git.sesse.net Git - ffmpeg/blob - 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 program is free software; you can redistribute it and/or modify
  11  * it under the terms of the GNU General Public License as published by
  12  * the Free Software Foundation; either version 2 of the License, or
  13  * (at your option) any later version.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  18  * GNU General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public License
  21  * along with this program; if not, write to the Free Software
  22  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  23  */
  24 #include <inttypes.h>
  25 #include "config.h"
  26 #include "rgb2rgb.h"
  27 #include "swscale.h"
  28 #include "swscale_internal.h"
  29 #include "x86_cpu.h"
  30 #include "bswap.h"
  31 #ifdef USE_FASTMEMCPY
  32 #include "libvo/fastmemcpy.h"
  33 #endif
  34
  35 #define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit
  36
  37 void (*rgb24to32)(const uint8_t *src,uint8_t *dst,long src_size);
  38 void (*rgb24to16)(const uint8_t *src,uint8_t *dst,long src_size);
  39 void (*rgb24to15)(const uint8_t *src,uint8_t *dst,long src_size);
  40 void (*rgb32to24)(const uint8_t *src,uint8_t *dst,long src_size);
  41 void (*rgb32to16)(const uint8_t *src,uint8_t *dst,long src_size);
  42 void (*rgb32to15)(const uint8_t *src,uint8_t *dst,long src_size);
  43 void (*rgb15to16)(const uint8_t *src,uint8_t *dst,long src_size);
  44 void (*rgb15to24)(const uint8_t *src,uint8_t *dst,long src_size);
  45 void (*rgb15to32)(const uint8_t *src,uint8_t *dst,long src_size);
  46 void (*rgb16to15)(const uint8_t *src,uint8_t *dst,long src_size);
  47 void (*rgb16to24)(const uint8_t *src,uint8_t *dst,long src_size);
  48 void (*rgb16to32)(const uint8_t *src,uint8_t *dst,long src_size);
  49 //void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  50 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  51 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  52 void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  53 void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  54 //void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  55 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  56 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  57
  58 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  59         long width, long height,
  60         long lumStride, long chromStride, long dstStride);
  61 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  62         long width, long height,
  63         long lumStride, long chromStride, long dstStride);
  64 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  65         long width, long height,
  66         long lumStride, long chromStride, long dstStride);
  67 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  68         long width, long height,
  69         long lumStride, long chromStride, long srcStride);
  70 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  71         long width, long height,
  72         long lumStride, long chromStride, long srcStride);
  73 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
  74         long srcStride, long dstStride);
  75 void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
  76                             long width, long height, long src1Stride,
  77                             long src2Stride, long dstStride);
  78 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
  79                         uint8_t *dst1, uint8_t *dst2,
  80                         long width, long height,
  81                         long srcStride1, long srcStride2,
  82                         long dstStride1, long dstStride2);
  83 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
  84                         uint8_t *dst,
  85                         long width, long height,
  86                         long srcStride1, long srcStride2,
  87                         long srcStride3, long dstStride);
  88
  89 #if defined(ARCH_X86) || defined(ARCH_X86_64)
  90 static const uint64_t mmx_null  __attribute__((aligned(8))) = 0x0000000000000000ULL;
  91 static const uint64_t mmx_one   __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL;
  92 static const uint64_t mask32b  attribute_used __attribute__((aligned(8))) = 0x000000FF000000FFULL;
  93 static const uint64_t mask32g  attribute_used __attribute__((aligned(8))) = 0x0000FF000000FF00ULL;
  94 static const uint64_t mask32r  attribute_used __attribute__((aligned(8))) = 0x00FF000000FF0000ULL;
  95 static const uint64_t mask32   __attribute__((aligned(8))) = 0x00FFFFFF00FFFFFFULL;
  96 static const uint64_t mask3216br __attribute__((aligned(8)))=0x00F800F800F800F8ULL;
  97 static const uint64_t mask3216g  __attribute__((aligned(8)))=0x0000FC000000FC00ULL;
  98 static const uint64_t mask3215g  __attribute__((aligned(8)))=0x0000F8000000F800ULL;
  99 static const uint64_t mul3216  __attribute__((aligned(8))) = 0x2000000420000004ULL;
 100 static const uint64_t mul3215  __attribute__((aligned(8))) = 0x2000000820000008ULL;
 101 static const uint64_t mask24b  attribute_used __attribute__((aligned(8))) = 0x00FF0000FF0000FFULL;
 102 static const uint64_t mask24g  attribute_used __attribute__((aligned(8))) = 0xFF0000FF0000FF00ULL;
 103 static const uint64_t mask24r  attribute_used __attribute__((aligned(8))) = 0x0000FF0000FF0000ULL;
 104 static const uint64_t mask24l  __attribute__((aligned(8))) = 0x0000000000FFFFFFULL;
 105 static const uint64_t mask24h  __attribute__((aligned(8))) = 0x0000FFFFFF000000ULL;
 106 static const uint64_t mask24hh  __attribute__((aligned(8))) = 0xffff000000000000ULL;
 107 static const uint64_t mask24hhh  __attribute__((aligned(8))) = 0xffffffff00000000ULL;
 108 static const uint64_t mask24hhhh  __attribute__((aligned(8))) = 0xffffffffffff0000ULL;
 109 static const uint64_t mask15b  __attribute__((aligned(8))) = 0x001F001F001F001FULL; /* 00000000 00011111  xxB */
 110 static const uint64_t mask15rg __attribute__((aligned(8))) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000  RGx */
 111 static const uint64_t mask15s  __attribute__((aligned(8))) = 0xFFE0FFE0FFE0FFE0ULL;
 112 static const uint64_t mask15g  __attribute__((aligned(8))) = 0x03E003E003E003E0ULL;
 113 static const uint64_t mask15r  __attribute__((aligned(8))) = 0x7C007C007C007C00ULL;
 114 #define mask16b mask15b
 115 static const uint64_t mask16g  __attribute__((aligned(8))) = 0x07E007E007E007E0ULL;
 116 static const uint64_t mask16r  __attribute__((aligned(8))) = 0xF800F800F800F800ULL;
 117 static const uint64_t red_16mask  __attribute__((aligned(8))) = 0x0000f8000000f800ULL;
 118 static const uint64_t green_16mask __attribute__((aligned(8)))= 0x000007e0000007e0ULL;
 119 static const uint64_t blue_16mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
 120 static const uint64_t red_15mask  __attribute__((aligned(8))) = 0x00007c000000f800ULL;
 121 static const uint64_t green_15mask __attribute__((aligned(8)))= 0x000003e0000007e0ULL;
 122 static const uint64_t blue_15mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
 123
 124 #ifdef FAST_BGR2YV12
 125 static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;
 126 static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
 127 static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
 128 #else
 129 static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;
 130 static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
 131 static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
 132 #endif
 133 static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;
 134 static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL;
 135 static const uint64_t w1111       attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;
 136
 137 #if 0
 138 static volatile uint64_t __attribute__((aligned(8))) b5Dither;
 139 static volatile uint64_t __attribute__((aligned(8))) g5Dither;
 140 static volatile uint64_t __attribute__((aligned(8))) g6Dither;
 141 static volatile uint64_t __attribute__((aligned(8))) r5Dither;
 142
 143 static uint64_t __attribute__((aligned(8))) dither4[2]={
 144         0x0103010301030103LL,
 145         0x0200020002000200LL,};
 146
 147 static uint64_t __attribute__((aligned(8))) dither8[2]={
 148         0x0602060206020602LL,
 149         0x0004000400040004LL,};
 150 #endif
 151 #endif /* defined(ARCH_X86) || defined(ARCH_X86_64) */
 152
 153 #define RGB2YUV_SHIFT 8
 154 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
 155 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
 156 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 157 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
 158 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
 159 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
 160 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
 161 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 162 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
 163
 164 //Note: we have C, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
 165 //Plain C versions
 166 #undef HAVE_MMX
 167 #undef HAVE_MMX2
 168 #undef HAVE_3DNOW
 169 #undef HAVE_SSE2
 170 #define RENAME(a) a ## _C
 171 #include "rgb2rgb_template.c"
 172
 173 #if defined(ARCH_X86) || defined(ARCH_X86_64)
 174
 175 //MMX versions
 176 #undef RENAME
 177 #define HAVE_MMX
 178 #undef HAVE_MMX2
 179 #undef HAVE_3DNOW
 180 #undef HAVE_SSE2
 181 #define RENAME(a) a ## _MMX
 182 #include "rgb2rgb_template.c"
 183
 184 //MMX2 versions
 185 #undef RENAME
 186 #define HAVE_MMX
 187 #define HAVE_MMX2
 188 #undef HAVE_3DNOW
 189 #undef HAVE_SSE2
 190 #define RENAME(a) a ## _MMX2
 191 #include "rgb2rgb_template.c"
 192
 193 //3DNOW versions
 194 #undef RENAME
 195 #define HAVE_MMX
 196 #undef HAVE_MMX2
 197 #define HAVE_3DNOW
 198 #undef HAVE_SSE2
 199 #define RENAME(a) a ## _3DNOW
 200 #include "rgb2rgb_template.c"
 201
 202 #endif //ARCH_X86 || ARCH_X86_64
 203
 204 /*
 205  rgb15->rgb16 Original by Strepto/Astral
 206  ported to gcc & bugfixed : A'rpi
 207  MMX2, 3DNOW optimization by Nick Kurshev
 208  32bit c version, and and&add trick by Michael Niedermayer
 209 */
 210
 211 void sws_rgb2rgb_init(int flags){
 212 #if defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX)
 213         if(flags & SWS_CPU_CAPS_MMX2){
 214                 rgb15to16= rgb15to16_MMX2;
 215                 rgb15to24= rgb15to24_MMX2;
 216                 rgb15to32= rgb15to32_MMX2;
 217                 rgb16to24= rgb16to24_MMX2;
 218                 rgb16to32= rgb16to32_MMX2;
 219                 rgb16to15= rgb16to15_MMX2;
 220                 rgb24to16= rgb24to16_MMX2;
 221                 rgb24to15= rgb24to15_MMX2;
 222                 rgb24to32= rgb24to32_MMX2;
 223                 rgb32to16= rgb32to16_MMX2;
 224                 rgb32to15= rgb32to15_MMX2;
 225                 rgb32to24= rgb32to24_MMX2;
 226                 rgb24tobgr15= rgb24tobgr15_MMX2;
 227                 rgb24tobgr16= rgb24tobgr16_MMX2;
 228                 rgb24tobgr24= rgb24tobgr24_MMX2;
 229                 rgb32tobgr32= rgb32tobgr32_MMX2;
 230                 rgb32tobgr16= rgb32tobgr16_MMX2;
 231                 rgb32tobgr15= rgb32tobgr15_MMX2;
 232                 yv12toyuy2= yv12toyuy2_MMX2;
 233                 yv12touyvy= yv12touyvy_MMX2;
 234                 yuv422ptoyuy2= yuv422ptoyuy2_MMX2;
 235                 yuy2toyv12= yuy2toyv12_MMX2;
 236 //              uyvytoyv12= uyvytoyv12_MMX2;
 237 //              yvu9toyv12= yvu9toyv12_MMX2;
 238                 planar2x= planar2x_MMX2;
 239                 rgb24toyv12= rgb24toyv12_MMX2;
 240                 interleaveBytes= interleaveBytes_MMX2;
 241                 vu9_to_vu12= vu9_to_vu12_MMX2;
 242                 yvu9_to_yuy2= yvu9_to_yuy2_MMX2;
 243         }else if(flags & SWS_CPU_CAPS_3DNOW){
 244                 rgb15to16= rgb15to16_3DNOW;
 245                 rgb15to24= rgb15to24_3DNOW;
 246                 rgb15to32= rgb15to32_3DNOW;
 247                 rgb16to24= rgb16to24_3DNOW;
 248                 rgb16to32= rgb16to32_3DNOW;
 249                 rgb16to15= rgb16to15_3DNOW;
 250                 rgb24to16= rgb24to16_3DNOW;
 251                 rgb24to15= rgb24to15_3DNOW;
 252                 rgb24to32= rgb24to32_3DNOW;
 253                 rgb32to16= rgb32to16_3DNOW;
 254                 rgb32to15= rgb32to15_3DNOW;
 255                 rgb32to24= rgb32to24_3DNOW;
 256                 rgb24tobgr15= rgb24tobgr15_3DNOW;
 257                 rgb24tobgr16= rgb24tobgr16_3DNOW;
 258                 rgb24tobgr24= rgb24tobgr24_3DNOW;
 259                 rgb32tobgr32= rgb32tobgr32_3DNOW;
 260                 rgb32tobgr16= rgb32tobgr16_3DNOW;
 261                 rgb32tobgr15= rgb32tobgr15_3DNOW;
 262                 yv12toyuy2= yv12toyuy2_3DNOW;
 263                 yv12touyvy= yv12touyvy_3DNOW;
 264                 yuv422ptoyuy2= yuv422ptoyuy2_3DNOW;
 265                 yuy2toyv12= yuy2toyv12_3DNOW;
 266 //              uyvytoyv12= uyvytoyv12_3DNOW;
 267 //              yvu9toyv12= yvu9toyv12_3DNOW;
 268                 planar2x= planar2x_3DNOW;
 269                 rgb24toyv12= rgb24toyv12_3DNOW;
 270                 interleaveBytes= interleaveBytes_3DNOW;
 271                 vu9_to_vu12= vu9_to_vu12_3DNOW;
 272                 yvu9_to_yuy2= yvu9_to_yuy2_3DNOW;
 273         }else if(flags & SWS_CPU_CAPS_MMX){
 274                 rgb15to16= rgb15to16_MMX;
 275                 rgb15to24= rgb15to24_MMX;
 276                 rgb15to32= rgb15to32_MMX;
 277                 rgb16to24= rgb16to24_MMX;
 278                 rgb16to32= rgb16to32_MMX;
 279                 rgb16to15= rgb16to15_MMX;
 280                 rgb24to16= rgb24to16_MMX;
 281                 rgb24to15= rgb24to15_MMX;
 282                 rgb24to32= rgb24to32_MMX;
 283                 rgb32to16= rgb32to16_MMX;
 284                 rgb32to15= rgb32to15_MMX;
 285                 rgb32to24= rgb32to24_MMX;
 286                 rgb24tobgr15= rgb24tobgr15_MMX;
 287                 rgb24tobgr16= rgb24tobgr16_MMX;
 288                 rgb24tobgr24= rgb24tobgr24_MMX;
 289                 rgb32tobgr32= rgb32tobgr32_MMX;
 290                 rgb32tobgr16= rgb32tobgr16_MMX;
 291                 rgb32tobgr15= rgb32tobgr15_MMX;
 292                 yv12toyuy2= yv12toyuy2_MMX;
 293                 yv12touyvy= yv12touyvy_MMX;
 294                 yuv422ptoyuy2= yuv422ptoyuy2_MMX;
 295                 yuy2toyv12= yuy2toyv12_MMX;
 296 //              uyvytoyv12= uyvytoyv12_MMX;
 297 //              yvu9toyv12= yvu9toyv12_MMX;
 298                 planar2x= planar2x_MMX;
 299                 rgb24toyv12= rgb24toyv12_MMX;
 300                 interleaveBytes= interleaveBytes_MMX;
 301                 vu9_to_vu12= vu9_to_vu12_MMX;
 302                 yvu9_to_yuy2= yvu9_to_yuy2_MMX;
 303         }else
 304 #endif /* defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX) */
 305         {
 306                 rgb15to16= rgb15to16_C;
 307                 rgb15to24= rgb15to24_C;
 308                 rgb15to32= rgb15to32_C;
 309                 rgb16to24= rgb16to24_C;
 310                 rgb16to32= rgb16to32_C;
 311                 rgb16to15= rgb16to15_C;
 312                 rgb24to16= rgb24to16_C;
 313                 rgb24to15= rgb24to15_C;
 314                 rgb24to32= rgb24to32_C;
 315                 rgb32to16= rgb32to16_C;
 316                 rgb32to15= rgb32to15_C;
 317                 rgb32to24= rgb32to24_C;
 318                 rgb24tobgr15= rgb24tobgr15_C;
 319                 rgb24tobgr16= rgb24tobgr16_C;
 320                 rgb24tobgr24= rgb24tobgr24_C;
 321                 rgb32tobgr32= rgb32tobgr32_C;
 322                 rgb32tobgr16= rgb32tobgr16_C;
 323                 rgb32tobgr15= rgb32tobgr15_C;
 324                 yv12toyuy2= yv12toyuy2_C;
 325                 yv12touyvy= yv12touyvy_C;
 326                 yuv422ptoyuy2= yuv422ptoyuy2_C;
 327                 yuy2toyv12= yuy2toyv12_C;
 328 //              uyvytoyv12= uyvytoyv12_C;
 329 //              yvu9toyv12= yvu9toyv12_C;
 330                 planar2x= planar2x_C;
 331                 rgb24toyv12= rgb24toyv12_C;
 332                 interleaveBytes= interleaveBytes_C;
 333                 vu9_to_vu12= vu9_to_vu12_C;
 334                 yvu9_to_yuy2= yvu9_to_yuy2_C;
 335         }
 336 }
 337
 338 /**
 339  * Pallete is assumed to contain bgr32
 340  */
 341 void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 342 {
 343         long i;
 344
 345 /*
 346         for(i=0; i<num_pixels; i++)
 347                 ((unsigned *)dst)[i] = ((unsigned *)palette)[ src[i] ];
 348 */
 349
 350         for(i=0; i<num_pixels; i++)
 351         {
 352                 #ifdef WORDS_BIGENDIAN
 353                         dst[3]= palette[ src[i]*4+2 ];
 354                         dst[2]= palette[ src[i]*4+1 ];
 355                         dst[1]= palette[ src[i]*4+0 ];
 356                 #else
 357                 //FIXME slow?
 358                         dst[0]= palette[ src[i]*4+2 ];
 359                         dst[1]= palette[ src[i]*4+1 ];
 360                         dst[2]= palette[ src[i]*4+0 ];
 361                         //dst[3]= 0; /* do we need this cleansing? */
 362                 #endif
 363                 dst+= 4;
 364         }
 365 }
 366
 367 void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 368 {
 369         long i;
 370         for(i=0; i<num_pixels; i++)
 371         {
 372                 #ifdef WORDS_BIGENDIAN
 373                         dst[3]= palette[ src[i]*4+0 ];
 374                         dst[2]= palette[ src[i]*4+1 ];
 375                         dst[1]= palette[ src[i]*4+2 ];
 376                 #else
 377                         //FIXME slow?
 378                         dst[0]= palette[ src[i]*4+0 ];
 379                         dst[1]= palette[ src[i]*4+1 ];
 380                         dst[2]= palette[ src[i]*4+2 ];
 381                         //dst[3]= 0; /* do we need this cleansing? */
 382                 #endif
 383
 384                 dst+= 4;
 385         }
 386 }
 387
 388 /**
 389  * Pallete is assumed to contain bgr32
 390  */
 391 void palette8torgb24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 392 {
 393         long i;
 394 /*
 395         writes 1 byte o much and might cause alignment issues on some architectures?
 396         for(i=0; i<num_pixels; i++)
 397                 ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
 398 */
 399         for(i=0; i<num_pixels; i++)
 400         {
 401                 //FIXME slow?
 402                 dst[0]= palette[ src[i]*4+2 ];
 403                 dst[1]= palette[ src[i]*4+1 ];
 404                 dst[2]= palette[ src[i]*4+0 ];
 405                 dst+= 3;
 406         }
 407 }
 408
 409 void palette8tobgr24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 410 {
 411         long i;
 412 /*
 413         writes 1 byte o much and might cause alignment issues on some architectures?
 414         for(i=0; i<num_pixels; i++)
 415                 ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
 416 */
 417         for(i=0; i<num_pixels; i++)
 418         {
 419                 //FIXME slow?
 420                 dst[0]= palette[ src[i]*4+0 ];
 421                 dst[1]= palette[ src[i]*4+1 ];
 422                 dst[2]= palette[ src[i]*4+2 ];
 423                 dst+= 3;
 424         }
 425 }
 426
 427 /**
 428  * Palette is assumed to contain bgr16, see rgb32to16 to convert the palette
 429  */
 430 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 431 {
 432         long i;
 433         for(i=0; i<num_pixels; i++)
 434                 ((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
 435 }
 436 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 437 {
 438         long i;
 439         for(i=0; i<num_pixels; i++)
 440                 ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
 441 }
 442
 443 /**
 444  * Pallete is assumed to contain bgr15, see rgb32to15 to convert the palette
 445  */
 446 void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 447 {
 448         long i;
 449         for(i=0; i<num_pixels; i++)
 450                 ((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
 451 }
 452 void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 453 {
 454         long i;
 455         for(i=0; i<num_pixels; i++)
 456                 ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
 457 }
 458
 459 void rgb32tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 460 {
 461         long i;
 462         long num_pixels = src_size >> 2;
 463         for(i=0; i<num_pixels; i++)
 464         {
 465                 #ifdef WORDS_BIGENDIAN
 466                         /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
 467                         dst[3*i + 0] = src[4*i + 1];
 468                         dst[3*i + 1] = src[4*i + 2];
 469                         dst[3*i + 2] = src[4*i + 3];
 470                 #else
 471                         dst[3*i + 0] = src[4*i + 2];
 472                         dst[3*i + 1] = src[4*i + 1];
 473                         dst[3*i + 2] = src[4*i + 0];
 474                 #endif
 475         }
 476 }
 477
 478 void rgb24tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 479 {
 480         long i;
 481         for(i=0; 3*i<src_size; i++)
 482         {
 483                 #ifdef WORDS_BIGENDIAN
 484                         /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
 485                         dst[4*i + 0] = 0;
 486                         dst[4*i + 1] = src[3*i + 0];
 487                         dst[4*i + 2] = src[3*i + 1];
 488                         dst[4*i + 3] = src[3*i + 2];
 489                 #else
 490                         dst[4*i + 0] = src[3*i + 2];
 491                         dst[4*i + 1] = src[3*i + 1];
 492                         dst[4*i + 2] = src[3*i + 0];
 493                         dst[4*i + 3] = 0;
 494                 #endif
 495         }
 496 }
 497
 498 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 499 {
 500         const uint16_t *end;
 501         uint8_t *d = (uint8_t *)dst;
 502         const uint16_t *s = (uint16_t *)src;
 503         end = s + src_size/2;
 504         while(s < end)
 505         {
 506                 register uint16_t bgr;
 507                 bgr = *s++;
 508                 #ifdef WORDS_BIGENDIAN
 509                         *d++ = 0;
 510                         *d++ = (bgr&0x1F)<<3;
 511                         *d++ = (bgr&0x7E0)>>3;
 512                         *d++ = (bgr&0xF800)>>8;
 513                 #else
 514                         *d++ = (bgr&0xF800)>>8;
 515                         *d++ = (bgr&0x7E0)>>3;
 516                         *d++ = (bgr&0x1F)<<3;
 517                         *d++ = 0;
 518                 #endif
 519         }
 520 }
 521
 522 void rgb16tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 523 {
 524         const uint16_t *end;
 525         uint8_t *d = (uint8_t *)dst;
 526         const uint16_t *s = (const uint16_t *)src;
 527         end = s + src_size/2;
 528         while(s < end)
 529         {
 530                 register uint16_t bgr;
 531                 bgr = *s++;
 532                 *d++ = (bgr&0xF800)>>8;
 533                 *d++ = (bgr&0x7E0)>>3;
 534                 *d++ = (bgr&0x1F)<<3;
 535         }
 536 }
 537
 538 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 539 {
 540         long i;
 541         long num_pixels = src_size >> 1;
 542
 543         for(i=0; i<num_pixels; i++)
 544         {
 545             unsigned b,g,r;
 546             register uint16_t rgb;
 547             rgb = src[2*i];
 548             r = rgb&0x1F;
 549             g = (rgb&0x7E0)>>5;
 550             b = (rgb&0xF800)>>11;
 551             dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
 552         }
 553 }
 554
 555 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 556 {
 557         long i;
 558         long num_pixels = src_size >> 1;
 559
 560         for(i=0; i<num_pixels; i++)
 561         {
 562             unsigned b,g,r;
 563             register uint16_t rgb;
 564             rgb = src[2*i];
 565             r = rgb&0x1F;
 566             g = (rgb&0x7E0)>>5;
 567             b = (rgb&0xF800)>>11;
 568             dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10);
 569         }
 570 }
 571
 572 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 573 {
 574         const uint16_t *end;
 575         uint8_t *d = (uint8_t *)dst;
 576         const uint16_t *s = (const uint16_t *)src;
 577         end = s + src_size/2;
 578         while(s < end)
 579         {
 580                 register uint16_t bgr;
 581                 bgr = *s++;
 582                 #ifdef WORDS_BIGENDIAN
 583                         *d++ = 0;
 584                         *d++ = (bgr&0x1F)<<3;
 585                         *d++ = (bgr&0x3E0)>>2;
 586                         *d++ = (bgr&0x7C00)>>7;
 587                 #else
 588                         *d++ = (bgr&0x7C00)>>7;
 589                         *d++ = (bgr&0x3E0)>>2;
 590                         *d++ = (bgr&0x1F)<<3;
 591                         *d++ = 0;
 592                 #endif
 593         }
 594 }
 595
 596 void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 597 {
 598         const uint16_t *end;
 599         uint8_t *d = (uint8_t *)dst;
 600         const uint16_t *s = (uint16_t *)src;
 601         end = s + src_size/2;
 602         while(s < end)
 603         {
 604                 register uint16_t bgr;
 605                 bgr = *s++;
 606                 *d++ = (bgr&0x7C00)>>7;
 607                 *d++ = (bgr&0x3E0)>>2;
 608                 *d++ = (bgr&0x1F)<<3;
 609         }
 610 }
 611
 612 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 613 {
 614         long i;
 615         long num_pixels = src_size >> 1;
 616
 617         for(i=0; i<num_pixels; i++)
 618         {
 619             unsigned b,g,r;
 620             register uint16_t rgb;
 621             rgb = src[2*i];
 622             r = rgb&0x1F;
 623             g = (rgb&0x3E0)>>5;
 624             b = (rgb&0x7C00)>>10;
 625             dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
 626         }
 627 }
 628
 629 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 630 {
 631         long i;
 632         long num_pixels = src_size >> 1;
 633
 634         for(i=0; i<num_pixels; i++)
 635         {
 636             unsigned b,g,r;
 637             register uint16_t rgb;
 638             rgb = src[2*i];
 639             r = rgb&0x1F;
 640             g = (rgb&0x3E0)>>5;
 641             b = (rgb&0x7C00)>>10;
 642             dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10);
 643         }
 644 }
 645
 646 void rgb8tobgr8(const uint8_t *src, uint8_t *dst, long src_size)
 647 {
 648         long i;
 649         long num_pixels = src_size;
 650         for(i=0; i<num_pixels; i++)
 651         {
 652             unsigned b,g,r;
 653             register uint8_t rgb;
 654             rgb = src[i];
 655             r = (rgb&0x07);
 656             g = (rgb&0x38)>>3;
 657             b = (rgb&0xC0)>>6;
 658             dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
 659         }
 660 }