1dc728ab9d87a4cbebe965cc0801e0dde3769fe1
[libav.git] / libswscale / rgb2rgb.c
1 /*
2 * software RGB to RGB converter
3 * pluralize by software PAL8 to RGB converter
4 * software YUV to YUV converter
5 * software YUV to RGB converter
6 * Written by Nick Kurshev.
7 * palette & YUV & runtime CPU stuff by Michael (michaelni@gmx.at)
8 *
9 * This file is part of FFmpeg.
10 *
11 * FFmpeg is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License as published by
13 * the Free Software Foundation; either version 2 of the License, or
14 * (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
19 * GNU General Public License for more details.
20 *
21 * You should have received a copy of the GNU General Public License
22 * 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 * The C code (not assembly, MMX, ...) of this file can be used
26 * under the LGPL license.
27 */
28 #include <inttypes.h>
29 #include "config.h"
30 #include "libavutil/x86_cpu.h"
31 #include "libavutil/bswap.h"
32 #include "rgb2rgb.h"
33 #include "swscale.h"
34 #include "swscale_internal.h"
35
36 #define FAST_BGR2YV12 // use 7-bit instead of 15-bit coefficients
37
38 void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
39 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
40 void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
41 void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
42 void (*rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size);
43 void (*rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size);
44 void (*rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size);
45 void (*rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
46 void (*rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size);
47 void (*rgb16to15)(const uint8_t *src, uint8_t *dst, long src_size);
48 void (*rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
49 void (*rgb16to32)(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 (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
52 void (*rgb24to15)(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 (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
55 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
56
57 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
58 long width, long height,
59 long lumStride, long chromStride, long dstStride);
60 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
61 long width, long height,
62 long lumStride, long chromStride, long dstStride);
63 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
64 long width, long height,
65 long lumStride, long chromStride, long dstStride);
66 void (*yuv422ptouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
67 long width, long height,
68 long lumStride, long chromStride, long dstStride);
69 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
70 long width, long height,
71 long lumStride, long chromStride, long srcStride);
72 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
73 long width, long height,
74 long lumStride, long chromStride, long srcStride);
75 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
76 long srcStride, long dstStride);
77 void (*interleaveBytes)(const uint8_t *src1, const uint8_t *src2, uint8_t *dst,
78 long width, long height, long src1Stride,
79 long src2Stride, long dstStride);
80 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
81 uint8_t *dst1, uint8_t *dst2,
82 long width, long height,
83 long srcStride1, long srcStride2,
84 long dstStride1, long dstStride2);
85 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
86 uint8_t *dst,
87 long width, long height,
88 long srcStride1, long srcStride2,
89 long srcStride3, long dstStride);
90 void (*uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
91 long width, long height,
92 long lumStride, long chromStride, long srcStride);
93 void (*uyvytoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
94 long width, long height,
95 long lumStride, long chromStride, long srcStride);
96 void (*yuyvtoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
97 long width, long height,
98 long lumStride, long chromStride, long srcStride);
99 void (*yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
100 long width, long height,
101 long lumStride, long chromStride, long srcStride);
102
103
104 #if ARCH_X86 && CONFIG_GPL
105 DECLARE_ASM_CONST(8, uint64_t, mmx_null) = 0x0000000000000000ULL;
106 DECLARE_ASM_CONST(8, uint64_t, mmx_one) = 0xFFFFFFFFFFFFFFFFULL;
107 DECLARE_ASM_CONST(8, uint64_t, mask32b) = 0x000000FF000000FFULL;
108 DECLARE_ASM_CONST(8, uint64_t, mask32g) = 0x0000FF000000FF00ULL;
109 DECLARE_ASM_CONST(8, uint64_t, mask32r) = 0x00FF000000FF0000ULL;
110 DECLARE_ASM_CONST(8, uint64_t, mask32a) = 0xFF000000FF000000ULL;
111 DECLARE_ASM_CONST(8, uint64_t, mask32) = 0x00FFFFFF00FFFFFFULL;
112 DECLARE_ASM_CONST(8, uint64_t, mask3216br) = 0x00F800F800F800F8ULL;
113 DECLARE_ASM_CONST(8, uint64_t, mask3216g) = 0x0000FC000000FC00ULL;
114 DECLARE_ASM_CONST(8, uint64_t, mask3215g) = 0x0000F8000000F800ULL;
115 DECLARE_ASM_CONST(8, uint64_t, mul3216) = 0x2000000420000004ULL;
116 DECLARE_ASM_CONST(8, uint64_t, mul3215) = 0x2000000820000008ULL;
117 DECLARE_ASM_CONST(8, uint64_t, mask24b) = 0x00FF0000FF0000FFULL;
118 DECLARE_ASM_CONST(8, uint64_t, mask24g) = 0xFF0000FF0000FF00ULL;
119 DECLARE_ASM_CONST(8, uint64_t, mask24r) = 0x0000FF0000FF0000ULL;
120 DECLARE_ASM_CONST(8, uint64_t, mask24l) = 0x0000000000FFFFFFULL;
121 DECLARE_ASM_CONST(8, uint64_t, mask24h) = 0x0000FFFFFF000000ULL;
122 DECLARE_ASM_CONST(8, uint64_t, mask24hh) = 0xffff000000000000ULL;
123 DECLARE_ASM_CONST(8, uint64_t, mask24hhh) = 0xffffffff00000000ULL;
124 DECLARE_ASM_CONST(8, uint64_t, mask24hhhh) = 0xffffffffffff0000ULL;
125 DECLARE_ASM_CONST(8, uint64_t, mask15b) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */
126 DECLARE_ASM_CONST(8, uint64_t, mask15rg) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */
127 DECLARE_ASM_CONST(8, uint64_t, mask15s) = 0xFFE0FFE0FFE0FFE0ULL;
128 DECLARE_ASM_CONST(8, uint64_t, mask15g) = 0x03E003E003E003E0ULL;
129 DECLARE_ASM_CONST(8, uint64_t, mask15r) = 0x7C007C007C007C00ULL;
130 #define mask16b mask15b
131 DECLARE_ASM_CONST(8, uint64_t, mask16g) = 0x07E007E007E007E0ULL;
132 DECLARE_ASM_CONST(8, uint64_t, mask16r) = 0xF800F800F800F800ULL;
133 DECLARE_ASM_CONST(8, uint64_t, red_16mask) = 0x0000f8000000f800ULL;
134 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
135 DECLARE_ASM_CONST(8, uint64_t, blue_16mask) = 0x0000001f0000001fULL;
136 DECLARE_ASM_CONST(8, uint64_t, red_15mask) = 0x00007c0000007c00ULL;
137 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
138 DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;
139 #endif /* ARCH_X86 */
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 versions, there is no 3DNOW + MMX2 one.
153 //plain C versions
154 #undef HAVE_MMX
155 #undef HAVE_MMX2
156 #undef HAVE_AMD3DNOW
157 #undef HAVE_SSE2
158 #define HAVE_MMX 0
159 #define HAVE_MMX2 0
160 #define HAVE_AMD3DNOW 0
161 #define HAVE_SSE2 0
162 #define RENAME(a) a ## _C
163 #include "rgb2rgb_template.c"
164
165 #if ARCH_X86 && CONFIG_GPL
166
167 //MMX versions
168 #undef RENAME
169 #undef HAVE_MMX
170 #define HAVE_MMX 1
171 #define RENAME(a) a ## _MMX
172 #include "rgb2rgb_template.c"
173
174 //MMX2 versions
175 #undef RENAME
176 #undef HAVE_MMX2
177 #define HAVE_MMX2 1
178 #define RENAME(a) a ## _MMX2
179 #include "rgb2rgb_template.c"
180
181 //3DNOW versions
182 #undef RENAME
183 #undef HAVE_MMX2
184 #undef HAVE_AMD3DNOW
185 #define HAVE_MMX2 0
186 #define HAVE_AMD3DNOW 1
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 32-bit C version, and and&add trick by Michael Niedermayer
197 */
198
199 void sws_rgb2rgb_init(int flags)
200 {
201 #if (HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX) && CONFIG_GPL
202 if (flags & SWS_CPU_CAPS_MMX2)
203 rgb2rgb_init_MMX2();
204 else if (flags & SWS_CPU_CAPS_3DNOW)
205 rgb2rgb_init_3DNOW();
206 else if (flags & SWS_CPU_CAPS_MMX)
207 rgb2rgb_init_MMX();
208 else
209 #endif /* HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX */
210 rgb2rgb_init_C();
211 }
212
213 /**
214 * Convert the palette to the same packet 32-bit format as the palette
215 */
216 void palette8topacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
217 {
218 long i;
219
220 for (i=0; i<num_pixels; i++)
221 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
222 }
223
224 /**
225 * Palette format: ABCD -> dst format: ABC
226 */
227 void palette8topacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
228 {
229 long i;
230
231 for (i=0; i<num_pixels; i++) {
232 //FIXME slow?
233 dst[0]= palette[src[i]*4+0];
234 dst[1]= palette[src[i]*4+1];
235 dst[2]= palette[src[i]*4+2];
236 dst+= 3;
237 }
238 }
239
240 /**
241 * Palette is assumed to contain BGR16, see rgb32to16 to convert the palette.
242 */
243 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
244 {
245 long i;
246 for (i=0; i<num_pixels; i++)
247 ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
248 }
249 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
250 {
251 long i;
252 for (i=0; i<num_pixels; i++)
253 ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
254 }
255
256 /**
257 * Palette is assumed to contain BGR15, see rgb32to15 to convert the palette.
258 */
259 void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
260 {
261 long i;
262 for (i=0; i<num_pixels; i++)
263 ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
264 }
265 void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
266 {
267 long i;
268 for (i=0; i<num_pixels; i++)
269 ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
270 }
271
272 void rgb32to24(const uint8_t *src, uint8_t *dst, long src_size)
273 {
274 long i;
275 long num_pixels = src_size >> 2;
276 for (i=0; i<num_pixels; i++) {
277 #if HAVE_BIGENDIAN
278 /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
279 dst[3*i + 0] = src[4*i + 1];
280 dst[3*i + 1] = src[4*i + 2];
281 dst[3*i + 2] = src[4*i + 3];
282 #else
283 dst[3*i + 0] = src[4*i + 2];
284 dst[3*i + 1] = src[4*i + 1];
285 dst[3*i + 2] = src[4*i + 0];
286 #endif
287 }
288 }
289
290 void rgb24to32(const uint8_t *src, uint8_t *dst, long src_size)
291 {
292 long i;
293 for (i=0; 3*i<src_size; i++) {
294 #if HAVE_BIGENDIAN
295 /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
296 dst[4*i + 0] = 255;
297 dst[4*i + 1] = src[3*i + 0];
298 dst[4*i + 2] = src[3*i + 1];
299 dst[4*i + 3] = src[3*i + 2];
300 #else
301 dst[4*i + 0] = src[3*i + 2];
302 dst[4*i + 1] = src[3*i + 1];
303 dst[4*i + 2] = src[3*i + 0];
304 dst[4*i + 3] = 255;
305 #endif
306 }
307 }
308
309 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
310 {
311 const uint16_t *end;
312 uint8_t *d = dst;
313 const uint16_t *s = (const uint16_t *)src;
314 end = s + src_size/2;
315 while (s < end) {
316 register uint16_t bgr;
317 bgr = *s++;
318 #if HAVE_BIGENDIAN
319 *d++ = 255;
320 *d++ = (bgr&0x1F)<<3;
321 *d++ = (bgr&0x7E0)>>3;
322 *d++ = (bgr&0xF800)>>8;
323 #else
324 *d++ = (bgr&0xF800)>>8;
325 *d++ = (bgr&0x7E0)>>3;
326 *d++ = (bgr&0x1F)<<3;
327 *d++ = 255;
328 #endif
329 }
330 }
331
332 void rgb16to24(const uint8_t *src, uint8_t *dst, long src_size)
333 {
334 const uint16_t *end;
335 uint8_t *d = dst;
336 const uint16_t *s = (const uint16_t *)src;
337 end = s + src_size/2;
338 while (s < end) {
339 register uint16_t bgr;
340 bgr = *s++;
341 *d++ = (bgr&0xF800)>>8;
342 *d++ = (bgr&0x7E0)>>3;
343 *d++ = (bgr&0x1F)<<3;
344 }
345 }
346
347 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
348 {
349 long i;
350 long num_pixels = src_size >> 1;
351
352 for (i=0; i<num_pixels; i++) {
353 unsigned rgb = ((const uint16_t*)src)[i];
354 ((uint16_t*)dst)[i] = (rgb>>11) | (rgb&0x7E0) | (rgb<<11);
355 }
356 }
357
358 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
359 {
360 long i;
361 long num_pixels = src_size >> 1;
362
363 for (i=0; i<num_pixels; i++) {
364 unsigned rgb = ((const uint16_t*)src)[i];
365 ((uint16_t*)dst)[i] = (rgb>>11) | ((rgb&0x7C0)>>1) | ((rgb&0x1F)<<10);
366 }
367 }
368
369 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
370 {
371 const uint16_t *end;
372 uint8_t *d = dst;
373 const uint16_t *s = (const uint16_t *)src;
374 end = s + src_size/2;
375 while (s < end) {
376 register uint16_t bgr;
377 bgr = *s++;
378 #if HAVE_BIGENDIAN
379 *d++ = 255;
380 *d++ = (bgr&0x1F)<<3;
381 *d++ = (bgr&0x3E0)>>2;
382 *d++ = (bgr&0x7C00)>>7;
383 #else
384 *d++ = (bgr&0x7C00)>>7;
385 *d++ = (bgr&0x3E0)>>2;
386 *d++ = (bgr&0x1F)<<3;
387 *d++ = 255;
388 #endif
389 }
390 }
391
392 void rgb15to24(const uint8_t *src, uint8_t *dst, long src_size)
393 {
394 const uint16_t *end;
395 uint8_t *d = dst;
396 const uint16_t *s = (const uint16_t *)src;
397 end = s + src_size/2;
398 while (s < end) {
399 register uint16_t bgr;
400 bgr = *s++;
401 *d++ = (bgr&0x7C00)>>7;
402 *d++ = (bgr&0x3E0)>>2;
403 *d++ = (bgr&0x1F)<<3;
404 }
405 }
406
407 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
408 {
409 long i;
410 long num_pixels = src_size >> 1;
411
412 for (i=0; i<num_pixels; i++) {
413 unsigned rgb = ((const uint16_t*)src)[i];
414 ((uint16_t*)dst)[i] = ((rgb&0x7C00)>>10) | ((rgb&0x3E0)<<1) | (rgb<<11);
415 }
416 }
417
418 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
419 {
420 long i;
421 long num_pixels = src_size >> 1;
422
423 for (i=0; i<num_pixels; i++) {
424 unsigned br;
425 unsigned rgb = ((const uint16_t*)src)[i];
426 br = rgb&0x7c1F;
427 ((uint16_t*)dst)[i] = (br>>10) | (rgb&0x3E0) | (br<<10);
428 }
429 }
430
431 void bgr8torgb8(const uint8_t *src, uint8_t *dst, long src_size)
432 {
433 long i;
434 long num_pixels = src_size;
435 for (i=0; i<num_pixels; i++) {
436 unsigned b,g,r;
437 register uint8_t rgb;
438 rgb = src[i];
439 r = (rgb&0x07);
440 g = (rgb&0x38)>>3;
441 b = (rgb&0xC0)>>6;
442 dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
443 }
444 }