?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
colorspace-test: fix build
[libswscale.git] / yuv2rgb.c
blob77d3d399daaffef00418e654b0b31f659d96803e
1 /*
2 * software YUV to RGB converter
3 *
4 * Copyright (C) 2009 Konstantin Shishkov
5 *
6 * 1,4,8bpp support and context / deglobalize stuff
7 * by Michael Niedermayer (michaelni@gmx.at)
8 *
9 * This file is part of FFmpeg.
10 *
11 * FFmpeg is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public
13 * License as published by the Free Software Foundation; either
14 * version 2.1 of the License, or (at your option) any later version.
15 *
16 * FFmpeg is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
20 *
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with FFmpeg; if not, write to the Free Software
23 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 */
25
26 #include <stdio.h>
27 #include <stdlib.h>
28 #include <inttypes.h>
29 #include <assert.h>
30
31 #include "config.h"
32 #include "rgb2rgb.h"
33 #include "swscale.h"
34 #include "swscale_internal.h"
35 #include "libavutil/x86_cpu.h"
36 #include "libavutil/bswap.h"
37
38 extern const uint8_t dither_4x4_16[4][8];
39 extern const uint8_t dither_8x8_32[8][8];
40 extern const uint8_t dither_8x8_73[8][8];
41 extern const uint8_t dither_8x8_220[8][8];
42
43 const int32_t ff_yuv2rgb_coeffs[8][4] = {
44 {117504, 138453, 13954, 34903}, /* no sequence_display_extension */
45 {117504, 138453, 13954, 34903}, /* ITU-R Rec. 709 (1990) */
46 {104597, 132201, 25675, 53279}, /* unspecified */
47 {104597, 132201, 25675, 53279}, /* reserved */
48 {104448, 132798, 24759, 53109}, /* FCC */
49 {104597, 132201, 25675, 53279}, /* ITU-R Rec. 624-4 System B, G */
50 {104597, 132201, 25675, 53279}, /* SMPTE 170M */
51 {117579, 136230, 16907, 35559} /* SMPTE 240M (1987) */
52 };
53
54 const int *sws_getCoefficients(int colorspace)
55 {
56 if (colorspace > 7 || colorspace < 0)
57 colorspace = SWS_CS_DEFAULT;
58 return ff_yuv2rgb_coeffs[colorspace];
59 }
60
61 #define LOADCHROMA(i) \
62 U = pu[i]; \
63 V = pv[i]; \
64 r = (void *)c->table_rV[V]; \
65 g = (void *)(c->table_gU[U] + c->table_gV[V]); \
66 b = (void *)c->table_bU[U];
67
68 #define PUTRGB(dst,src,i) \
69 Y = src[2*i]; \
70 dst[2*i ] = r[Y] + g[Y] + b[Y]; \
71 Y = src[2*i+1]; \
72 dst[2*i+1] = r[Y] + g[Y] + b[Y];
73
74 #define PUTRGB24(dst,src,i) \
75 Y = src[2*i]; \
76 dst[6*i+0] = r[Y]; dst[6*i+1] = g[Y]; dst[6*i+2] = b[Y]; \
77 Y = src[2*i+1]; \
78 dst[6*i+3] = r[Y]; dst[6*i+4] = g[Y]; dst[6*i+5] = b[Y];
79
80 #define PUTBGR24(dst,src,i) \
81 Y = src[2*i]; \
82 dst[6*i+0] = b[Y]; dst[6*i+1] = g[Y]; dst[6*i+2] = r[Y]; \
83 Y = src[2*i+1]; \
84 dst[6*i+3] = b[Y]; dst[6*i+4] = g[Y]; dst[6*i+5] = r[Y];
85
86 #define PUTRGBA(dst,ysrc,asrc,i,s) \
87 Y = ysrc[2*i]; \
88 dst[2*i ] = r[Y] + g[Y] + b[Y] + (asrc[2*i ]<<s); \
89 Y = ysrc[2*i+1]; \
90 dst[2*i+1] = r[Y] + g[Y] + b[Y] + (asrc[2*i+1]<<s);
91
92 #define PUTRGB48(dst,src,i) \
93 Y = src[2*i]; \
94 dst[12*i+ 0] = dst[12*i+ 1] = r[Y]; \
95 dst[12*i+ 2] = dst[12*i+ 3] = g[Y]; \
96 dst[12*i+ 4] = dst[12*i+ 5] = b[Y]; \
97 Y = src[2*i+1]; \
98 dst[12*i+ 6] = dst[12*i+ 7] = r[Y]; \
99 dst[12*i+ 8] = dst[12*i+ 9] = g[Y]; \
100 dst[12*i+10] = dst[12*i+11] = b[Y];
101
102 #define YUV2RGBFUNC(func_name, dst_type, alpha) \
103 static int func_name(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY, \
104 int srcSliceH, uint8_t* dst[], int dstStride[]) \
105 {\
106 int y;\
107 \
108 if (!alpha && c->srcFormat == PIX_FMT_YUV422P) {\
109 srcStride[1] *= 2;\
110 srcStride[2] *= 2;\
111 }\
112 for (y=0; y<srcSliceH; y+=2) {\
113 dst_type *dst_1 = (dst_type*)(dst[0] + (y+srcSliceY )*dstStride[0]);\
114 dst_type *dst_2 = (dst_type*)(dst[0] + (y+srcSliceY+1)*dstStride[0]);\
115 dst_type av_unused *r, *b;\
116 dst_type *g;\
117 const uint8_t *py_1 = src[0] + y*srcStride[0];\
118 const uint8_t *py_2 = py_1 + srcStride[0];\
119 const uint8_t *pu = src[1] + (y>>1)*srcStride[1];\
120 const uint8_t *pv = src[2] + (y>>1)*srcStride[2];\
121 const uint8_t av_unused *pa_1, *pa_2;\
122 unsigned int h_size = c->dstW>>3;\
123 if (alpha) {\
124 pa_1 = src[3] + y*srcStride[3];\
125 pa_2 = pa_1 + srcStride[3];\
126 }\
127 while (h_size--) {\
128 int av_unused U, V;\
129 int Y;\
130
131 #define ENDYUV2RGBLINE(dst_delta)\
132 pu += 4;\
133 pv += 4;\
134 py_1 += 8;\
135 py_2 += 8;\
136 dst_1 += dst_delta;\
137 dst_2 += dst_delta;\
138 }\
139 if (c->dstW & 4) {\
140 int av_unused Y, U, V;\
141
142 #define ENDYUV2RGBFUNC()\
143 }\
144 }\
145 return srcSliceH;\
146 }
147
148 #define CLOSEYUV2RGBFUNC(dst_delta)\
149 ENDYUV2RGBLINE(dst_delta)\
150 ENDYUV2RGBFUNC()
151
152 YUV2RGBFUNC(yuv2rgb_c_48, uint8_t, 0)
153 LOADCHROMA(0);
154 PUTRGB48(dst_1,py_1,0);
155 PUTRGB48(dst_2,py_2,0);
156
157 LOADCHROMA(1);
158 PUTRGB48(dst_2,py_2,1);
159 PUTRGB48(dst_1,py_1,1);
160
161 LOADCHROMA(2);
162 PUTRGB48(dst_1,py_1,2);
163 PUTRGB48(dst_2,py_2,2);
164
165 LOADCHROMA(3);
166 PUTRGB48(dst_2,py_2,3);
167 PUTRGB48(dst_1,py_1,3);
168 ENDYUV2RGBLINE(48)
169 LOADCHROMA(0);
170 PUTRGB48(dst_1,py_1,0);
171 PUTRGB48(dst_2,py_2,0);
172
173 LOADCHROMA(1);
174 PUTRGB48(dst_2,py_2,1);
175 PUTRGB48(dst_1,py_1,1);
176 ENDYUV2RGBFUNC()
177
178 YUV2RGBFUNC(yuv2rgb_c_32, uint32_t, 0)
179 LOADCHROMA(0);
180 PUTRGB(dst_1,py_1,0);
181 PUTRGB(dst_2,py_2,0);
182
183 LOADCHROMA(1);
184 PUTRGB(dst_2,py_2,1);
185 PUTRGB(dst_1,py_1,1);
186
187 LOADCHROMA(2);
188 PUTRGB(dst_1,py_1,2);
189 PUTRGB(dst_2,py_2,2);
190
191 LOADCHROMA(3);
192 PUTRGB(dst_2,py_2,3);
193 PUTRGB(dst_1,py_1,3);
194 ENDYUV2RGBLINE(8)
195 LOADCHROMA(0);
196 PUTRGB(dst_1,py_1,0);
197 PUTRGB(dst_2,py_2,0);
198
199 LOADCHROMA(1);
200 PUTRGB(dst_2,py_2,1);
201 PUTRGB(dst_1,py_1,1);
202 ENDYUV2RGBFUNC()
203
204 YUV2RGBFUNC(yuva2rgba_c, uint32_t, 1)
205 LOADCHROMA(0);
206 PUTRGBA(dst_1,py_1,pa_1,0,24);
207 PUTRGBA(dst_2,py_2,pa_2,0,24);
208
209 LOADCHROMA(1);
210 PUTRGBA(dst_2,py_2,pa_1,1,24);
211 PUTRGBA(dst_1,py_1,pa_2,1,24);
212
213 LOADCHROMA(2);
214 PUTRGBA(dst_1,py_1,pa_1,2,24);
215 PUTRGBA(dst_2,py_2,pa_2,2,24);
216
217 LOADCHROMA(3);
218 PUTRGBA(dst_2,py_2,pa_1,3,24);
219 PUTRGBA(dst_1,py_1,pa_2,3,24);
220 pa_1 += 8;\
221 pa_2 += 8;\
222 ENDYUV2RGBLINE(8)
223 LOADCHROMA(0);
224 PUTRGBA(dst_1,py_1,pa_1,0,24);
225 PUTRGBA(dst_2,py_2,pa_2,0,24);
226
227 LOADCHROMA(1);
228 PUTRGBA(dst_2,py_2,pa_1,1,24);
229 PUTRGBA(dst_1,py_1,pa_2,1,24);
230 ENDYUV2RGBFUNC()
231
232 YUV2RGBFUNC(yuva2argb_c, uint32_t, 1)
233 LOADCHROMA(0);
234 PUTRGBA(dst_1,py_1,pa_1,0,0);
235 PUTRGBA(dst_2,py_2,pa_2,0,0);
236
237 LOADCHROMA(1);
238 PUTRGBA(dst_2,py_2,pa_2,1,0);
239 PUTRGBA(dst_1,py_1,pa_1,1,0);
240
241 LOADCHROMA(2);
242 PUTRGBA(dst_1,py_1,pa_1,2,0);
243 PUTRGBA(dst_2,py_2,pa_2,2,0);
244
245 LOADCHROMA(3);
246 PUTRGBA(dst_2,py_2,pa_2,3,0);
247 PUTRGBA(dst_1,py_1,pa_1,3,0);
248 pa_1 += 8;\
249 pa_2 += 8;\
250 ENDYUV2RGBLINE(8)
251 LOADCHROMA(0);
252 PUTRGBA(dst_1,py_1,pa_1,0,0);
253 PUTRGBA(dst_2,py_2,pa_2,0,0);
254
255 LOADCHROMA(1);
256 PUTRGBA(dst_2,py_2,pa_2,1,0);
257 PUTRGBA(dst_1,py_1,pa_1,1,0);
258 ENDYUV2RGBFUNC()
259
260 YUV2RGBFUNC(yuv2rgb_c_24_rgb, uint8_t, 0)
261 LOADCHROMA(0);
262 PUTRGB24(dst_1,py_1,0);
263 PUTRGB24(dst_2,py_2,0);
264
265 LOADCHROMA(1);
266 PUTRGB24(dst_2,py_2,1);
267 PUTRGB24(dst_1,py_1,1);
268
269 LOADCHROMA(2);
270 PUTRGB24(dst_1,py_1,2);
271 PUTRGB24(dst_2,py_2,2);
272
273 LOADCHROMA(3);
274 PUTRGB24(dst_2,py_2,3);
275 PUTRGB24(dst_1,py_1,3);
276 ENDYUV2RGBLINE(24)
277 LOADCHROMA(0);
278 PUTRGB24(dst_1,py_1,0);
279 PUTRGB24(dst_2,py_2,0);
280
281 LOADCHROMA(1);
282 PUTRGB24(dst_2,py_2,1);
283 PUTRGB24(dst_1,py_1,1);
284 ENDYUV2RGBFUNC()
285
286 // only trivial mods from yuv2rgb_c_24_rgb
287 YUV2RGBFUNC(yuv2rgb_c_24_bgr, uint8_t, 0)
288 LOADCHROMA(0);
289 PUTBGR24(dst_1,py_1,0);
290 PUTBGR24(dst_2,py_2,0);
291
292 LOADCHROMA(1);
293 PUTBGR24(dst_2,py_2,1);
294 PUTBGR24(dst_1,py_1,1);
295
296 LOADCHROMA(2);
297 PUTBGR24(dst_1,py_1,2);
298 PUTBGR24(dst_2,py_2,2);
299
300 LOADCHROMA(3);
301 PUTBGR24(dst_2,py_2,3);
302 PUTBGR24(dst_1,py_1,3);
303 ENDYUV2RGBLINE(24)
304 LOADCHROMA(0);
305 PUTBGR24(dst_1,py_1,0);
306 PUTBGR24(dst_2,py_2,0);
307
308 LOADCHROMA(1);
309 PUTBGR24(dst_2,py_2,1);
310 PUTBGR24(dst_1,py_1,1);
311 ENDYUV2RGBFUNC()
312
313 // This is exactly the same code as yuv2rgb_c_32 except for the types of
314 // r, g, b, dst_1, dst_2
315 YUV2RGBFUNC(yuv2rgb_c_16, uint16_t, 0)
316 LOADCHROMA(0);
317 PUTRGB(dst_1,py_1,0);
318 PUTRGB(dst_2,py_2,0);
319
320 LOADCHROMA(1);
321 PUTRGB(dst_2,py_2,1);
322 PUTRGB(dst_1,py_1,1);
323
324 LOADCHROMA(2);
325 PUTRGB(dst_1,py_1,2);
326 PUTRGB(dst_2,py_2,2);
327
328 LOADCHROMA(3);
329 PUTRGB(dst_2,py_2,3);
330 PUTRGB(dst_1,py_1,3);
331 CLOSEYUV2RGBFUNC(8)
332
333 #if 0 // Currently unused
334 // This is exactly the same code as yuv2rgb_c_32 except for the types of
335 // r, g, b, dst_1, dst_2
336 YUV2RGBFUNC(yuv2rgb_c_8, uint8_t, 0)
337 LOADCHROMA(0);
338 PUTRGB(dst_1,py_1,0);
339 PUTRGB(dst_2,py_2,0);
340
341 LOADCHROMA(1);
342 PUTRGB(dst_2,py_2,1);
343 PUTRGB(dst_1,py_1,1);
344
345 LOADCHROMA(2);
346 PUTRGB(dst_1,py_1,2);
347 PUTRGB(dst_2,py_2,2);
348
349 LOADCHROMA(3);
350 PUTRGB(dst_2,py_2,3);
351 PUTRGB(dst_1,py_1,3);
352 CLOSEYUV2RGBFUNC(8)
353 #endif
354
355 // r, g, b, dst_1, dst_2
356 YUV2RGBFUNC(yuv2rgb_c_12_ordered_dither, uint16_t, 0)
357 const uint8_t *d16 = dither_4x4_16[y&3];
358 #define PUTRGB12(dst,src,i,o) \
359 Y = src[2*i]; \
360 dst[2*i] = r[Y+d16[0+o]] + g[Y+d16[0+o]] + b[Y+d16[0+o]]; \
361 Y = src[2*i+1]; \
362 dst[2*i+1] = r[Y+d16[1+o]] + g[Y+d16[1+o]] + b[Y+d16[1+o]];
363
364 LOADCHROMA(0);
365 PUTRGB12(dst_1,py_1,0,0);
366 PUTRGB12(dst_2,py_2,0,0+8);
367
368 LOADCHROMA(1);
369 PUTRGB12(dst_2,py_2,1,2+8);
370 PUTRGB12(dst_1,py_1,1,2);
371
372 LOADCHROMA(2);
373 PUTRGB12(dst_1,py_1,2,4);
374 PUTRGB12(dst_2,py_2,2,4+8);
375
376 LOADCHROMA(3);
377 PUTRGB12(dst_2,py_2,3,6+8);
378 PUTRGB12(dst_1,py_1,3,6);
379 CLOSEYUV2RGBFUNC(8)
380
381 // r, g, b, dst_1, dst_2
382 YUV2RGBFUNC(yuv2rgb_c_8_ordered_dither, uint8_t, 0)
383 const uint8_t *d32 = dither_8x8_32[y&7];
384 const uint8_t *d64 = dither_8x8_73[y&7];
385 #define PUTRGB8(dst,src,i,o) \
386 Y = src[2*i]; \
387 dst[2*i] = r[Y+d32[0+o]] + g[Y+d32[0+o]] + b[Y+d64[0+o]]; \
388 Y = src[2*i+1]; \
389 dst[2*i+1] = r[Y+d32[1+o]] + g[Y+d32[1+o]] + b[Y+d64[1+o]];
390
391 LOADCHROMA(0);
392 PUTRGB8(dst_1,py_1,0,0);
393 PUTRGB8(dst_2,py_2,0,0+8);
394
395 LOADCHROMA(1);
396 PUTRGB8(dst_2,py_2,1,2+8);
397 PUTRGB8(dst_1,py_1,1,2);
398
399 LOADCHROMA(2);
400 PUTRGB8(dst_1,py_1,2,4);
401 PUTRGB8(dst_2,py_2,2,4+8);
402
403 LOADCHROMA(3);
404 PUTRGB8(dst_2,py_2,3,6+8);
405 PUTRGB8(dst_1,py_1,3,6);
406 CLOSEYUV2RGBFUNC(8)
407
408 #if 0 // Currently unused
409 // This is exactly the same code as yuv2rgb_c_32 except for the types of
410 // r, g, b, dst_1, dst_2
411 YUV2RGBFUNC(yuv2rgb_c_4, uint8_t, 0)
412 int acc;
413 #define PUTRGB4(dst,src,i) \
414 Y = src[2*i]; \
415 acc = r[Y] + g[Y] + b[Y]; \
416 Y = src[2*i+1]; \
417 acc |= (r[Y] + g[Y] + b[Y])<<4; \
418 dst[i] = acc;
419
420 LOADCHROMA(0);
421 PUTRGB4(dst_1,py_1,0);
422 PUTRGB4(dst_2,py_2,0);
423
424 LOADCHROMA(1);
425 PUTRGB4(dst_2,py_2,1);
426 PUTRGB4(dst_1,py_1,1);
427
428 LOADCHROMA(2);
429 PUTRGB4(dst_1,py_1,2);
430 PUTRGB4(dst_2,py_2,2);
431
432 LOADCHROMA(3);
433 PUTRGB4(dst_2,py_2,3);
434 PUTRGB4(dst_1,py_1,3);
435 CLOSEYUV2RGBFUNC(4)
436 #endif
437
438 YUV2RGBFUNC(yuv2rgb_c_4_ordered_dither, uint8_t, 0)
439 const uint8_t *d64 = dither_8x8_73[y&7];
440 const uint8_t *d128 = dither_8x8_220[y&7];
441 int acc;
442
443 #define PUTRGB4D(dst,src,i,o) \
444 Y = src[2*i]; \
445 acc = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]]; \
446 Y = src[2*i+1]; \
447 acc |= (r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]])<<4; \
448 dst[i]= acc;
449
450 LOADCHROMA(0);
451 PUTRGB4D(dst_1,py_1,0,0);
452 PUTRGB4D(dst_2,py_2,0,0+8);
453
454 LOADCHROMA(1);
455 PUTRGB4D(dst_2,py_2,1,2+8);
456 PUTRGB4D(dst_1,py_1,1,2);
457
458 LOADCHROMA(2);
459 PUTRGB4D(dst_1,py_1,2,4);
460 PUTRGB4D(dst_2,py_2,2,4+8);
461
462 LOADCHROMA(3);
463 PUTRGB4D(dst_2,py_2,3,6+8);
464 PUTRGB4D(dst_1,py_1,3,6);
465 CLOSEYUV2RGBFUNC(4)
466
467 #if 0 // Currently unused
468 // This is exactly the same code as yuv2rgb_c_32 except for the types of
469 // r, g, b, dst_1, dst_2
470 YUV2RGBFUNC(yuv2rgb_c_4b, uint8_t, 0)
471 LOADCHROMA(0);
472 PUTRGB(dst_1,py_1,0);
473 PUTRGB(dst_2,py_2,0);
474
475 LOADCHROMA(1);
476 PUTRGB(dst_2,py_2,1);
477 PUTRGB(dst_1,py_1,1);
478
479 LOADCHROMA(2);
480 PUTRGB(dst_1,py_1,2);
481 PUTRGB(dst_2,py_2,2);
482
483 LOADCHROMA(3);
484 PUTRGB(dst_2,py_2,3);
485 PUTRGB(dst_1,py_1,3);
486 CLOSEYUV2RGBFUNC(8)
487 #endif
488
489 YUV2RGBFUNC(yuv2rgb_c_4b_ordered_dither, uint8_t, 0)
490 const uint8_t *d64 = dither_8x8_73[y&7];
491 const uint8_t *d128 = dither_8x8_220[y&7];
492
493 #define PUTRGB4DB(dst,src,i,o) \
494 Y = src[2*i]; \
495 dst[2*i] = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]]; \
496 Y = src[2*i+1]; \
497 dst[2*i+1] = r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]];
498
499 LOADCHROMA(0);
500 PUTRGB4DB(dst_1,py_1,0,0);
501 PUTRGB4DB(dst_2,py_2,0,0+8);
502
503 LOADCHROMA(1);
504 PUTRGB4DB(dst_2,py_2,1,2+8);
505 PUTRGB4DB(dst_1,py_1,1,2);
506
507 LOADCHROMA(2);
508 PUTRGB4DB(dst_1,py_1,2,4);
509 PUTRGB4DB(dst_2,py_2,2,4+8);
510
511 LOADCHROMA(3);
512 PUTRGB4DB(dst_2,py_2,3,6+8);
513 PUTRGB4DB(dst_1,py_1,3,6);
514 CLOSEYUV2RGBFUNC(8)
515
516 YUV2RGBFUNC(yuv2rgb_c_1_ordered_dither, uint8_t, 0)
517 const uint8_t *d128 = dither_8x8_220[y&7];
518 char out_1 = 0, out_2 = 0;
519 g= c->table_gU[128] + c->table_gV[128];
520
521 #define PUTRGB1(out,src,i,o) \
522 Y = src[2*i]; \
523 out+= out + g[Y+d128[0+o]]; \
524 Y = src[2*i+1]; \
525 out+= out + g[Y+d128[1+o]];
526
527 PUTRGB1(out_1,py_1,0,0);
528 PUTRGB1(out_2,py_2,0,0+8);
529
530 PUTRGB1(out_2,py_2,1,2+8);
531 PUTRGB1(out_1,py_1,1,2);
532
533 PUTRGB1(out_1,py_1,2,4);
534 PUTRGB1(out_2,py_2,2,4+8);
535
536 PUTRGB1(out_2,py_2,3,6+8);
537 PUTRGB1(out_1,py_1,3,6);
538
539 dst_1[0]= out_1;
540 dst_2[0]= out_2;
541 CLOSEYUV2RGBFUNC(1)
542
543 SwsFunc ff_yuv2rgb_get_func_ptr(SwsContext *c)
544 {
545 SwsFunc t = NULL;
546 #if HAVE_MMX
547 t = ff_yuv2rgb_init_mmx(c);
548 #endif
549 #if HAVE_VIS
550 t = ff_yuv2rgb_init_vis(c);
551 #endif
552 #if CONFIG_MLIB
553 t = ff_yuv2rgb_init_mlib(c);
554 #endif
555 #if HAVE_ALTIVEC
556 if (c->flags & SWS_CPU_CAPS_ALTIVEC)
557 t = ff_yuv2rgb_init_altivec(c);
558 #endif
559
560 #if ARCH_BFIN
561 if (c->flags & SWS_CPU_CAPS_BFIN)
562 t = ff_yuv2rgb_get_func_ptr_bfin(c);
563 #endif
564
565 if (t)
566 return t;
567
568 av_log(c, AV_LOG_WARNING, "No accelerated colorspace conversion found from %s to %s.\n", sws_format_name(c->srcFormat), sws_format_name(c->dstFormat));
569
570 switch (c->dstFormat) {
571 case PIX_FMT_RGB48BE:
572 case PIX_FMT_RGB48LE: return yuv2rgb_c_48;
573 case PIX_FMT_ARGB:
574 case PIX_FMT_ABGR: if (CONFIG_SWSCALE_ALPHA && c->srcFormat == PIX_FMT_YUVA420P) return yuva2argb_c;
575 case PIX_FMT_RGBA:
576 case PIX_FMT_BGRA: return (CONFIG_SWSCALE_ALPHA && c->srcFormat == PIX_FMT_YUVA420P) ? yuva2rgba_c : yuv2rgb_c_32;
577 case PIX_FMT_RGB24: return yuv2rgb_c_24_rgb;
578 case PIX_FMT_BGR24: return yuv2rgb_c_24_bgr;
579 case PIX_FMT_RGB565:
580 case PIX_FMT_BGR565:
581 case PIX_FMT_RGB555:
582 case PIX_FMT_BGR555: return yuv2rgb_c_16;
583 case PIX_FMT_RGB444:
584 case PIX_FMT_BGR444: return yuv2rgb_c_12_ordered_dither;
585 case PIX_FMT_RGB8:
586 case PIX_FMT_BGR8: return yuv2rgb_c_8_ordered_dither;
587 case PIX_FMT_RGB4:
588 case PIX_FMT_BGR4: return yuv2rgb_c_4_ordered_dither;
589 case PIX_FMT_RGB4_BYTE:
590 case PIX_FMT_BGR4_BYTE: return yuv2rgb_c_4b_ordered_dither;
591 case PIX_FMT_MONOBLACK: return yuv2rgb_c_1_ordered_dither;
592 default:
593 assert(0);
594 }
595 return NULL;
596 }
597
598 static void fill_table(uint8_t* table[256], const int elemsize, const int inc, void *y_tab)
599 {
600 int i;
601 int64_t cb = 0;
602 uint8_t *y_table = y_tab;
603
604 y_table -= elemsize * (inc >> 9);
605
606 for (i = 0; i < 256; i++) {
607 table[i] = y_table + elemsize * (cb >> 16);
608 cb += inc;
609 }
610 }
611
612 static void fill_gv_table(int table[256], const int elemsize, const int inc)
613 {
614 int i;
615 int64_t cb = 0;
616 int off = -(inc >> 9);
617
618 for (i = 0; i < 256; i++) {
619 table[i] = elemsize * (off + (cb >> 16));
620 cb += inc;
621 }
622 }
623
624 static uint16_t roundToInt16(int64_t f)
625 {
626 int r= (f + (1<<15))>>16;
627 if (r<-0x7FFF) return 0x8000;
628 else if (r> 0x7FFF) return 0x7FFF;
629 else return r;
630 }
631
632 av_cold int ff_yuv2rgb_c_init_tables(SwsContext *c, const int inv_table[4], int fullRange,
633 int brightness, int contrast, int saturation)
634 {
635 const int isRgb = c->dstFormat==PIX_FMT_RGB32
636 || c->dstFormat==PIX_FMT_RGB32_1
637 || c->dstFormat==PIX_FMT_BGR24
638 || c->dstFormat==PIX_FMT_RGB565BE
639 || c->dstFormat==PIX_FMT_RGB565LE
640 || c->dstFormat==PIX_FMT_RGB555BE
641 || c->dstFormat==PIX_FMT_RGB555LE
642 || c->dstFormat==PIX_FMT_RGB444BE
643 || c->dstFormat==PIX_FMT_RGB444LE
644 || c->dstFormat==PIX_FMT_RGB8
645 || c->dstFormat==PIX_FMT_RGB4
646 || c->dstFormat==PIX_FMT_RGB4_BYTE
647 || c->dstFormat==PIX_FMT_MONOBLACK;
648 const int isNotNe = c->dstFormat==PIX_FMT_NE(RGB565LE,RGB565BE)
649 || c->dstFormat==PIX_FMT_NE(RGB555LE,RGB555BE)
650 || c->dstFormat==PIX_FMT_NE(RGB444LE,RGB444BE)
651 || c->dstFormat==PIX_FMT_NE(BGR565LE,BGR565BE)
652 || c->dstFormat==PIX_FMT_NE(BGR555LE,BGR555BE)
653 || c->dstFormat==PIX_FMT_NE(BGR444LE,BGR444BE);
654 const int bpp = c->dstFormatBpp;
655 uint8_t *y_table;
656 uint16_t *y_table16;
657 uint32_t *y_table32;
658 int i, base, rbase, gbase, bbase, abase, needAlpha;
659 const int yoffs = fullRange ? 384 : 326;
660
661 int64_t crv = inv_table[0];
662 int64_t cbu = inv_table[1];
663 int64_t cgu = -inv_table[2];
664 int64_t cgv = -inv_table[3];
665 int64_t cy = 1<<16;
666 int64_t oy = 0;
667
668 int64_t yb = 0;
669
670 if (!fullRange) {
671 cy = (cy*255) / 219;
672 oy = 16<<16;
673 } else {
674 crv = (crv*224) / 255;
675 cbu = (cbu*224) / 255;
676 cgu = (cgu*224) / 255;
677 cgv = (cgv*224) / 255;
678 }
679
680 cy = (cy *contrast ) >> 16;
681 crv = (crv*contrast * saturation) >> 32;
682 cbu = (cbu*contrast * saturation) >> 32;
683 cgu = (cgu*contrast * saturation) >> 32;
684 cgv = (cgv*contrast * saturation) >> 32;
685 oy -= 256*brightness;
686
687 c->uOffset= 0x0400040004000400LL;
688 c->vOffset= 0x0400040004000400LL;
689 c->yCoeff= roundToInt16(cy *8192) * 0x0001000100010001ULL;
690 c->vrCoeff= roundToInt16(crv*8192) * 0x0001000100010001ULL;
691 c->ubCoeff= roundToInt16(cbu*8192) * 0x0001000100010001ULL;
692 c->vgCoeff= roundToInt16(cgv*8192) * 0x0001000100010001ULL;
693 c->ugCoeff= roundToInt16(cgu*8192) * 0x0001000100010001ULL;
694 c->yOffset= roundToInt16(oy * 8) * 0x0001000100010001ULL;
695
696 c->yuv2rgb_y_coeff = (int16_t)roundToInt16(cy <<13);
697 c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
698 c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
699 c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
700 c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
701 c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
702
703 //scale coefficients by cy
704 crv = ((crv << 16) + 0x8000) / cy;
705 cbu = ((cbu << 16) + 0x8000) / cy;
706 cgu = ((cgu << 16) + 0x8000) / cy;
707 cgv = ((cgv << 16) + 0x8000) / cy;
708
709 av_free(c->yuvTable);
710
711 switch (bpp) {
712 case 1:
713 c->yuvTable = av_malloc(1024);
714 y_table = c->yuvTable;
715 yb = -(384<<16) - oy;
716 for (i = 0; i < 1024-110; i++) {
717 y_table[i+110] = av_clip_uint8((yb + 0x8000) >> 16) >> 7;
718 yb += cy;
719 }
720 fill_table(c->table_gU, 1, cgu, y_table + yoffs);
721 fill_gv_table(c->table_gV, 1, cgv);
722 break;
723 case 4:
724 case 4|128:
725 rbase = isRgb ? 3 : 0;
726 gbase = 1;
727 bbase = isRgb ? 0 : 3;
728 c->yuvTable = av_malloc(1024*3);
729 y_table = c->yuvTable;
730 yb = -(384<<16) - oy;
731 for (i = 0; i < 1024-110; i++) {
732 int yval = av_clip_uint8((yb + 0x8000) >> 16);
733 y_table[i+110 ] = (yval >> 7) << rbase;
734 y_table[i+ 37+1024] = ((yval + 43) / 85) << gbase;
735 y_table[i+110+2048] = (yval >> 7) << bbase;
736 yb += cy;
737 }
738 fill_table(c->table_rV, 1, crv, y_table + yoffs);
739 fill_table(c->table_gU, 1, cgu, y_table + yoffs + 1024);
740 fill_table(c->table_bU, 1, cbu, y_table + yoffs + 2048);
741 fill_gv_table(c->table_gV, 1, cgv);
742 break;
743 case 8:
744 rbase = isRgb ? 5 : 0;
745 gbase = isRgb ? 2 : 3;
746 bbase = isRgb ? 0 : 6;
747 c->yuvTable = av_malloc(1024*3);
748 y_table = c->yuvTable;
749 yb = -(384<<16) - oy;
750 for (i = 0; i < 1024-38; i++) {
751 int yval = av_clip_uint8((yb + 0x8000) >> 16);
752 y_table[i+16 ] = ((yval + 18) / 36) << rbase;
753 y_table[i+16+1024] = ((yval + 18) / 36) << gbase;
754 y_table[i+37+2048] = ((yval + 43) / 85) << bbase;
755 yb += cy;
756 }
757 fill_table(c->table_rV, 1, crv, y_table + yoffs);
758 fill_table(c->table_gU, 1, cgu, y_table + yoffs + 1024);
759 fill_table(c->table_bU, 1, cbu, y_table + yoffs + 2048);
760 fill_gv_table(c->table_gV, 1, cgv);
761 break;
762 case 12:
763 rbase = isRgb ? 8 : 0;
764 gbase = 4;
765 bbase = isRgb ? 0 : 8;
766 c->yuvTable = av_malloc(1024*3*2);
767 y_table16 = c->yuvTable;
768 yb = -(384<<16) - oy;
769 for (i = 0; i < 1024; i++) {
770 uint8_t yval = av_clip_uint8((yb + 0x8000) >> 16);
771 y_table16[i ] = (yval >> 4) << rbase;
772 y_table16[i+1024] = (yval >> 4) << gbase;
773 y_table16[i+2048] = (yval >> 4) << bbase;
774 yb += cy;
775 }
776 if (isNotNe)
777 for (i = 0; i < 1024*3; i++)
778 y_table16[i] = av_bswap16(y_table16[i]);
779 fill_table(c->table_rV, 2, crv, y_table16 + yoffs);
780 fill_table(c->table_gU, 2, cgu, y_table16 + yoffs + 1024);
781 fill_table(c->table_bU, 2, cbu, y_table16 + yoffs + 2048);
782 fill_gv_table(c->table_gV, 2, cgv);
783 break;
784 case 15:
785 case 16:
786 rbase = isRgb ? bpp - 5 : 0;
787 gbase = 5;
788 bbase = isRgb ? 0 : (bpp - 5);
789 c->yuvTable = av_malloc(1024*3*2);
790 y_table16 = c->yuvTable;
791 yb = -(384<<16) - oy;
792 for (i = 0; i < 1024; i++) {
793 uint8_t yval = av_clip_uint8((yb + 0x8000) >> 16);
794 y_table16[i ] = (yval >> 3) << rbase;
795 y_table16[i+1024] = (yval >> (18 - bpp)) << gbase;
796 y_table16[i+2048] = (yval >> 3) << bbase;
797 yb += cy;
798 }
799 if(isNotNe)
800 for (i = 0; i < 1024*3; i++)
801 y_table16[i] = av_bswap16(y_table16[i]);
802 fill_table(c->table_rV, 2, crv, y_table16 + yoffs);
803 fill_table(c->table_gU, 2, cgu, y_table16 + yoffs + 1024);
804 fill_table(c->table_bU, 2, cbu, y_table16 + yoffs + 2048);
805 fill_gv_table(c->table_gV, 2, cgv);
806 break;
807 case 24:
808 case 48:
809 c->yuvTable = av_malloc(1024);
810 y_table = c->yuvTable;
811 yb = -(384<<16) - oy;
812 for (i = 0; i < 1024; i++) {
813 y_table[i] = av_clip_uint8((yb + 0x8000) >> 16);
814 yb += cy;
815 }
816 fill_table(c->table_rV, 1, crv, y_table + yoffs);
817 fill_table(c->table_gU, 1, cgu, y_table + yoffs);
818 fill_table(c->table_bU, 1, cbu, y_table + yoffs);
819 fill_gv_table(c->table_gV, 1, cgv);
820 break;
821 case 32:
822 base = (c->dstFormat == PIX_FMT_RGB32_1 || c->dstFormat == PIX_FMT_BGR32_1) ? 8 : 0;
823 rbase = base + (isRgb ? 16 : 0);
824 gbase = base + 8;
825 bbase = base + (isRgb ? 0 : 16);
826 needAlpha = CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat);
827 if (!needAlpha)
828 abase = (base + 24) & 31;
829 c->yuvTable = av_malloc(1024*3*4);
830 y_table32 = c->yuvTable;
831 yb = -(384<<16) - oy;
832 for (i = 0; i < 1024; i++) {
833 uint8_t yval = av_clip_uint8((yb + 0x8000) >> 16);
834 y_table32[i ] = (yval << rbase) + (needAlpha ? 0 : (255 << abase));
835 y_table32[i+1024] = yval << gbase;
836 y_table32[i+2048] = yval << bbase;
837 yb += cy;
838 }
839 fill_table(c->table_rV, 4, crv, y_table32 + yoffs);
840 fill_table(c->table_gU, 4, cgu, y_table32 + yoffs + 1024);
841 fill_table(c->table_bU, 4, cbu, y_table32 + yoffs + 2048);
842 fill_gv_table(c->table_gV, 4, cgv);
843 break;
844 default:
845 c->yuvTable = NULL;
846 av_log(c, AV_LOG_ERROR, "%ibpp not supported by yuv2rgb\n", bpp);
847 return -1;
848 }
849 return 0;
850 }