remove palette8torgb15 and palette8tobgr15
[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
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 #include <inttypes.h>
26 #include "config.h"
27 #include "libavutil/x86_cpu.h"
28 #include "libavutil/bswap.h"
29 #include "rgb2rgb.h"
30 #include "swscale.h"
31 #include "swscale_internal.h"
32
33 #define FAST_BGR2YV12 // use 7-bit instead of 15-bit coefficients
34
35 void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
36 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
37 void (*rgb24tobgr15)(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 (*rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size);
40 void (*rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size);
41 void (*rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size);
42 void (*rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
43 void (*rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size);
44 void (*rgb16to15)(const uint8_t *src, uint8_t *dst, long src_size);
45 void (*rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
46 void (*rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size);
47 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
48 void (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
49 void (*rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size);
50 void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
51 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
52 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
53
54 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
55 long width, long height,
56 long lumStride, long chromStride, long dstStride);
57 void (*yv12touyvy)(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 (*yuv422ptoyuy2)(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 (*yuv422ptouyvy)(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 (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
67 long width, long height,
68 long lumStride, long chromStride, long srcStride);
69 void (*rgb24toyv12)(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 (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
73 long srcStride, long dstStride);
74 void (*interleaveBytes)(const uint8_t *src1, const uint8_t *src2, uint8_t *dst,
75 long width, long height, long src1Stride,
76 long src2Stride, long dstStride);
77 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
78 uint8_t *dst1, uint8_t *dst2,
79 long width, long height,
80 long srcStride1, long srcStride2,
81 long dstStride1, long dstStride2);
82 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
83 uint8_t *dst,
84 long width, long height,
85 long srcStride1, long srcStride2,
86 long srcStride3, long dstStride);
87 void (*uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
88 long width, long height,
89 long lumStride, long chromStride, long srcStride);
90 void (*uyvytoyuv422)(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 (*yuyvtoyuv420)(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 (*yuyvtoyuv422)(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
100
101 #if ARCH_X86
102 DECLARE_ASM_CONST(8, uint64_t, mmx_null) = 0x0000000000000000ULL;
103 DECLARE_ASM_CONST(8, uint64_t, mmx_one) = 0xFFFFFFFFFFFFFFFFULL;
104 DECLARE_ASM_CONST(8, uint64_t, mask32b) = 0x000000FF000000FFULL;
105 DECLARE_ASM_CONST(8, uint64_t, mask32g) = 0x0000FF000000FF00ULL;
106 DECLARE_ASM_CONST(8, uint64_t, mask32r) = 0x00FF000000FF0000ULL;
107 DECLARE_ASM_CONST(8, uint64_t, mask32a) = 0xFF000000FF000000ULL;
108 DECLARE_ASM_CONST(8, uint64_t, mask32) = 0x00FFFFFF00FFFFFFULL;
109 DECLARE_ASM_CONST(8, uint64_t, mask3216br) = 0x00F800F800F800F8ULL;
110 DECLARE_ASM_CONST(8, uint64_t, mask3216g) = 0x0000FC000000FC00ULL;
111 DECLARE_ASM_CONST(8, uint64_t, mask3215g) = 0x0000F8000000F800ULL;
112 DECLARE_ASM_CONST(8, uint64_t, mul3216) = 0x2000000420000004ULL;
113 DECLARE_ASM_CONST(8, uint64_t, mul3215) = 0x2000000820000008ULL;
114 DECLARE_ASM_CONST(8, uint64_t, mask24b) = 0x00FF0000FF0000FFULL;
115 DECLARE_ASM_CONST(8, uint64_t, mask24g) = 0xFF0000FF0000FF00ULL;
116 DECLARE_ASM_CONST(8, uint64_t, mask24r) = 0x0000FF0000FF0000ULL;
117 DECLARE_ASM_CONST(8, uint64_t, mask24l) = 0x0000000000FFFFFFULL;
118 DECLARE_ASM_CONST(8, uint64_t, mask24h) = 0x0000FFFFFF000000ULL;
119 DECLARE_ASM_CONST(8, uint64_t, mask24hh) = 0xffff000000000000ULL;
120 DECLARE_ASM_CONST(8, uint64_t, mask24hhh) = 0xffffffff00000000ULL;
121 DECLARE_ASM_CONST(8, uint64_t, mask24hhhh) = 0xffffffffffff0000ULL;
122 DECLARE_ASM_CONST(8, uint64_t, mask15b) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */
123 DECLARE_ASM_CONST(8, uint64_t, mask15rg) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */
124 DECLARE_ASM_CONST(8, uint64_t, mask15s) = 0xFFE0FFE0FFE0FFE0ULL;
125 DECLARE_ASM_CONST(8, uint64_t, mask15g) = 0x03E003E003E003E0ULL;
126 DECLARE_ASM_CONST(8, uint64_t, mask15r) = 0x7C007C007C007C00ULL;
127 #define mask16b mask15b
128 DECLARE_ASM_CONST(8, uint64_t, mask16g) = 0x07E007E007E007E0ULL;
129 DECLARE_ASM_CONST(8, uint64_t, mask16r) = 0xF800F800F800F800ULL;
130 DECLARE_ASM_CONST(8, uint64_t, red_16mask) = 0x0000f8000000f800ULL;
131 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
132 DECLARE_ASM_CONST(8, uint64_t, blue_16mask) = 0x0000001f0000001fULL;
133 DECLARE_ASM_CONST(8, uint64_t, red_15mask) = 0x00007c0000007c00ULL;
134 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
135 DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;
136 #endif /* ARCH_X86 */
137
138 #define RGB2YUV_SHIFT 8
139 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
140 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
141 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
142 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
143 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
144 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
145 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
146 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
147 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
148
149 //Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
150 //plain C versions
151 #undef HAVE_MMX
152 #undef HAVE_MMX2
153 #undef HAVE_AMD3DNOW
154 #undef HAVE_SSE2
155 #define HAVE_MMX 0
156 #define HAVE_MMX2 0
157 #define HAVE_AMD3DNOW 0
158 #define HAVE_SSE2 0
159 #define RENAME(a) a ## _C
160 #include "rgb2rgb_template.c"
161
162 #if ARCH_X86
163
164 //MMX versions
165 #undef RENAME
166 #undef HAVE_MMX
167 #define HAVE_MMX 1
168 #define RENAME(a) a ## _MMX
169 #include "rgb2rgb_template.c"
170
171 //MMX2 versions
172 #undef RENAME
173 #undef HAVE_MMX2
174 #define HAVE_MMX2 1
175 #define RENAME(a) a ## _MMX2
176 #include "rgb2rgb_template.c"
177
178 //3DNOW versions
179 #undef RENAME
180 #undef HAVE_MMX2
181 #undef HAVE_AMD3DNOW
182 #define HAVE_MMX2 0
183 #define HAVE_AMD3DNOW 1
184 #define RENAME(a) a ## _3DNOW
185 #include "rgb2rgb_template.c"
186
187 #endif //ARCH_X86 || ARCH_X86_64
188
189 /*
190 RGB15->RGB16 original by Strepto/Astral
191 ported to gcc & bugfixed : A'rpi
192 MMX2, 3DNOW optimization by Nick Kurshev
193 32-bit C version, and and&add trick by Michael Niedermayer
194 */
195
196 void sws_rgb2rgb_init(int flags)
197 {
198 #if HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX
199 if (flags & SWS_CPU_CAPS_MMX2)
200 rgb2rgb_init_MMX2();
201 else if (flags & SWS_CPU_CAPS_3DNOW)
202 rgb2rgb_init_3DNOW();
203 else if (flags & SWS_CPU_CAPS_MMX)
204 rgb2rgb_init_MMX();
205 else
206 #endif /* HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX */
207 rgb2rgb_init_C();
208 }
209
210 /**
211 * Convert the palette to the same packet 32-bit format as the palette
212 */
213 void palette8topacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
214 {
215 long i;
216
217 for (i=0; i<num_pixels; i++)
218 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
219 }
220
221 /**
222 * Palette format: ABCD -> dst format: ABC
223 */
224 void palette8topacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
225 {
226 long i;
227
228 for (i=0; i<num_pixels; i++) {
229 //FIXME slow?
230 dst[0]= palette[src[i]*4+0];
231 dst[1]= palette[src[i]*4+1];
232 dst[2]= palette[src[i]*4+2];
233 dst+= 3;
234 }
235 }
236
237 /**
238 * Palette is assumed to contain BGR16, see rgb32to16 to convert the palette.
239 */
240 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
241 {
242 long i;
243 for (i=0; i<num_pixels; i++)
244 ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
245 }
246 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
247 {
248 long i;
249 for (i=0; i<num_pixels; i++)
250 ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
251 }
252
253 void rgb32to24(const uint8_t *src, uint8_t *dst, long src_size)
254 {
255 long i;
256 long num_pixels = src_size >> 2;
257 for (i=0; i<num_pixels; i++) {
258 #if HAVE_BIGENDIAN
259 /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
260 dst[3*i + 0] = src[4*i + 1];
261 dst[3*i + 1] = src[4*i + 2];
262 dst[3*i + 2] = src[4*i + 3];
263 #else
264 dst[3*i + 0] = src[4*i + 2];
265 dst[3*i + 1] = src[4*i + 1];
266 dst[3*i + 2] = src[4*i + 0];
267 #endif
268 }
269 }
270
271 void rgb24to32(const uint8_t *src, uint8_t *dst, long src_size)
272 {
273 long i;
274 for (i=0; 3*i<src_size; i++) {
275 #if HAVE_BIGENDIAN
276 /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
277 dst[4*i + 0] = 255;
278 dst[4*i + 1] = src[3*i + 0];
279 dst[4*i + 2] = src[3*i + 1];
280 dst[4*i + 3] = src[3*i + 2];
281 #else
282 dst[4*i + 0] = src[3*i + 2];
283 dst[4*i + 1] = src[3*i + 1];
284 dst[4*i + 2] = src[3*i + 0];
285 dst[4*i + 3] = 255;
286 #endif
287 }
288 }
289
290 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
291 {
292 const uint16_t *end;
293 uint8_t *d = dst;
294 const uint16_t *s = (const uint16_t *)src;
295 end = s + src_size/2;
296 while (s < end) {
297 register uint16_t bgr;
298 bgr = *s++;
299 #if HAVE_BIGENDIAN
300 *d++ = 255;
301 *d++ = (bgr&0x1F)<<3;
302 *d++ = (bgr&0x7E0)>>3;
303 *d++ = (bgr&0xF800)>>8;
304 #else
305 *d++ = (bgr&0xF800)>>8;
306 *d++ = (bgr&0x7E0)>>3;
307 *d++ = (bgr&0x1F)<<3;
308 *d++ = 255;
309 #endif
310 }
311 }
312
313 void rgb16to24(const uint8_t *src, uint8_t *dst, long src_size)
314 {
315 const uint16_t *end;
316 uint8_t *d = dst;
317 const uint16_t *s = (const uint16_t *)src;
318 end = s + src_size/2;
319 while (s < end) {
320 register uint16_t bgr;
321 bgr = *s++;
322 *d++ = (bgr&0xF800)>>8;
323 *d++ = (bgr&0x7E0)>>3;
324 *d++ = (bgr&0x1F)<<3;
325 }
326 }
327
328 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
329 {
330 long i;
331 long num_pixels = src_size >> 1;
332
333 for (i=0; i<num_pixels; i++) {
334 unsigned rgb = ((const uint16_t*)src)[i];
335 ((uint16_t*)dst)[i] = (rgb>>11) | (rgb&0x7E0) | (rgb<<11);
336 }
337 }
338
339 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
340 {
341 long i;
342 long num_pixels = src_size >> 1;
343
344 for (i=0; i<num_pixels; i++) {
345 unsigned rgb = ((const uint16_t*)src)[i];
346 ((uint16_t*)dst)[i] = (rgb>>11) | ((rgb&0x7C0)>>1) | ((rgb&0x1F)<<10);
347 }
348 }
349
350 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
351 {
352 const uint16_t *end;
353 uint8_t *d = dst;
354 const uint16_t *s = (const uint16_t *)src;
355 end = s + src_size/2;
356 while (s < end) {
357 register uint16_t bgr;
358 bgr = *s++;
359 #if HAVE_BIGENDIAN
360 *d++ = 255;
361 *d++ = (bgr&0x1F)<<3;
362 *d++ = (bgr&0x3E0)>>2;
363 *d++ = (bgr&0x7C00)>>7;
364 #else
365 *d++ = (bgr&0x7C00)>>7;
366 *d++ = (bgr&0x3E0)>>2;
367 *d++ = (bgr&0x1F)<<3;
368 *d++ = 255;
369 #endif
370 }
371 }
372
373 void rgb15to24(const uint8_t *src, uint8_t *dst, long src_size)
374 {
375 const uint16_t *end;
376 uint8_t *d = dst;
377 const uint16_t *s = (const uint16_t *)src;
378 end = s + src_size/2;
379 while (s < end) {
380 register uint16_t bgr;
381 bgr = *s++;
382 *d++ = (bgr&0x7C00)>>7;
383 *d++ = (bgr&0x3E0)>>2;
384 *d++ = (bgr&0x1F)<<3;
385 }
386 }
387
388 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
389 {
390 long i;
391 long num_pixels = src_size >> 1;
392
393 for (i=0; i<num_pixels; i++) {
394 unsigned rgb = ((const uint16_t*)src)[i];
395 ((uint16_t*)dst)[i] = ((rgb&0x7C00)>>10) | ((rgb&0x3E0)<<1) | (rgb<<11);
396 }
397 }
398
399 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
400 {
401 long i;
402 long num_pixels = src_size >> 1;
403
404 for (i=0; i<num_pixels; i++) {
405 unsigned br;
406 unsigned rgb = ((const uint16_t*)src)[i];
407 br = rgb&0x7c1F;
408 ((uint16_t*)dst)[i] = (br>>10) | (rgb&0x3E0) | (br<<10);
409 }
410 }
411
412 void bgr8torgb8(const uint8_t *src, uint8_t *dst, long src_size)
413 {
414 long i;
415 long num_pixels = src_size;
416 for (i=0; i<num_pixels; i++) {
417 unsigned b,g,r;
418 register uint8_t rgb;
419 rgb = src[i];
420 r = (rgb&0x07);
421 g = (rgb&0x38)>>3;
422 b = (rgb&0xC0)>>6;
423 dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
424 }
425 }
426
427 #define DEFINE_SHUFFLE_BYTES(a, b, c, d) \
428 void shuffle_bytes_##a##b##c##d(const uint8_t *src, uint8_t *dst, long src_size) \
429 { \
430 long i; \
431 \
432 for (i = 0; i < src_size; i+=4) { \
433 dst[i + 0] = src[i + a]; \
434 dst[i + 1] = src[i + b]; \
435 dst[i + 2] = src[i + c]; \
436 dst[i + 3] = src[i + d]; \
437 } \
438 }
439
440 DEFINE_SHUFFLE_BYTES(0, 3, 2, 1);
441 DEFINE_SHUFFLE_BYTES(1, 2, 3, 0);
442 DEFINE_SHUFFLE_BYTES(2, 1, 0, 3);
443 DEFINE_SHUFFLE_BYTES(3, 0, 1, 2);
444 DEFINE_SHUFFLE_BYTES(3, 2, 1, 0);
445