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