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) (under GPL)
   9  */
  10 #include <inttypes.h>
  11 #include "config.h"
  12 #include "rgb2rgb.h"
  13 #include "swscale.h"
  14 #include "swscale_internal.h"
  15 #include "x86_cpu.h"
  16 #include "bswap.h"
  17 #ifdef USE_FASTMEMCPY
  18 #include "libvo/fastmemcpy.h"
  19 #endif
  20
  21 #define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit
  22
  23 void (*rgb24to32)(const uint8_t *src,uint8_t *dst,long src_size);
  24 void (*rgb24to16)(const uint8_t *src,uint8_t *dst,long src_size);
  25 void (*rgb24to15)(const uint8_t *src,uint8_t *dst,long src_size);
  26 void (*rgb32to24)(const uint8_t *src,uint8_t *dst,long src_size);
  27 void (*rgb32to16)(const uint8_t *src,uint8_t *dst,long src_size);
  28 void (*rgb32to15)(const uint8_t *src,uint8_t *dst,long src_size);
  29 void (*rgb15to16)(const uint8_t *src,uint8_t *dst,long src_size);
  30 void (*rgb15to24)(const uint8_t *src,uint8_t *dst,long src_size);
  31 void (*rgb15to32)(const uint8_t *src,uint8_t *dst,long src_size);
  32 void (*rgb16to15)(const uint8_t *src,uint8_t *dst,long src_size);
  33 void (*rgb16to24)(const uint8_t *src,uint8_t *dst,long src_size);
  34 void (*rgb16to32)(const uint8_t *src,uint8_t *dst,long src_size);
  35 //void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  36 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  37 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  38 void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  39 void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  40 //void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  41 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  42 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  43
  44 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  45         long width, long height,
  46         long lumStride, long chromStride, long dstStride);
  47 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  48         long width, long height,
  49         long lumStride, long chromStride, long dstStride);
  50 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  51         long width, long height,
  52         long lumStride, long chromStride, long dstStride);
  53 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  54         long width, long height,
  55         long lumStride, long chromStride, long srcStride);
  56 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  57         long width, long height,
  58         long lumStride, long chromStride, long srcStride);
  59 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
  60         long srcStride, long dstStride);
  61 void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
  62                             long width, long height, long src1Stride,
  63                             long src2Stride, long dstStride);
  64 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
  65                         uint8_t *dst1, uint8_t *dst2,
  66                         long width, long height,
  67                         long srcStride1, long srcStride2,
  68                         long dstStride1, long dstStride2);
  69 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
  70                         uint8_t *dst,
  71                         long width, long height,
  72                         long srcStride1, long srcStride2,
  73                         long srcStride3, long dstStride);
  74
  75 #if defined(HAVE_MMX)
  76 static const uint64_t mmx_null  __attribute__((aligned(8))) = 0x0000000000000000ULL;
  77 static const uint64_t mmx_one   __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL;
  78 #endif
  79 #if defined(ARCH_X86) || defined(ARCH_X86_64)
  80 static const uint64_t mask32b  attribute_used __attribute__((aligned(8))) = 0x000000FF000000FFULL;
  81 static const uint64_t mask32g  attribute_used __attribute__((aligned(8))) = 0x0000FF000000FF00ULL;
  82 static const uint64_t mask32r  attribute_used __attribute__((aligned(8))) = 0x00FF000000FF0000ULL;
  83 static const uint64_t mask32   __attribute__((aligned(8))) = 0x00FFFFFF00FFFFFFULL;
  84 static const uint64_t mask3216br __attribute__((aligned(8)))=0x00F800F800F800F8ULL;
  85 static const uint64_t mask3216g  __attribute__((aligned(8)))=0x0000FC000000FC00ULL;
  86 static const uint64_t mask3215g  __attribute__((aligned(8)))=0x0000F8000000F800ULL;
  87 static const uint64_t mul3216  __attribute__((aligned(8))) = 0x2000000420000004ULL;
  88 static const uint64_t mul3215  __attribute__((aligned(8))) = 0x2000000820000008ULL;
  89 static const uint64_t mask24b  attribute_used __attribute__((aligned(8))) = 0x00FF0000FF0000FFULL;
  90 static const uint64_t mask24g  attribute_used __attribute__((aligned(8))) = 0xFF0000FF0000FF00ULL;
  91 static const uint64_t mask24r  attribute_used __attribute__((aligned(8))) = 0x0000FF0000FF0000ULL;
  92 static const uint64_t mask24l  __attribute__((aligned(8))) = 0x0000000000FFFFFFULL;
  93 static const uint64_t mask24h  __attribute__((aligned(8))) = 0x0000FFFFFF000000ULL;
  94 static const uint64_t mask24hh  __attribute__((aligned(8))) = 0xffff000000000000ULL;
  95 static const uint64_t mask24hhh  __attribute__((aligned(8))) = 0xffffffff00000000ULL;
  96 static const uint64_t mask24hhhh  __attribute__((aligned(8))) = 0xffffffffffff0000ULL;
  97 static const uint64_t mask15b  __attribute__((aligned(8))) = 0x001F001F001F001FULL; /* 00000000 00011111  xxB */
  98 static const uint64_t mask15rg __attribute__((aligned(8))) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000  RGx */
  99 static const uint64_t mask15s  __attribute__((aligned(8))) = 0xFFE0FFE0FFE0FFE0ULL;
 100 static const uint64_t mask15g  __attribute__((aligned(8))) = 0x03E003E003E003E0ULL;
 101 static const uint64_t mask15r  __attribute__((aligned(8))) = 0x7C007C007C007C00ULL;
 102 #define mask16b mask15b
 103 static const uint64_t mask16g  __attribute__((aligned(8))) = 0x07E007E007E007E0ULL;
 104 static const uint64_t mask16r  __attribute__((aligned(8))) = 0xF800F800F800F800ULL;
 105 static const uint64_t red_16mask  __attribute__((aligned(8))) = 0x0000f8000000f800ULL;
 106 static const uint64_t green_16mask __attribute__((aligned(8)))= 0x000007e0000007e0ULL;
 107 static const uint64_t blue_16mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
 108 static const uint64_t red_15mask  __attribute__((aligned(8))) = 0x00007c000000f800ULL;
 109 static const uint64_t green_15mask __attribute__((aligned(8)))= 0x000003e0000007e0ULL;
 110 static const uint64_t blue_15mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
 111
 112 #ifdef FAST_BGR2YV12
 113 static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;
 114 static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
 115 static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
 116 #else
 117 static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;
 118 static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
 119 static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
 120 #endif
 121 static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;
 122 static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL;
 123 static const uint64_t w1111       attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;
 124
 125 #if 0
 126 static volatile uint64_t __attribute__((aligned(8))) b5Dither;
 127 static volatile uint64_t __attribute__((aligned(8))) g5Dither;
 128 static volatile uint64_t __attribute__((aligned(8))) g6Dither;
 129 static volatile uint64_t __attribute__((aligned(8))) r5Dither;
 130
 131 static uint64_t __attribute__((aligned(8))) dither4[2]={
 132         0x0103010301030103LL,
 133         0x0200020002000200LL,};
 134
 135 static uint64_t __attribute__((aligned(8))) dither8[2]={
 136         0x0602060206020602LL,
 137         0x0004000400040004LL,};
 138 #endif
 139 #endif /* defined(ARCH_X86) || defined(ARCH_X86_64) */
 140
 141 #define RGB2YUV_SHIFT 8
 142 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
 143 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
 144 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 145 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
 146 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
 147 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
 148 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
 149 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 150 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
 151
 152 //Note: we have C, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
 153 //Plain C versions
 154 #undef HAVE_MMX
 155 #undef HAVE_MMX2
 156 #undef HAVE_3DNOW
 157 #undef HAVE_SSE2
 158 #define RENAME(a) a ## _C
 159 #include "rgb2rgb_template.c"
 160
 161 #if defined(ARCH_X86) || defined(ARCH_X86_64)
 162
 163 //MMX versions
 164 #undef RENAME
 165 #define HAVE_MMX
 166 #undef HAVE_MMX2
 167 #undef HAVE_3DNOW
 168 #undef HAVE_SSE2
 169 #define RENAME(a) a ## _MMX
 170 #include "rgb2rgb_template.c"
 171
 172 //MMX2 versions
 173 #undef RENAME
 174 #define HAVE_MMX
 175 #define HAVE_MMX2
 176 #undef HAVE_3DNOW
 177 #undef HAVE_SSE2
 178 #define RENAME(a) a ## _MMX2
 179 #include "rgb2rgb_template.c"
 180
 181 //3DNOW versions
 182 #undef RENAME
 183 #define HAVE_MMX
 184 #undef HAVE_MMX2
 185 #define HAVE_3DNOW
 186 #undef HAVE_SSE2
 187 #define RENAME(a) a ## _3DNOW
 188 #include "rgb2rgb_template.c"
 189
 190 #endif //ARCH_X86 || ARCH_X86_64
 191
 192 /*
 193  rgb15->rgb16 Original by Strepto/Astral
 194  ported to gcc & bugfixed : A'rpi
 195  MMX2, 3DNOW optimization by Nick Kurshev
 196  32bit c version, and and&add trick by Michael Niedermayer
 197 */
 198
 199 void sws_rgb2rgb_init(int flags){
 200 #if defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX)
 201         if(flags & SWS_CPU_CAPS_MMX2){
 202                 rgb15to16= rgb15to16_MMX2;
 203                 rgb15to24= rgb15to24_MMX2;
 204                 rgb15to32= rgb15to32_MMX2;
 205                 rgb16to24= rgb16to24_MMX2;
 206                 rgb16to32= rgb16to32_MMX2;
 207                 rgb16to15= rgb16to15_MMX2;
 208                 rgb24to16= rgb24to16_MMX2;
 209                 rgb24to15= rgb24to15_MMX2;
 210                 rgb24to32= rgb24to32_MMX2;
 211                 rgb32to16= rgb32to16_MMX2;
 212                 rgb32to15= rgb32to15_MMX2;
 213                 rgb32to24= rgb32to24_MMX2;
 214                 rgb24tobgr15= rgb24tobgr15_MMX2;
 215                 rgb24tobgr16= rgb24tobgr16_MMX2;
 216                 rgb24tobgr24= rgb24tobgr24_MMX2;
 217                 rgb32tobgr32= rgb32tobgr32_MMX2;
 218                 rgb32tobgr16= rgb32tobgr16_MMX2;
 219                 rgb32tobgr15= rgb32tobgr15_MMX2;
 220                 yv12toyuy2= yv12toyuy2_MMX2;
 221                 yv12touyvy= yv12touyvy_MMX2;
 222                 yuv422ptoyuy2= yuv422ptoyuy2_MMX2;
 223                 yuy2toyv12= yuy2toyv12_MMX2;
 224 //              uyvytoyv12= uyvytoyv12_MMX2;
 225 //              yvu9toyv12= yvu9toyv12_MMX2;
 226                 planar2x= planar2x_MMX2;
 227                 rgb24toyv12= rgb24toyv12_MMX2;
 228                 interleaveBytes= interleaveBytes_MMX2;
 229                 vu9_to_vu12= vu9_to_vu12_MMX2;
 230                 yvu9_to_yuy2= yvu9_to_yuy2_MMX2;
 231         }else if(flags & SWS_CPU_CAPS_3DNOW){
 232                 rgb15to16= rgb15to16_3DNOW;
 233                 rgb15to24= rgb15to24_3DNOW;
 234                 rgb15to32= rgb15to32_3DNOW;
 235                 rgb16to24= rgb16to24_3DNOW;
 236                 rgb16to32= rgb16to32_3DNOW;
 237                 rgb16to15= rgb16to15_3DNOW;
 238                 rgb24to16= rgb24to16_3DNOW;
 239                 rgb24to15= rgb24to15_3DNOW;
 240                 rgb24to32= rgb24to32_3DNOW;
 241                 rgb32to16= rgb32to16_3DNOW;
 242                 rgb32to15= rgb32to15_3DNOW;
 243                 rgb32to24= rgb32to24_3DNOW;
 244                 rgb24tobgr15= rgb24tobgr15_3DNOW;
 245                 rgb24tobgr16= rgb24tobgr16_3DNOW;
 246                 rgb24tobgr24= rgb24tobgr24_3DNOW;
 247                 rgb32tobgr32= rgb32tobgr32_3DNOW;
 248                 rgb32tobgr16= rgb32tobgr16_3DNOW;
 249                 rgb32tobgr15= rgb32tobgr15_3DNOW;
 250                 yv12toyuy2= yv12toyuy2_3DNOW;
 251                 yv12touyvy= yv12touyvy_3DNOW;
 252                 yuv422ptoyuy2= yuv422ptoyuy2_3DNOW;
 253                 yuy2toyv12= yuy2toyv12_3DNOW;
 254 //              uyvytoyv12= uyvytoyv12_3DNOW;
 255 //              yvu9toyv12= yvu9toyv12_3DNOW;
 256                 planar2x= planar2x_3DNOW;
 257                 rgb24toyv12= rgb24toyv12_3DNOW;
 258                 interleaveBytes= interleaveBytes_3DNOW;
 259                 vu9_to_vu12= vu9_to_vu12_3DNOW;
 260                 yvu9_to_yuy2= yvu9_to_yuy2_3DNOW;
 261         }else if(flags & SWS_CPU_CAPS_MMX){
 262                 rgb15to16= rgb15to16_MMX;
 263                 rgb15to24= rgb15to24_MMX;
 264                 rgb15to32= rgb15to32_MMX;
 265                 rgb16to24= rgb16to24_MMX;
 266                 rgb16to32= rgb16to32_MMX;
 267                 rgb16to15= rgb16to15_MMX;
 268                 rgb24to16= rgb24to16_MMX;
 269                 rgb24to15= rgb24to15_MMX;
 270                 rgb24to32= rgb24to32_MMX;
 271                 rgb32to16= rgb32to16_MMX;
 272                 rgb32to15= rgb32to15_MMX;
 273                 rgb32to24= rgb32to24_MMX;
 274                 rgb24tobgr15= rgb24tobgr15_MMX;
 275                 rgb24tobgr16= rgb24tobgr16_MMX;
 276                 rgb24tobgr24= rgb24tobgr24_MMX;
 277                 rgb32tobgr32= rgb32tobgr32_MMX;
 278                 rgb32tobgr16= rgb32tobgr16_MMX;
 279                 rgb32tobgr15= rgb32tobgr15_MMX;
 280                 yv12toyuy2= yv12toyuy2_MMX;
 281                 yv12touyvy= yv12touyvy_MMX;
 282                 yuv422ptoyuy2= yuv422ptoyuy2_MMX;
 283                 yuy2toyv12= yuy2toyv12_MMX;
 284 //              uyvytoyv12= uyvytoyv12_MMX;
 285 //              yvu9toyv12= yvu9toyv12_MMX;
 286                 planar2x= planar2x_MMX;
 287                 rgb24toyv12= rgb24toyv12_MMX;
 288                 interleaveBytes= interleaveBytes_MMX;
 289                 vu9_to_vu12= vu9_to_vu12_MMX;
 290                 yvu9_to_yuy2= yvu9_to_yuy2_MMX;
 291         }else
 292 #endif /* defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX) */
 293         {
 294                 rgb15to16= rgb15to16_C;
 295                 rgb15to24= rgb15to24_C;
 296                 rgb15to32= rgb15to32_C;
 297                 rgb16to24= rgb16to24_C;
 298                 rgb16to32= rgb16to32_C;
 299                 rgb16to15= rgb16to15_C;
 300                 rgb24to16= rgb24to16_C;
 301                 rgb24to15= rgb24to15_C;
 302                 rgb24to32= rgb24to32_C;
 303                 rgb32to16= rgb32to16_C;
 304                 rgb32to15= rgb32to15_C;
 305                 rgb32to24= rgb32to24_C;
 306                 rgb24tobgr15= rgb24tobgr15_C;
 307                 rgb24tobgr16= rgb24tobgr16_C;
 308                 rgb24tobgr24= rgb24tobgr24_C;
 309                 rgb32tobgr32= rgb32tobgr32_C;
 310                 rgb32tobgr16= rgb32tobgr16_C;
 311                 rgb32tobgr15= rgb32tobgr15_C;
 312                 yv12toyuy2= yv12toyuy2_C;
 313                 yv12touyvy= yv12touyvy_C;
 314                 yuv422ptoyuy2= yuv422ptoyuy2_C;
 315                 yuy2toyv12= yuy2toyv12_C;
 316 //              uyvytoyv12= uyvytoyv12_C;
 317 //              yvu9toyv12= yvu9toyv12_C;
 318                 planar2x= planar2x_C;
 319                 rgb24toyv12= rgb24toyv12_C;
 320                 interleaveBytes= interleaveBytes_C;
 321                 vu9_to_vu12= vu9_to_vu12_C;
 322                 yvu9_to_yuy2= yvu9_to_yuy2_C;
 323         }
 324 }
 325
 326 /**
 327  * Pallete is assumed to contain bgr32
 328  */
 329 void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 330 {
 331         long i;
 332
 333 /*
 334         for(i=0; i<num_pixels; i++)
 335                 ((unsigned *)dst)[i] = ((unsigned *)palette)[ src[i] ];
 336 */
 337
 338         for(i=0; i<num_pixels; i++)
 339         {
 340                 #ifdef WORDS_BIGENDIAN
 341                         dst[3]= palette[ src[i]*4+2 ];
 342                         dst[2]= palette[ src[i]*4+1 ];
 343                         dst[1]= palette[ src[i]*4+0 ];
 344                 #else
 345                 //FIXME slow?
 346                         dst[0]= palette[ src[i]*4+2 ];
 347                         dst[1]= palette[ src[i]*4+1 ];
 348                         dst[2]= palette[ src[i]*4+0 ];
 349                         //dst[3]= 0; /* do we need this cleansing? */
 350                 #endif
 351                 dst+= 4;
 352         }
 353 }
 354
 355 void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 356 {
 357         long i;
 358         for(i=0; i<num_pixels; i++)
 359         {
 360                 #ifdef WORDS_BIGENDIAN
 361                         dst[3]= palette[ src[i]*4+0 ];
 362                         dst[2]= palette[ src[i]*4+1 ];
 363                         dst[1]= palette[ src[i]*4+2 ];
 364                 #else
 365                         //FIXME slow?
 366                         dst[0]= palette[ src[i]*4+0 ];
 367                         dst[1]= palette[ src[i]*4+1 ];
 368                         dst[2]= palette[ src[i]*4+2 ];
 369                         //dst[3]= 0; /* do we need this cleansing? */
 370                 #endif
 371
 372                 dst+= 4;
 373         }
 374 }
 375
 376 /**
 377  * Pallete is assumed to contain bgr32
 378  */
 379 void palette8torgb24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 380 {
 381         long i;
 382 /*
 383         writes 1 byte o much and might cause alignment issues on some architectures?
 384         for(i=0; i<num_pixels; i++)
 385                 ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
 386 */
 387         for(i=0; i<num_pixels; i++)
 388         {
 389                 //FIXME slow?
 390                 dst[0]= palette[ src[i]*4+2 ];
 391                 dst[1]= palette[ src[i]*4+1 ];
 392                 dst[2]= palette[ src[i]*4+0 ];
 393                 dst+= 3;
 394         }
 395 }
 396
 397 void palette8tobgr24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 398 {
 399         long i;
 400 /*
 401         writes 1 byte o much and might cause alignment issues on some architectures?
 402         for(i=0; i<num_pixels; i++)
 403                 ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
 404 */
 405         for(i=0; i<num_pixels; i++)
 406         {
 407                 //FIXME slow?
 408                 dst[0]= palette[ src[i]*4+0 ];
 409                 dst[1]= palette[ src[i]*4+1 ];
 410                 dst[2]= palette[ src[i]*4+2 ];
 411                 dst+= 3;
 412         }
 413 }
 414
 415 /**
 416  * Palette is assumed to contain bgr16, see rgb32to16 to convert the palette
 417  */
 418 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 419 {
 420         long i;
 421         for(i=0; i<num_pixels; i++)
 422                 ((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
 423 }
 424 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 425 {
 426         long i;
 427         for(i=0; i<num_pixels; i++)
 428                 ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
 429 }
 430
 431 /**
 432  * Pallete is assumed to contain bgr15, see rgb32to15 to convert the palette
 433  */
 434 void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 435 {
 436         long i;
 437         for(i=0; i<num_pixels; i++)
 438                 ((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
 439 }
 440 void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 441 {
 442         long i;
 443         for(i=0; i<num_pixels; i++)
 444                 ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
 445 }
 446
 447 void rgb32tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 448 {
 449         long i;
 450         long num_pixels = src_size >> 2;
 451         for(i=0; i<num_pixels; i++)
 452         {
 453                 #ifdef WORDS_BIGENDIAN
 454                         /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
 455                         dst[3*i + 0] = src[4*i + 1];
 456                         dst[3*i + 1] = src[4*i + 2];
 457                         dst[3*i + 2] = src[4*i + 3];
 458                 #else
 459                         dst[3*i + 0] = src[4*i + 2];
 460                         dst[3*i + 1] = src[4*i + 1];
 461                         dst[3*i + 2] = src[4*i + 0];
 462                 #endif
 463         }
 464 }
 465
 466 void rgb24tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 467 {
 468         long i;
 469         for(i=0; 3*i<src_size; i++)
 470         {
 471                 #ifdef WORDS_BIGENDIAN
 472                         /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
 473                         dst[4*i + 0] = 0;
 474                         dst[4*i + 1] = src[3*i + 0];
 475                         dst[4*i + 2] = src[3*i + 1];
 476                         dst[4*i + 3] = src[3*i + 2];
 477                 #else
 478                         dst[4*i + 0] = src[3*i + 2];
 479                         dst[4*i + 1] = src[3*i + 1];
 480                         dst[4*i + 2] = src[3*i + 0];
 481                         dst[4*i + 3] = 0;
 482                 #endif
 483         }
 484 }
 485
 486 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 487 {
 488         const uint16_t *end;
 489         uint8_t *d = (uint8_t *)dst;
 490         const uint16_t *s = (uint16_t *)src;
 491         end = s + src_size/2;
 492         while(s < end)
 493         {
 494                 register uint16_t bgr;
 495                 bgr = *s++;
 496                 #ifdef WORDS_BIGENDIAN
 497                         *d++ = 0;
 498                         *d++ = (bgr&0x1F)<<3;
 499                         *d++ = (bgr&0x7E0)>>3;
 500                         *d++ = (bgr&0xF800)>>8;
 501                 #else
 502                         *d++ = (bgr&0xF800)>>8;
 503                         *d++ = (bgr&0x7E0)>>3;
 504                         *d++ = (bgr&0x1F)<<3;
 505                         *d++ = 0;
 506                 #endif
 507         }
 508 }
 509
 510 void rgb16tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 511 {
 512         const uint16_t *end;
 513         uint8_t *d = (uint8_t *)dst;
 514         const uint16_t *s = (const uint16_t *)src;
 515         end = s + src_size/2;
 516         while(s < end)
 517         {
 518                 register uint16_t bgr;
 519                 bgr = *s++;
 520                 *d++ = (bgr&0xF800)>>8;
 521                 *d++ = (bgr&0x7E0)>>3;
 522                 *d++ = (bgr&0x1F)<<3;
 523         }
 524 }
 525
 526 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 527 {
 528         long i;
 529         long num_pixels = src_size >> 1;
 530
 531         for(i=0; i<num_pixels; i++)
 532         {
 533             unsigned b,g,r;
 534             register uint16_t rgb;
 535             rgb = src[2*i];
 536             r = rgb&0x1F;
 537             g = (rgb&0x7E0)>>5;
 538             b = (rgb&0xF800)>>11;
 539             dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
 540         }
 541 }
 542
 543 void rgb16tobgr15(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&0x1F)<<5) | ((r&0x1F)<<10);
 557         }
 558 }
 559
 560 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 561 {
 562         const uint16_t *end;
 563         uint8_t *d = (uint8_t *)dst;
 564         const uint16_t *s = (const uint16_t *)src;
 565         end = s + src_size/2;
 566         while(s < end)
 567         {
 568                 register uint16_t bgr;
 569                 bgr = *s++;
 570                 #ifdef WORDS_BIGENDIAN
 571                         *d++ = 0;
 572                         *d++ = (bgr&0x1F)<<3;
 573                         *d++ = (bgr&0x3E0)>>2;
 574                         *d++ = (bgr&0x7C00)>>7;
 575                 #else
 576                         *d++ = (bgr&0x7C00)>>7;
 577                         *d++ = (bgr&0x3E0)>>2;
 578                         *d++ = (bgr&0x1F)<<3;
 579                         *d++ = 0;
 580                 #endif
 581         }
 582 }
 583
 584 void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
 585 {
 586         const uint16_t *end;
 587         uint8_t *d = (uint8_t *)dst;
 588         const uint16_t *s = (uint16_t *)src;
 589         end = s + src_size/2;
 590         while(s < end)
 591         {
 592                 register uint16_t bgr;
 593                 bgr = *s++;
 594                 *d++ = (bgr&0x7C00)>>7;
 595                 *d++ = (bgr&0x3E0)>>2;
 596                 *d++ = (bgr&0x1F)<<3;
 597         }
 598 }
 599
 600 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 601 {
 602         long i;
 603         long num_pixels = src_size >> 1;
 604
 605         for(i=0; i<num_pixels; i++)
 606         {
 607             unsigned b,g,r;
 608             register uint16_t rgb;
 609             rgb = src[2*i];
 610             r = rgb&0x1F;
 611             g = (rgb&0x3E0)>>5;
 612             b = (rgb&0x7C00)>>10;
 613             dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
 614         }
 615 }
 616
 617 void rgb15tobgr15(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&0x1F)<<5) | ((r&0x1F)<<10);
 631         }
 632 }
 633
 634 void rgb8tobgr8(const uint8_t *src, uint8_t *dst, long src_size)
 635 {
 636         long i;
 637         long num_pixels = src_size;
 638         for(i=0; i<num_pixels; i++)
 639         {
 640             unsigned b,g,r;
 641             register uint8_t rgb;
 642             rgb = src[i];
 643             r = (rgb&0x07);
 644             g = (rgb&0x38)>>3;
 645             b = (rgb&0xC0)>>6;
 646             dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
 647         }
 648 }