint8_t*)dest)[2*i2+0]= Y1;\
 874             ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
 875             ((uint8_t*)dest)[2*i2+2]= Y2;\
 876             ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
 877         }                \
 878         break;\
 879     }
 880
 881 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 882                                   const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 883                                   const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
 884 {
 885     int i;
 886     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)
 887 }
 888
 889 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 890                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 891                                     const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
 892 {
 893     int i;
 894     int step= c->dstFormatBpp/8;
 895     int aidx= 3;
 896
 897     switch(c->dstFormat) {
 898     case PIX_FMT_ARGB:
 899         dest++;
 900         aidx= 0;
 901     case PIX_FMT_RGB24:
 902         aidx--;
 903     case PIX_FMT_RGBA:
 904         if (CONFIG_SMALL) {
 905             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
 906             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
 907                 dest[aidx]= needAlpha ? A : 255;
 908                 dest[0]= R>>22;
 909                 dest[1]= G>>22;
 910                 dest[2]= B>>22;
 911                 dest+= step;
 912             }
 913         } else {
 914             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 915                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
 916                     dest[aidx]= A;
 917                     dest[0]= R>>22;
 918                     dest[1]= G>>22;
 919                     dest[2]= B>>22;
 920                     dest+= step;
 921                 }
 922             } else {
 923                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
 924                     dest[aidx]= 255;
 925                     dest[0]= R>>22;
 926                     dest[1]= G>>22;
 927                     dest[2]= B>>22;
 928                     dest+= step;
 929                 }
 930             }
 931         }
 932         break;
 933     case PIX_FMT_ABGR:
 934         dest++;
 935         aidx= 0;
 936     case PIX_FMT_BGR24:
 937         aidx--;
 938     case PIX_FMT_BGRA:
 939         if (CONFIG_SMALL) {
 940             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
 941             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
 942                 dest[aidx]= needAlpha ? A : 255;
 943                 dest[0]= B>>22;
 944                 dest[1]= G>>22;
 945                 dest[2]= R>>22;
 946                 dest+= step;
 947             }
 948         } else {
 949             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 950                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
 951                     dest[aidx]= A;
 952                     dest[0]= B>>22;
 953                     dest[1]= G>>22;
 954                     dest[2]= R>>22;
 955                     dest+= step;
 956                 }
 957             } else {
 958                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
 959                     dest[aidx]= 255;
 960                     dest[0]= B>>22;
 961                     dest[1]= G>>22;
 962                     dest[2]= R>>22;
 963                     dest+= step;
 964                 }
 965             }
 966         }
 967         break;
 968     default:
 969         assert(0);
 970     }
 971 }
 972
 973 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
 974 {
 975     int i;
 976     uint8_t *ptr = plane + stride*y;
 977     for (i=0; i<height; i++) {
 978         memset(ptr, val, width);
 979         ptr += stride;
 980     }
 981 }
 982
 983 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width,
 984                             uint32_t *unused)
 985 {
 986     int i;
 987     for (i = 0; i < width; i++) {
 988         int r = src[i*6+0];
 989         int g = src[i*6+2];
 990         int b = src[i*6+4];
 991
 992         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
 993     }
 994 }
 995
 996 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
 997                              const uint8_t *src1, const uint8_t *src2,
 998                              int width, uint32_t *unused)
 999 {
1000     int i;
1001     assert(src1==src2);
1002     for (i = 0; i < width; i++) {
1003         int r = src1[6*i + 0];
1004         int g = src1[6*i + 2];
1005         int b = src1[6*i + 4];
1006
1007         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1008         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1009     }
1010 }
1011
1012 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1013                                   const uint8_t *src1, const uint8_t *src2,
1014                                   int width, uint32_t *unused)
1015 {
1016     int i;
1017     assert(src1==src2);
1018     for (i = 0; i < width; i++) {
1019         int r= src1[12*i + 0] + src1[12*i + 6];
1020         int g= src1[12*i + 2] + src1[12*i + 8];
1021         int b= src1[12*i + 4] + src1[12*i + 10];
1022
1023         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1024         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1025     }
1026 }
1027
1028 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1029 static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1030 {\
1031     int i;\
1032     for (i=0; i<width; i++) {\
1033         int b= (((const type*)src)[i]>>shb)&maskb;\
1034         int g= (((const type*)src)[i]>>shg)&maskg;\
1035         int r= (((const type*)src)[i]>>shr)&maskr;\
1036 \
1037         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1038     }\
1039 }
1040
1041 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1042 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1043 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1044 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1045 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1046 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1047
1048 static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1049 {
1050     int i;
1051     for (i=0; i<width; i++) {
1052         dst[i]= src[4*i];
1053     }
1054 }
1055
1056 #define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1057 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1058 {\
1059     int i;\
1060     for (i=0; i<width; i++) {\
1061         int b= (((const type*)src)[i]&maskb)>>shb;\
1062         int g= (((const type*)src)[i]&maskg)>>shg;\
1063         int r= (((const type*)src)[i]&maskr)>>shr;\
1064 \
1065         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1066         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1067     }\
1068 }\
1069 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1070 {\
1071     int i;\
1072     for (i=0; i<width; i++) {\
1073         int pix0= ((const type*)src)[2*i+0];\
1074         int pix1= ((const type*)src)[2*i+1];\
1075         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1076         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1077         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1078         g&= maskg|(2*maskg);\
1079 \
1080         g>>=shg;\
1081 \
1082         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1083         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1084     }\
1085 }
1086
1087 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1088 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1089 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1090 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1091 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1092 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1093
1094 static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1095 {
1096     int i;
1097     for (i=0; i<width; i++) {
1098         int d= src[i];
1099
1100         dst[i]= pal[d] & 0xFF;
1101     }
1102 }
1103
1104 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1105                            const uint8_t *src1, const uint8_t *src2,
1106                            long width, uint32_t *pal)
1107 {
1108     int i;
1109     assert(src1 == src2);
1110     for (i=0; i<width; i++) {
1111         int p= pal[src1[i]];
1112
1113         dstU[i]= p>>8;
1114         dstV[i]= p>>16;
1115     }
1116 }
1117
1118 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1119 {
1120     int i, j;
1121     for (i=0; i<width/8; i++) {
1122         int d= ~src[i];
1123         for(j=0; j<8; j++)
1124             dst[8*i+j]= ((d>>(7-j))&1)*255;
1125     }
1126 }
1127
1128 static inline void monoblack2Y(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 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1139 //Plain C versions
1140 #if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1141 #define COMPILE_C
1142 #endif
1143
1144 #if ARCH_PPC
1145 #if HAVE_ALTIVEC
1146 #define COMPILE_ALTIVEC
1147 #endif
1148 #endif //ARCH_PPC
1149
1150 #if ARCH_X86
1151
1152 #if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1153 #define COMPILE_MMX
1154 #endif
1155
1156 #if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1157 #define COMPILE_MMX2
1158 #endif
1159
1160 #if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1161 #define COMPILE_3DNOW
1162 #endif
1163 #endif //ARCH_X86
1164
1165 #define COMPILE_TEMPLATE_MMX 0
1166 #define COMPILE_TEMPLATE_MMX2 0
1167 #define COMPILE_TEMPLATE_AMD3DNOW 0
1168 #define COMPILE_TEMPLATE_ALTIVEC 0
1169
1170 #ifdef COMPILE_C
1171 #define RENAME(a) a ## _C
1172 #include "swscale_template.c"
1173 #endif
1174
1175 #ifdef COMPILE_ALTIVEC
1176 #undef RENAME
1177 #undef COMPILE_TEMPLATE_ALTIVEC
1178 #define COMPILE_TEMPLATE_ALTIVEC 1
1179 #define RENAME(a) a ## _altivec
1180 #include "swscale_template.c"
1181 #endif
1182
1183 #if ARCH_X86
1184
1185 //MMX versions
1186 #ifdef COMPILE_MMX
1187 #undef RENAME
1188 #undef COMPILE_TEMPLATE_MMX
1189 #undef COMPILE_TEMPLATE_MMX2
1190 #undef COMPILE_TEMPLATE_AMD3DNOW
1191 #define COMPILE_TEMPLATE_MMX 1
1192 #define COMPILE_TEMPLATE_MMX2 0
1193 #define COMPILE_TEMPLATE_AMD3DNOW 0
1194 #define RENAME(a) a ## _MMX
1195 #include "swscale_template.c"
1196 #endif
1197
1198 //MMX2 versions
1199 #ifdef COMPILE_MMX2
1200 #undef RENAME
1201 #undef COMPILE_TEMPLATE_MMX
1202 #undef COMPILE_TEMPLATE_MMX2
1203 #undef COMPILE_TEMPLATE_AMD3DNOW
1204 #define COMPILE_TEMPLATE_MMX 1
1205 #define COMPILE_TEMPLATE_MMX2 1
1206 #define COMPILE_TEMPLATE_AMD3DNOW 0
1207 #define RENAME(a) a ## _MMX2
1208 #include "swscale_template.c"
1209 #endif
1210
1211 //3DNOW versions
1212 #ifdef COMPILE_3DNOW
1213 #undef RENAME
1214 #undef COMPILE_TEMPLATE_MMX
1215 #undef COMPILE_TEMPLATE_MMX2
1216 #undef COMPILE_TEMPLATE_AMD3DNOW
1217 #define COMPILE_TEMPLATE_MMX 1
1218 #define COMPILE_TEMPLATE_MMX2 0
1219 #define COMPILE_TEMPLATE_AMD3DNOW 1
1220 #define RENAME(a) a ## _3DNow
1221 #include "swscale_template.c"
1222 #endif
1223
1224 #endif //ARCH_X86
1225
1226 SwsFunc ff_getSwsFunc(SwsContext *c)
1227 {
1228 #if CONFIG_RUNTIME_CPUDETECT
1229     int flags = c->flags;
1230
1231 #if ARCH_X86 && CONFIG_GPL
1232     // ordered per speed fastest first
1233     if (flags & SWS_CPU_CAPS_MMX2) {
1234         sws_init_swScale_MMX2(c);
1235         return swScale_MMX2;
1236     } else if (flags & SWS_CPU_CAPS_3DNOW) {
1237         sws_init_swScale_3DNow(c);
1238         return swScale_3DNow;
1239     } else if (flags & SWS_CPU_CAPS_MMX) {
1240         sws_init_swScale_MMX(c);
1241         return swScale_MMX;
1242     } else {
1243         sws_init_swScale_C(c);
1244         return swScale_C;
1245     }
1246
1247 #else
1248 #if ARCH_PPC && COMPILE_ALTIVEC
1249     if (flags & SWS_CPU_CAPS_ALTIVEC) {
1250         sws_init_swScale_altivec(c);
1251         return swScale_altivec;
1252     } else {
1253         sws_init_swScale_C(c);
1254         return swScale_C;
1255     }
1256 #endif
1257     sws_init_swScale_C(c);
1258     return swScale_C;
1259 #endif /* ARCH_X86 && CONFIG_GPL */
1260 #else //CONFIG_RUNTIME_CPUDETECT
1261 #if   COMPILE_TEMPLATE_MMX2
1262     sws_init_swScale_MMX2(c);
1263     return swScale_MMX2;
1264 #elif COMPILE_TEMPLATE_AMD3DNOW
1265     sws_init_swScale_3DNow(c);
1266     return swScale_3DNow;
1267 #elif COMPILE_TEMPLATE_MMX
1268     sws_init_swScale_MMX(c);
1269     return swScale_MMX;
1270 #elif COMPILE_TEMPLATE_ALTIVEC
1271     sws_init_swScale_altivec(c);
1272     return swScale_altivec;
1273 #else
1274     sws_init_swScale_C(c);
1275     return swScale_C;
1276 #endif
1277 #endif //!CONFIG_RUNTIME_CPUDETECT
1278 }
1279
1280 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1281                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1282 {
1283     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1284     /* Copy Y plane */
1285     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1286         memcpy(dst, src[0], srcSliceH*dstStride[0]);
1287     else {
1288         int i;
1289         const uint8_t *srcPtr= src[0];
1290         uint8_t *dstPtr= dst;
1291         for (i=0; i<srcSliceH; i++) {
1292             memcpy(dstPtr, srcPtr, c->srcW);
1293             srcPtr+= srcStride[0];
1294             dstPtr+= dstStride[0];
1295         }
1296     }
1297     dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1298     if (c->dstFormat == PIX_FMT_NV12)
1299         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1300     else
1301         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1302
1303     return srcSliceH;
1304 }
1305
1306 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1307                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1308 {
1309     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1310
1311     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1312
1313     return srcSliceH;
1314 }
1315
1316 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1317                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1318 {
1319     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1320
1321     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1322
1323     return srcSliceH;
1324 }
1325
1326 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1327                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1328 {
1329     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1330
1331     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1332
1333     return srcSliceH;
1334 }
1335
1336 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1337                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1338 {
1339     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1340
1341     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1342
1343     return srcSliceH;
1344 }
1345
1346 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1347                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1348 {
1349     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1350     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1351     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1352
1353     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1354
1355     if (dstParam[3])
1356         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1357
1358     return srcSliceH;
1359 }
1360
1361 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1362                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1363 {
1364     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1365     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1366     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1367
1368     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1369
1370     return srcSliceH;
1371 }
1372
1373 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1374                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1375 {
1376     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1377     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1378     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1379
1380     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1381
1382     if (dstParam[3])
1383         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1384
1385     return srcSliceH;
1386 }
1387
1388 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1389                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1390 {
1391     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1392     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1393     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1394
1395     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1396
1397     return srcSliceH;
1398 }
1399
1400 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1401                            int srcSliceH, uint8_t* dst[], int dstStride[])
1402 {
1403     const enum PixelFormat srcFormat= c->srcFormat;
1404     const enum PixelFormat dstFormat= c->dstFormat;
1405     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1406                  const uint8_t *palette)=NULL;
1407     int i;
1408     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1409     const uint8_t *srcPtr= src[0];
1410
1411     if (usePal(srcFormat)) {
1412         switch (dstFormat) {
1413         case PIX_FMT_RGB32  : conv = palette8topacked32; break;
1414         case PIX_FMT_BGR32  : conv = palette8topacked32; break;
1415         case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
1416         case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
1417         case PIX_FMT_RGB24  : conv = palette8topacked24; break;
1418         case PIX_FMT_BGR24  : conv = palette8topacked24; break;
1419         }
1420     }
1421
1422     if (!conv)
1423         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1424                sws_format_name(srcFormat), sws_format_name(dstFormat));
1425     else {
1426         for (i=0; i<srcSliceH; i++) {
1427             conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1428             srcPtr+= srcStride[0];
1429             dstPtr+= dstStride[0];
1430         }
1431     }
1432
1433     return srcSliceH;
1434 }
1435
1436 #define isRGBA32(x) (            \
1437            (x) == PIX_FMT_ARGB   \
1438         || (x) == PIX_FMT_RGBA   \
1439         || (x) == PIX_FMT_BGRA   \
1440         || (x) == PIX_FMT_ABGR   \
1441         )
1442
1443 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1444 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1445                            int srcSliceH, uint8_t* dst[], int dstStride[])
1446 {
1447     const enum PixelFormat srcFormat= c->srcFormat;
1448     const enum PixelFormat dstFormat= c->dstFormat;
1449     const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1450     const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1451     const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1452     const int dstId= c->dstFormatBpp >> 2;
1453     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1454
1455 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1456
1457     if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1458         if (     CONV_IS(ABGR, RGBA)
1459               || CONV_IS(ARGB, BGRA)
1460               || CONV_IS(BGRA, ARGB)
1461               || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1462         else if (CONV_IS(ABGR, ARGB)
1463               || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1464         else if (CONV_IS(ABGR, BGRA)
1465               || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1466         else if (CONV_IS(BGRA, RGBA)
1467               || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1468         else if (CONV_IS(BGRA, ABGR)
1469               || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1470     } else
1471     /* BGR -> BGR */
1472     if (  (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1473        || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1474         switch(srcId | (dstId<<4)) {
1475         case 0x34: conv= rgb16to15; break;
1476         case 0x36: conv= rgb24to15; break;
1477         case 0x38: conv= rgb32to15; break;
1478         case 0x43: conv= rgb15to16; break;
1479         case 0x46: conv= rgb24to16; break;
1480         case 0x48: conv= rgb32to16; break;
1481         case 0x63: conv= rgb15to24; break;
1482         case 0x64: conv= rgb16to24; break;
1483         case 0x68: conv= rgb32to24; break;
1484         case 0x83: conv= rgb15to32; break;
1485         case 0x84: conv= rgb16to32; break;
1486         case 0x86: conv= rgb24to32; break;
1487         }
1488     } else if (  (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1489              || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1490         switch(srcId | (dstId<<4)) {
1491         case 0x33: conv= rgb15tobgr15; break;
1492         case 0x34: conv= rgb16tobgr15; break;
1493         case 0x36: conv= rgb24tobgr15; break;
1494         case 0x38: conv= rgb32tobgr15; break;
1495         case 0x43: conv= rgb15tobgr16; break;
1496         case 0x44: conv= rgb16tobgr16; break;
1497         case 0x46: conv= rgb24tobgr16; break;
1498         case 0x48: conv= rgb32tobgr16; break;
1499         case 0x63: conv= rgb15tobgr24; break;
1500         case 0x64: conv= rgb16tobgr24; break;
1501         case 0x66: conv= rgb24tobgr24; break;
1502         case 0x68: conv= rgb32tobgr24; break;
1503         case 0x83: conv= rgb15tobgr32; break;
1504         case 0x84: conv= rgb16tobgr32; break;
1505         case 0x86: conv= rgb24tobgr32; break;
1506         }
1507     }
1508
1509     if (!conv) {
1510         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1511                sws_format_name(srcFormat), sws_format_name(dstFormat));
1512     } else {
1513         const uint8_t *srcPtr= src[0];
1514               uint8_t *dstPtr= dst[0];
1515         if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1516             srcPtr += ALT32_CORR;
1517
1518         if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1519             dstPtr += ALT32_CORR;
1520
1521         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1522             conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1523         else {
1524             int i;
1525             dstPtr += dstStride[0]*srcSliceY;
1526
1527             for (i=0; i<srcSliceH; i++) {
1528                 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1529                 srcPtr+= srcStride[0];
1530                 dstPtr+= dstStride[0];
1531             }
1532         }
1533     }
1534     return srcSliceH;
1535 }
1536
1537 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1538                               int srcSliceH, uint8_t* dst[], int dstStride[])
1539 {
1540     rgb24toyv12(
1541         src[0],
1542         dst[0]+ srcSliceY    *dstStride[0],
1543         dst[1]+(srcSliceY>>1)*dstStride[1],
1544         dst[2]+(srcSliceY>>1)*dstStride[2],
1545         c->srcW, srcSliceH,
1546         dstStride[0], dstStride[1], srcStride[0]);
1547     if (dst[3])
1548         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1549     return srcSliceH;
1550 }
1551
1552 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1553                              int srcSliceH, uint8_t* dst[], int dstStride[])
1554 {
1555     int i;
1556
1557     /* copy Y */
1558     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
1559         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1560     else {
1561         const uint8_t *srcPtr= src[0];
1562         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1563
1564         for (i=0; i<srcSliceH; i++) {
1565             memcpy(dstPtr, srcPtr, c->srcW);
1566             srcPtr+= srcStride[0];
1567             dstPtr+= dstStride[0];
1568         }
1569     }
1570
1571     if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {
1572         planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1573                  srcSliceH >> 2, srcStride[1], dstStride[1]);
1574         planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1575                  srcSliceH >> 2, srcStride[2], dstStride[2]);
1576     } else {
1577         planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1578                  srcSliceH >> 2, srcStride[1], dstStride[2]);
1579         planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1580                  srcSliceH >> 2, srcStride[2], dstStride[1]);
1581     }
1582     if (dst[3])
1583         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1584     return srcSliceH;
1585 }
1586
1587 /* unscaled copy like stuff (assumes nearly identical formats) */
1588 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1589                              int srcSliceH, uint8_t* dst[], int dstStride[])
1590 {
1591     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1592         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1593     else {
1594         int i;
1595         const uint8_t *srcPtr= src[0];
1596         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1597         int length=0;
1598
1599         /* universal length finder */
1600         while(length+c->srcW <= FFABS(dstStride[0])
1601            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1602         assert(length!=0);
1603
1604         for (i=0; i<srcSliceH; i++) {
1605             memcpy(dstPtr, srcPtr, length);
1606             srcPtr+= srcStride[0];
1607             dstPtr+= dstStride[0];
1608         }
1609     }
1610     return srcSliceH;
1611 }
1612
1613 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1614                              int srcSliceH, uint8_t* dst[], int dstStride[])
1615 {
1616     int plane, i, j;
1617     for (plane=0; plane<4; plane++) {
1618         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
1619         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1620         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1621         const uint8_t *srcPtr= src[plane];
1622         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1623
1624         if (!dst[plane]) continue;
1625         // ignore palette for GRAY8
1626         if (plane == 1 && !dst[2]) continue;
1627         if (!src[plane] || (plane == 1 && !src[2])) {
1628             if(is16BPS(c->dstFormat))
1629                 length*=2;
1630             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1631         } else {
1632             if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1633                 if (!isBE(c->srcFormat)) srcPtr++;
1634                 for (i=0; i<height; i++) {
1635                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1636                     srcPtr+= srcStride[plane];
1637                     dstPtr+= dstStride[plane];
1638                 }
1639             } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1640                 for (i=0; i<height; i++) {
1641                     for (j=0; j<length; j++) {
1642                         dstPtr[ j<<1   ] = srcPtr[j];
1643                         dstPtr[(j<<1)+1] = srcPtr[j];
1644                     }
1645                     srcPtr+= srcStride[plane];
1646                     dstPtr+= dstStride[plane];
1647                 }
1648             } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1649                   && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1650
1651                 for (i=0; i<height; i++) {
1652                     for (j=0; j<length; j++)
1653                         ((uint16_t*)dstPtr)[j] = bswap_16(((const uint16_t*)srcPtr)[j]);
1654                     srcPtr+= srcStride[plane];
1655                     dstPtr+= dstStride[plane];
1656                 }
1657             } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
1658                 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
1659             else {
1660                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1661                     length*=2;
1662                 for (i=0; i<height; i++) {
1663                     memcpy(dstPtr, srcPtr, length);
1664                     srcPtr+= srcStride[plane];
1665                     dstPtr+= dstStride[plane];
1666                 }
1667             }
1668         }
1669     }
1670     return srcSliceH;
1671 }
1672
1673 int ff_hardcodedcpuflags(void)
1674 {
1675     int flags = 0;
1676 #if   COMPILE_TEMPLATE_MMX2
1677     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1678 #elif COMPILE_TEMPLATE_AMD3DNOW
1679     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1680 #elif COMPILE_TEMPLATE_MMX
1681     flags |= SWS_CPU_CAPS_MMX;
1682 #elif COMPILE_TEMPLATE_ALTIVEC
1683     flags |= SWS_CPU_CAPS_ALTIVEC;
1684 #elif ARCH_BFIN
1685     flags |= SWS_CPU_CAPS_BFIN;
1686 #endif
1687     return flags;
1688 }
1689
1690 void ff_get_unscaled_swscale(SwsContext *c)
1691 {
1692     const enum PixelFormat srcFormat = c->srcFormat;
1693     const enum PixelFormat dstFormat = c->dstFormat;
1694     const int flags = c->flags;
1695     const int dstH = c->dstH;
1696     int needsDither;
1697
1698     needsDither= isAnyRGB(dstFormat)
1699         &&  c->dstFormatBpp < 24
1700         && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1701
1702     /* yv12_to_nv12 */
1703     if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1704         c->swScale= planarToNv12Wrapper;
1705     }
1706     /* yuv2bgr */
1707     if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1708         && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1709         c->swScale= ff_yuv2rgb_get_func_ptr(c);
1710     }
1711
1712     if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1713         c->swScale= yvu9ToYv12Wrapper;
1714     }
1715
1716     /* bgr24toYV12 */
1717     if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1718         c->swScale= bgr24ToYv12Wrapper;
1719
1720     /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1721     if (   isAnyRGB(srcFormat)
1722         && isAnyRGB(dstFormat)
1723         && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
1724         && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
1725         && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
1726         && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
1727         && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1728         && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1729         && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1730         && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1731         && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
1732         && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
1733         && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1734         c->swScale= rgbToRgbWrapper;
1735
1736     if ((usePal(srcFormat) && (
1737         dstFormat == PIX_FMT_RGB32   ||
1738         dstFormat == PIX_FMT_RGB32_1 ||
1739         dstFormat == PIX_FMT_RGB24   ||
1740         dstFormat == PIX_FMT_BGR32   ||
1741         dstFormat == PIX_FMT_BGR32_1 ||
1742         dstFormat == PIX_FMT_BGR24)))
1743         c->swScale= palToRgbWrapper;
1744
1745     if (srcFormat == PIX_FMT_YUV422P) {
1746         if (dstFormat == PIX_FMT_YUYV422)
1747             c->swScale= yuv422pToYuy2Wrapper;
1748         else if (dstFormat == PIX_FMT_UYVY422)
1749             c->swScale= yuv422pToUyvyWrapper;
1750     }
1751
1752     /* LQ converters if -sws 0 or -sws 4*/
1753     if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1754         /* yv12_to_yuy2 */
1755         if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1756             if (dstFormat == PIX_FMT_YUYV422)
1757                 c->swScale= planarToYuy2Wrapper;
1758             else if (dstFormat == PIX_FMT_UYVY422)
1759                 c->swScale= planarToUyvyWrapper;
1760         }
1761     }
1762     if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1763         c->swScale= yuyvToYuv420Wrapper;
1764     if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1765         c->swScale= uyvyToYuv420Wrapper;
1766     if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1767         c->swScale= yuyvToYuv422Wrapper;
1768     if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1769         c->swScale= uyvyToYuv422Wrapper;
1770
1771 #ifdef COMPILE_ALTIVEC
1772     if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1773         !(c->flags & SWS_BITEXACT) &&
1774         srcFormat == PIX_FMT_YUV420P) {
1775         // unscaled YV12 -> packed YUV, we want speed
1776         if (dstFormat == PIX_FMT_YUYV422)
1777             c->swScale= yv12toyuy2_unscaled_altivec;
1778         else if (dstFormat == PIX_FMT_UYVY422)
1779             c->swScale= yv12touyvy_unscaled_altivec;
1780     }
1781 #endif
1782
1783     /* simple copy */
1784     if (  srcFormat == dstFormat
1785         || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1786         || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1787         || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1788         || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1789         || (isGray(dstFormat) && isGray(srcFormat))
1790         || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1791             && c->chrDstHSubSample == c->chrSrcHSubSample
1792             && c->chrDstVSubSample == c->chrSrcVSubSample
1793             && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1794             && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1795     {
1796         if (isPacked(c->srcFormat))
1797             c->swScale= packedCopyWrapper;
1798         else /* Planar YUV or gray */
1799             c->swScale= planarCopyWrapper;
1800     }
1801 #if ARCH_BFIN
1802     if (flags & SWS_CPU_CAPS_BFIN)
1803         ff_bfin_get_unscaled_swscale (c);
1804 #endif
1805 }
1806
1807 static void reset_ptr(const uint8_t* src[], int format)
1808 {
1809     if(!isALPHA(format))
1810         src[3]=NULL;
1811     if(!isPlanarYUV(format)) {
1812         src[3]=src[2]=NULL;
1813
1814         if (!usePal(format))
1815             src[1]= NULL;
1816     }
1817 }
1818
1819 /**
1820  * swscale wrapper, so we don't need to export the SwsContext.
1821  * Assumes planar YUV to be in YUV order instead of YVU.
1822  */
1823 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1824               int srcSliceH, uint8_t* const dst[], const int dstStride[])
1825 {
1826     int i;
1827     const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1828     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1829
1830     // do not mess up sliceDir if we have a "trailing" 0-size slice
1831     if (srcSliceH == 0)
1832         return 0;
1833
1834     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1835         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1836         return 0;
1837     }
1838     if (c->sliceDir == 0) {
1839         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1840     }
1841
1842     if (usePal(c->srcFormat)) {
1843         for (i=0; i<256; i++) {
1844             int p, r, g, b,y,u,v;
1845             if(c->srcFormat == PIX_FMT_PAL8) {
1846                 p=((const uint32_t*)(src[1]))[i];
1847                 r= (p>>16)&0xFF;
1848                 g= (p>> 8)&0xFF;
1849                 b=  p     &0xFF;
1850             } else if(c->srcFormat == PIX_FMT_RGB8) {
1851                 r= (i>>5    )*36;
1852                 g= ((i>>2)&7)*36;
1853                 b= (i&3     )*85;
1854             } else if(c->srcFormat == PIX_FMT_BGR8) {
1855                 b= (i>>6    )*85;
1856                 g= ((i>>3)&7)*36;
1857                 r= (i&7     )*36;
1858             } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
1859                 r= (i>>3    )*255;
1860                 g= ((i>>1)&3)*85;
1861                 b= (i&1     )*255;
1862             } else if(c->srcFormat == PIX_FMT_GRAY8) {
1863                 r = g = b = i;
1864             } else {
1865                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
1866                 b= (i>>3    )*255;
1867                 g= ((i>>1)&3)*85;
1868                 r= (i&1     )*255;
1869             }
1870             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1871             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1872             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1873             c->pal_yuv[i]= y + (u<<8) + (v<<16);
1874
1875             switch(c->dstFormat) {
1876             case PIX_FMT_BGR32:
1877 #if !HAVE_BIGENDIAN
1878             case PIX_FMT_RGB24:
1879 #endif
1880                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);
1881                 break;
1882             case PIX_FMT_BGR32_1:
1883 #if HAVE_BIGENDIAN
1884             case PIX_FMT_BGR24:
1885 #endif
1886                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
1887                 break;
1888             case PIX_FMT_RGB32_1:
1889 #if HAVE_BIGENDIAN
1890             case PIX_FMT_RGB24:
1891 #endif
1892                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
1893                 break;
1894             case PIX_FMT_RGB32:
1895 #if !HAVE_BIGENDIAN
1896             case PIX_FMT_BGR24:
1897 #endif
1898             default:
1899                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);
1900             }
1901         }
1902     }
1903
1904     // copy strides, so they can safely be modified
1905     if (c->sliceDir == 1) {
1906         // slices go from top to bottom
1907         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
1908         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
1909
1910         reset_ptr(src2, c->srcFormat);
1911         reset_ptr((const uint8_t**)dst2, c->dstFormat);
1912
1913         /* reset slice direction at end of frame */
1914         if (srcSliceY + srcSliceH == c->srcH)
1915             c->sliceDir = 0;
1916
1917         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
1918     } else {
1919         // slices go from bottom to top => we flip the image internally
1920         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
1921         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
1922
1923         src2[0] += (srcSliceH-1)*srcStride[0];
1924         if (!usePal(c->srcFormat))
1925             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
1926         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
1927         src2[3] += (srcSliceH-1)*srcStride[3];
1928         dst2[0] += ( c->dstH                      -1)*dstStride[0];
1929         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
1930         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
1931         dst2[3] += ( c->dstH                      -1)*dstStride[3];
1932
1933         reset_ptr(src2, c->srcFormat);
1934         reset_ptr((const uint8_t**)dst2, c->dstFormat);
1935
1936         /* reset slice direction at end of frame */
1937         if (!srcSliceY)
1938             c->sliceDir = 0;
1939
1940         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
1941     }
1942 }
1943
1944 #if LIBSWSCALE_VERSION_MAJOR < 1
1945 int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
1946                       int srcSliceH, uint8_t* dst[], int dstStride[])
1947 {
1948     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
1949 }
1950 #endif