libswscale/rgb2rgb.c

   1 /*
   2  *
   3  *  rgb2rgb.c, Software RGB to RGB convertor
   4  *  pluralize by Software PAL8 to RGB convertor
   5  *               Software YUV to YUV convertor
   6  *               Software YUV to RGB convertor
   7  *  Written by Nick Kurshev.
   8  *  palette & YUV & runtime CPU stuff by Michael (michaelni@gmx.at)
   9  *
  10  * This file is part of FFmpeg.
  11  *
  12  * FFmpeg is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License as published by
  14  * the Free Software Foundation; either version 2 of the License, or
  15  * (at your option) any later version.
  16  *
  17  * FFmpeg is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  20  * GNU General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU General Public License
  23  * along with FFmpeg; if not, write to the Free Software
  24  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  25  *
  26  * the C code (not assembly, mmx, ...) of this file can be used
  27  * under the LGPL license too
  28  */
  29 #include <inttypes.h>
  30 #include "config.h"
  31 #include "rgb2rgb.h"
  32 #include "swscale.h"
  33 #include "swscale_internal.h"
  34 #include "x86_cpu.h"
  35 #include "bswap.h"
  36 #ifdef USE_FASTMEMCPY
  37 #include "libvo/fastmemcpy.h"
  38 #endif
  39
  40 #define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit
  41
  42 void (*rgb24to32)(const uint8_t *src,uint8_t *dst,long src_size);
  43 void (*rgb24to16)(const uint8_t *src,uint8_t *dst,long src_size);
  44 void (*rgb24to15)(const uint8_t *src,uint8_t *dst,long src_size);
  45 void (*rgb32to24)(const uint8_t *src,uint8_t *dst,long src_size);
  46 void (*rgb32to16)(const uint8_t *src,uint8_t *dst,long src_size);
  47 void (*rgb32to15)(const uint8_t *src,uint8_t *dst,long src_size);
  48 void (*rgb15to16)(const uint8_t *src,uint8_t *dst,long src_size);
  49 void (*rgb15to24)(const uint8_t *src,uint8_t *dst,long src_size);
  50 void (*rgb15to32)(const uint8_t *src,uint8_t *dst,long src_size);
  51 void (*rgb16to15)(const uint8_t *src,uint8_t *dst,long src_size);
  52 void (*rgb16to24)(const uint8_t *src,uint8_t *dst,long src_size);
  53 void (*rgb16to32)(const uint8_t *src,uint8_t *dst,long src_size);
  54 //void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  55 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  56 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  57 void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  58 void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  59 //void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  60 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  61 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  62
  63 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  64         long width, long height,
  65         long lumStride, long chromStride, long dstStride);
  66 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  67         long width, long height,
  68         long lumStride, long chromStride, long dstStride);
  69 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  70         long width, long height,
  71         long lumStride, long chromStride, long dstStride);
  72 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  73         long width, long height,
  74         long lumStride, long chromStride, long srcStride);
  75 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  76         long width, long height,
  77         long lumStride, long chromStride, long srcStride);
  78 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
  79         long srcStride, long dstStride);
  80 void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
  81                             long width, long height, long src1Stride,
  82                             long src2Stride, long dstStride);
  83 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
  84                         uint8_t *dst1, uint8_t *dst2,
  85                         long width, long height,
  86                         long srcStride1, long srcStride2,
  87                         long dstStride1, long dstStride2);
  88 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
  89                         uint8_t *dst,
  90                         long width, long height,
  91                         long srcStride1, long srcStride2,
  92                         long srcStride3, long dstStride);
  93
  94 #if defined(ARCH_X86) && defined(CONFIG_GPL)
  95 static const uint64_t mmx_null  __attribute__((aligned(8))) = 0x0000000000000000ULL;
  96 static const uint64_t mmx_one   __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL;
  97 static const uint64_t mask32b  attribute_used __attribute__((aligned(8))) = 0x000000FF000000FFULL;
  98 static const uint64_t mask32g  attribute_used __attribute__((aligned(8))) = 0x0000FF000000FF00ULL;
  99 static const uint64_t mask32r  attribute_used __attribute__((aligned(8))) = 0x00FF000000FF0000ULL;
 100 static const uint64_t mask32   __attribute__((aligned(8))) = 0x00FFFFFF00FFFFFFULL;
 101 static const uint64_t mask3216br __attribute__((aligned(8)))=0x00F800F800F800F8ULL;
 102 static const uint64_t mask3216g  __attribute__((aligned(8)))=0x0000FC000000FC00ULL;
 103 static const uint64_t mask3215g  __attribute__((aligned(8)))=0x0000F8000000F800ULL;
 104 static const uint64_t mul3216  __attribute__((aligned(8))) = 0x2000000420000004ULL;
 105 static const uint64_t mul3215  __attribute__((aligned(8))) = 0x2000000820000008ULL;
 106 static const uint64_t mask24b  attribute_used __attribute__((aligned(8))) = 0x00FF0000FF0000FFULL;
 107 static const uint64_t mask24g  attribute_used __attribute__((aligned(8))) = 0xFF0000FF0000FF00ULL;
 108 static const uint64_t mask24r  attribute_used __attribute__((aligned(8))) = 0x0000FF0000FF0000ULL;
 109 static const uint64_t mask24l  __attribute__((aligned(8))) = 0x0000000000FFFFFFULL;
 110 static const uint64_t mask24h  __attribute__((aligned(8))) = 0x0000FFFFFF000000ULL;
 111 static const uint64_t mask24hh  __attribute__((aligned(8))) = 0xffff000000000000ULL;
 112 static const uint64_t mask24hhh  __attribute__((aligned(8))) = 0xffffffff00000000ULL;
 113 static const uint64_t mask24hhhh  __attribute__((aligned(8))) = 0xffffffffffff0000ULL;
 114 static const uint64_t mask15b  __attribute__((aligned(8))) = 0x001F001F001F001FULL; /* 00000000 00011111  xxB */
 115 static const uint64_t mask15rg __attribute__((aligned(8))) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000  RGx */
 116 static const uint64_t mask15s  __attribute__((aligned(8))) = 0xFFE0FFE0FFE0FFE0ULL;
 117 static const uint64_t mask15g  __attribute__((aligned(8))) = 0x03E003E003E003E0ULL;
 118 static const uint64_t mask15r  __attribute__((aligned(8))) = 0x7C007C007C007C00ULL;
 119 #define mask16b mask15b
 120 static const uint64_t mask16g  __attribute__((aligned(8))) = 0x07E007E007E007E0ULL;
 121 static const uint64_t mask16r  __attribute__((aligned(8))) = 0xF800F800F800F800ULL;
 122 static const uint64_t red_16mask  __attribute__((aligned(8))) = 0x0000f8000000f800ULL;
 123 static const uint64_t green_16mask __attribute__((aligned(8)))= 0x000007e0000007e0ULL;
 124 static const uint64_t blue_16mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
 125 static const uint64_t red_15mask  __attribute__((aligned(8))) = 0x00007c000000f800ULL;
 126 static const uint64_t green_15mask __attribute__((aligned(8)))= 0x000003e0000007e0ULL;
 127 static const uint64_t blue_15mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
 128
 129 #ifdef FAST_BGR2YV12
 130 static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;
 131 static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
 132 static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
 133 #else
 134 static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;
 135 static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
 136 static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
 137 #endif
 138 static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;
 139 static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL;
 140 static const uint64_t w1111       attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;
 141
 142 #if 0
 143 static volatile uint64_t __attribute__((aligned(8))) b5Dither;
 144 static volatile uint64_t __attribute__((aligned(8))) g5Dither;
 145 static volatile uint64_t __attribute__((aligned(8))) g6Dither;
 146 static volatile uint64_t __attribute__((aligned(8))) r5Dither;
 147
 148 static uint64_t __attribute__((aligned(8))) dither4[2]={
 149         0x0103010301030103LL,
 150         0x0200020002000200LL,};
 151
 152 static uint64_t __attribute__((aligned(8))) dither8[2]={
 153         0x0602060206020602LL,
 154         0x0004000400040004LL,};
 155 #endif
 156 #endif /* defined(ARCH_X86) */
 157
 158 #define RGB2YUV_SHIFT 8
 159 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
 160 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
 161 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 162 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
 163 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
 164 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
 165 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
 166 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 167 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
 168
 169 //Note: we have C, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
 170 //Plain C versions
 171 #undef HAVE_MMX
 172 #undef HAVE_MMX2
 173 #undef HAVE_3DNOW
 174 #undef HAVE_SSE2
 175 #define RENAME(a) a ## _C
 176 #include "rgb2rgb_template.c"
 177
 178 #if defined(ARCH_X86) && defined(CONFIG_GPL)
 179
 180 //MMX versions
 181 #undef RENAME
 182 #define HAVE_MMX
 183 #undef HAVE_MMX2
 184 #undef HAVE_3DNOW
 185 #undef HAVE_SSE2
 186 #define RENAME(a) a ## _MMX
 187 #include "rgb2rgb_template.c"
 188
 189 //MMX2 versions
 190 #undef RENAME
 191 #define HAVE_MMX
 192 #define HAVE_MMX2
 193 #undef HAVE_3DNOW
 194 #undef HAVE_SSE2
 195 #define RENAME(a) a ## _MMX2
 196 #include "rgb2rgb_template.c"
 197
 198 //3DNOW versions
 199 #undef RENAME
 200 #define HAVE_MMX
 201 #undef HAVE_MMX2
 202 #define HAVE_3DNOW
 203 #undef HAVE_SSE2
 204 #define RENAME(a) a ## _3DNOW
 205 #include "rgb2rgb_template.c"
 206
 207 #endif //ARCH_X86 || ARCH_X86_64
 208
 209 /*
 210  rgb15->rgb16 Original by Strepto/Astral
 211  ported to gcc & bugfixed : A'rpi
 212  MMX2, 3DNOW optimization by Nick Kurshev
 213  32bit c version, and and&add trick by Michael Niedermayer
 214 */
 215
 216 void sws_rgb2rgb_init(int flags){
 217 #if (defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX))  && defined(CONFIG_GPL)
 218         if(flags & SWS_CPU_CAPS_MMX2){
 219                 rgb15to16= rgb15to16_MMX2;
 220                 rgb15to24= rgb15to24_MMX2;
 221                 rgb15to32= rgb15to32_MMX2;
 222                 rgb16to24= rgb16to24_MMX2;
 223                 rgb16to32= rgb16to32_MMX2;
 224                 rgb16to15= rgb16to15_MMX2;
 225                 rgb24to16= rgb24to16_MMX2;
 226                 rgb24to15= rgb24to15_MMX2;
 227                 rgb24to32= rgb24to32_MMX2;
 228                 rgb32to16= rgb32to16_MMX2;
 229                 rgb32to15= rgb32to15_MMX2;
 230                 rgb32to24= rgb32to24_MMX2;
 231                 rgb24tobgr15= rgb24tobgr15_MMX2;
 232                 rgb24tobgr16= rgb24tobgr16_MMX2;
 233                 rgb24tobgr24= rgb24tobgr24_MMX2;
 234                 rgb32tobgr32= rgb32tobgr32_MMX2;
 235                 rgb32tobgr16= rgb32tobgr16_MMX2;
 236                 rgb32tobgr15= rgb32tobgr15_MMX2;
 237                 yv12toyuy2= yv12toyuy2_MMX2;
 238                 yv12touyvy= yv12touyvy_MMX2;
 239                 yuv422ptoyuy2= yuv422ptoyuy2_MMX2;
 240                 yuy2toyv12= yuy2toyv12_MMX2;
 241 //              uyvytoyv12= uyvytoyv12_MMX2;
 242 //              yvu9toyv12= yvu9toyv12_MMX2;
 243                 planar2x= planar2x_MMX2;
 244                 rgb24toyv12= rgb24toyv12_MMX2;
 245                 interleaveBytes= interleaveBytes_MMX2;
 246                 vu9_to_vu12= vu9_to_vu12_MMX2;
 247                 yvu9_to_yuy2= yvu9_to_yuy2_MMX2;
 248         }else if(flags & SWS_CPU_CAPS_3DNOW){
 249                 rgb15to16= rgb15to16_3DNOW;
 250                 rgb15to24= rgb15to24_3DNOW;
 251                 rgb15to32= rgb15to32_3DNOW;
 252                 rgb16to24= rgb16to24_3DNOW;
 253                 rgb16to32= rgb16to32_3DNOW;
 254                 rgb16to15= rgb16to15_3DNOW;
 255                 rgb24to16= rgb24to16_3DNOW;
 256                 rgb24to15= rgb24to15_3DNOW;
 257                 rgb24to32= rgb24to32_3DNOW;
 258                 rgb32to16= rgb32to16_3DNOW;
 259                 rgb32to15= rgb32to15_3DNOW;
 260                 rgb32to24= rgb32to24_3DNOW;
 261                 rgb24tobgr15= rgb24tobgr15_3DNOW;
 262                 rgb24tobgr16= rgb24tobgr16_3DNOW;
 263                 rgb24tobgr24= rgb24tobgr24_3DNOW;
 264                 rgb32tobgr32= rgb32tobgr32_3DNOW;
 265                 rgb32tobgr16= rgb32tobgr16_3DNOW;
 266                 rgb32tobgr15= rgb32tobgr15_3DNOW;
 267                 yv12toyuy2= yv12toyuy2_3DNOW;
 268                 yv12touyvy= yv12touyvy_3DNOW;
 269                 yuv422ptoyuy2= yuv422ptoyuy2_3DNOW;
 270                 yuy2toyv12= yuy2toyv12_3DNOW;
 271 //              uyvytoyv12= uyvytoyv12_3DNOW;
 272 //              yvu9toyv12= yvu9toyv12_3DNOW;
 273                 planar2x= planar2x_3DNOW;
 274                 rgb24toyv12= rgb24toyv12_3DNOW;
 275                 interleaveBytes= interleaveBytes_3DNOW;
 276                 vu9_to_vu12= vu9_to_vu12_3DNOW;
 277                 yvu9_to_yuy2= yvu9_to_yuy2_3DNOW;
 278         }else if(flags & SWS_CPU_CAPS_MMX){
 279                 rgb15to16= rgb15to16_MMX;
 280                 rgb15to24= rgb15to24_MMX;
 281                 rgb15to32= rgb15to32_MMX;
 282                 rgb16to24= rgb16to24_MMX;
 283                 rgb16to32= rgb16to32_MMX;
 284                 rgb16to15= rgb16to15_MMX;
 285                 rgb24to16= rgb24to16_MMX;
 286                 rgb24to15= rgb24to15_MMX;
 287                 rgb24to32= rgb24to32_MMX;
 288                 rgb32to16= rgb32to16_MMX;
 289                 rgb32to15= rgb32to15_MMX;
 290                 rgb32to24= rgb32to24_MMX;
 291                 rgb24tobgr15= rgb24tobgr15_MMX;
 292                 rgb24tobgr16= rgb24tobgr16_MMX;
 293                 rgb24tobgr24= rgb24tobgr24_MMX;
 294                 rgb32tobgr32= rgb32tobgr32_MMX;
 295                 rgb32tobgr16= rgb32tobgr16_MMX;
 296                 rgb32tobgr15= rgb32tobgr15_MMX;
 297                 yv12toyuy2= yv12toyuy2_MMX;
 298                 yv12touyvy= yv12touyvy_MMX;
 299                 yuv422ptoyuy2= yuv422ptoyuy2_MMX;
 300                 yuy2toyv12= yuy2toyv12_MMX;
 301 //              uyvytoyv12= uyvytoyv12_MMX;
 302 //              yvu9toyv12= yvu9toyv12_MMX;
 303                 planar2x= planar2x_MMX;
 304                 rgb24toyv12= rgb24toyv12_MMX;
 305                 interleaveBytes= interleaveBytes_MMX;
 306                 vu9_to_vu12= vu9_to_vu12_MMX;
 307                 yvu9_to_yuy2= yvu9_to_yuy2_MMX;
 308         }else
 309 #endif /* defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX) */
 310         {
 311                 rgb15to16= rgb15to16_C;
 312                 rgb15to24= rgb15to24_C;
 313                 rgb15to32= rgb15to32_C;
 314                 rgb16to24= rgb16to24_C;
 315                 rgb16to32= rgb16to32_C;
 316                 rgb16to15= rgb16to15_C;
 317                 rgb24to16= rgb24to16_C;
 318                 rgb24to15= rgb24to15_C;
 319                 rgb24to32= rgb24to32_C;
 320                 rgb32to16= rgb32to16_C;
 321                 rgb32to15= rgb32to15_C;
 322                 rgb32to24= rgb32to24_C;
 323                 rgb24tobgr15= rgb24tobgr15_C;
 324                 rgb24tobgr16= rgb24tobgr16_C;
 325                 rgb24tobgr24= rgb24tobgr24_C;
 326                 rgb32tobgr32= rgb32tobgr32_C;
 327                 rgb32tobgr16= rgb32tobgr16_C;
 328                 rgb32tobgr15= rgb32tobgr15_C;
 329                 yv12toyuy2= yv12toyuy2_C;
 330                 yv12touyvy= yv12touyvy_C;
 331                 yuv422ptoyuy2= yuv422ptoyuy2_C;
 332                 yuy2toyv12= yuy2toyv12_C;
 333 //              uyvytoyv12= uyvytoyv12_C;
 334 //              yvu9toyv12= yvu9toyv12_C;
 335                 planar2x= planar2x_C;
 336                 rgb24toyv12= rgb24toyv12_C;
 337                 interleaveBytes= interleaveBytes_C;
 338                 vu9_to_vu12= vu9_to_vu12_C;
 339                 yvu9_to_yuy2= yvu9_to_yuy2_C;
 340         }
 341 }
 342
 343 /**
 344  * Palette is assumed to contain BGR32.
 345  */
 346 void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 347 {
 348         long i;
 349
 350 /*
 351         for(i=0; i<num_pixels; i++)
 352                 ((unsigned *)dst)[i] = ((unsigned *)palette)[ src[i] ];
 353 */
 354
 355         for(i=0; i<num_pixels; i++)
 356         {
 357                 #ifdef WORDS_BIGENDIAN
 358                         dst[3]= palette[ src[i]*4+2 ];
 359                         dst[2]= palette[ src[i]*4+1 ];
 360                         dst[1]= palette[ src[i]*4+0 ];
 361                 #else
 362                 //FIXME slow?
 363                         dst[0]= palette[ src[i]*4+2 ];
 364                         dst[1]= palette[ src[i]*4+1 ];
 365                         dst[2]= palette[ src[i]*4+0 ];
 366                         //dst[3]= 0; /* do we need this cleansing? */
 367                 #endif
 368                 dst+= 4;
 369         }
 370 }
 371
 372 void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 373 {
 374         long i;
 375         for(i=0; i<num_pixels; i++)
 376         {
 377                 #ifdef WORDS_BIGENDIAN
 378                         dst[3]= palette[ src[i]*4+0 ];
 379                         dst[2]= palette[ src[i]*4+1 ];
 380                         dst[1]= palette[ src[i]*4+2 ];
 381                 #else
 382                         //FIXME slow?
 383                         dst[0]= palette[ src[i]*4+0 ];
 384                         dst[1]= palette[ src[i]*4+1 ];
 385                         dst[2]= palette[ src[i]*4+2 ];
 386                         //dst[3]= 0; /* do we need this cleansing? */
 387                 #endif
 388
 389                 dst+= 4;
 390         }
 391 }
 392
 393 /**
 394  * Palette is assumed to contain BGR32.
 395  */
 396 void palette8torgb24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 397 {
 398         long i;
 399 /*
 400         writes 1 byte o much and might cause alignment issues on some architectures?
 401         for(i=0; i<num_pixels; i++)
 402                 ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
 403 */
 404         for(i=0; i<num_pixels; i++)
 405         {
 406                 //FIXME slow?
 407                 dst[0]= palette[ src[i]*4+2 ];
 408                 dst[1]= palette[ src[i]*4+1 ];
 409                 dst[2]= palette[ src[i]*4+0 ];
 410                 dst+= 3;
 411         }
 412 }
 413
 414 void palette8tobgr24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 415 {
 416         long i;
 417 /*
 418         writes 1 byte o much and might cause alignment issues on some architectures?
 419         for(i=0; i<num_pixels; i++)
 420                 ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
 421 */
 422         for(i=0; i<num_pixels; i++)
 423         {
 424                 //FIXME slow?
 425                 dst[0]= palette[ src[i]*4+0 ];
 426                 dst[1]= palette[ src[i]*4+1 ];
 427                 dst[2]= palette[ src[i]*4+2 ];
 428                 dst+= 3;
 429         }
 430 }
 431
 432 /**
 433  * Palette is assumed to contain bgr16, see rgb32to16 to convert the palette
 434  */
 435 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 436 {
 437         long i;
 438         for(i=0; i<num_pixels; i++)
 439                 ((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
 440 }
 441 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 442 {
 443         long i;
 444         for(i=0; i<num_pixels; i++)
 445                 ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
 446 }
 447
 448 /**
 449  * Palette is assumed to contain BGR15, see rgb32to15 to convert the palette.
 450  */
 451 void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 452 {
 453         long i;
 454         for(i=0; i<num_pixels; i++)
 455                 ((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
 456 }
 457 void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 458 {
 459         long i;
 460         for(i=0; i<num_pixels; i++)
 461                 ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
 462 }
 463
 464 void rgb32tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 465 {
 466         long i;
 467         long num_pixels = src_size >> 2;
 468         for(i=0; i<num_pixels; i++)
 469         {
 470                 #ifdef WORDS_BIGENDIAN
 471                         /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
 472                         dst[3*i + 0] = src[4*i + 1];
 473                         dst[3*i + 1] = src[4*i + 2];
 474                         dst[3*i + 2] = src[4*i + 3];
 475                 #else
 476                         dst[3*i + 0] = src[4*i + 2];
 477                         dst[3*i + 1] = src[4*i + 1];
 478                         dst[3*i + 2] = src[4*i + 0];
 479                 #endif
 480         }
 481 }
 482
 483 void rgb24tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 484 {
 485         long i;
 486         for(i=0; 3*i<src_size; i++)
 487         {
 488                 #ifdef WORDS_BIGENDIAN
 489                         /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
 490                         dst[4*i + 0] = 0;
 491                         dst[4*i + 1] = src[3*i + 0];
 492                         dst[4*i + 2] = src[3*i + 1];
 493                         dst[4*i + 3] = src[3*i + 2];
 494                 #else
 495                         dst[4*i + 0] = src[3*i + 2];
 496                         dst[4*i + 1] = src[3*i + 1];
 497                         dst[4*i + 2] = src[3*i + 0];
 498                         dst[4*i + 3] = 0;
 499                 #endif
 500         }
 501 }
 502
 503 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 504 {
 505         const uint16_t *end;
 506         uint8_t *d = (uint8_t *)dst;
 507         const uint16_t *s = (uint16_t *)src;
 508         end = s + src_size/2;
 509         while(s < end)
 510         {
 511                 register uint16_t bgr;
 512                 bgr = *s++;
 513                 #ifdef WORDS_BIGENDIAN
 514                         *d++ = 0;
 515                         *d++ = (bgr&0x1F)<<3;
 516                         *d++ = (bgr&0x7E0)>>3;
 517                         *d++ = (bgr&0xF800)>>8;
 518                 #else
 519                         *d++ = (bgr&0xF800)>>8;
 520                         *d++ = (bgr&0x7E0)>>3;
 521                         *d++ = (bgr&0x1F)<<3;
 522                         *d++ = 0;
 523                 #endif
 524         }
 525 }
 526
 527 void rgb16tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 528 {
 529         const uint16_t *end;
 530         uint8_t *d = (uint8_t *)dst;
 531         const uint16_t *s = (const uint16_t *)src;
 532         end = s + src_size/2;
 533         while(s < end)
 534         {
 535                 register uint16_t bgr;
 536                 bgr = *s++;
 537                 *d++ = (bgr&0xF800)>>8;
 538                 *d++ = (bgr&0x7E0)>>3;
 539                 *d++ = (bgr&0x1F)<<3;
 540         }
 541 }
 542
 543 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 544 {
 545         long i;
 546         long num_pixels = src_size >> 1;
 547
 548         for(i=0; i<num_pixels; i++)
 549         {
 550             unsigned b,g,r;
 551             register uint16_t rgb;
 552             rgb = src[2*i];
 553             r = rgb&0x1F;
 554             g = (rgb&0x7E0)>>5;
 555             b = (rgb&0xF800)>>11;
 556             dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
 557         }
 558 }
 559
 560 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 561 {
 562         long i;
 563         long num_pixels = src_size >> 1;
 564
 565         for(i=0; i<num_pixels; i++)
 566         {
 567             unsigned b,g,r;
 568             register uint16_t rgb;
 569             rgb = src[2*i];
 570             r = rgb&0x1F;
 571             g = (rgb&0x7E0)>>5;
 572             b = (rgb&0xF800)>>11;
 573             dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10);
 574         }
 575 }
 576
 577 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 578 {
 579         const uint16_t *end;
 580         uint8_t *d = (uint8_t *)dst;
 581         const uint16_t *s = (const uint16_t *)src;
 582         end = s + src_size/2;
 583         while(s < end)
 584         {
 585                 register uint16_t bgr;
 586                 bgr = *s++;
 587                 #ifdef WORDS_BIGENDIAN
 588                         *d++ = 0;
 589                         *d++ = (bgr&0x1F)<<3;
 590                         *d++ = (bgr&0x3E0)>>2;
 591                         *d++ = (bgr&0x7C00)>>7;
 592                 #else
 593                         *d++ = (bgr&0x7C00)>>7;
 594                         *d++ = (bgr&0x3E0)>>2;
 595                         *d++ = (bgr&0x1F)<<3;
 596                         *d++ = 0;
 597                 #endif
 598         }
 599 }
 600
 601 void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 602 {
 603         const uint16_t *end;
 604         uint8_t *d = (uint8_t *)dst;
 605         const uint16_t *s = (uint16_t *)src;
 606         end = s + src_size/2;
 607         while(s < end)
 608         {
 609                 register uint16_t bgr;
 610                 bgr = *s++;
 611                 *d++ = (bgr&0x7C00)>>7;
 612                 *d++ = (bgr&0x3E0)>>2;
 613                 *d++ = (bgr&0x1F)<<3;
 614         }
 615 }
 616
 617 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 618 {
 619         long i;
 620         long num_pixels = src_size >> 1;
 621
 622         for(i=0; i<num_pixels; i++)
 623         {
 624             unsigned b,g,r;
 625             register uint16_t rgb;
 626             rgb = src[2*i];
 627             r = rgb&0x1F;
 628             g = (rgb&0x3E0)>>5;
 629             b = (rgb&0x7C00)>>10;
 630             dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
 631         }
 632 }
 633
 634 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 635 {
 636         long i;
 637         long num_pixels = src_size >> 1;
 638
 639         for(i=0; i<num_pixels; i++)
 640         {
 641             unsigned b,g,r;
 642             register uint16_t rgb;
 643             rgb = src[2*i];
 644             r = rgb&0x1F;
 645             g = (rgb&0x3E0)>>5;
 646             b = (rgb&0x7C00)>>10;
 647             dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10);
 648         }
 649 }
 650
 651 void rgb8tobgr8(const uint8_t *src, uint8_t *dst, long src_size)
 652 {
 653         long i;
 654         long num_pixels = src_size;
 655         for(i=0; i<num_pixels; i++)
 656         {
 657             unsigned b,g,r;
 658             register uint8_t rgb;
 659             rgb = src[i];
 660             r = (rgb&0x07);
 661             g = (rgb&0x38)>>3;
 662             b = (rgb&0xC0)>>6;
 663             dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
 664         }
 665 }