d3763cc5845b33ef7f9e7a4ad8e207fc7e2bf93f
[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 Libav.
10 *
11 * Libav 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 * Libav 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 Libav; 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 void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
34 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
35 void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
36 void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
37 void (*rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size);
38 void (*rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size);
39 void (*rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size);
40 void (*rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
41 void (*rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size);
42 void (*rgb16to15)(const uint8_t *src, uint8_t *dst, long src_size);
43 void (*rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
44 void (*rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size);
45 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
46 void (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
47 void (*rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size);
48 void (*shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst, long src_size);
49 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
50 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
51
52 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
53 long width, long height,
54 long lumStride, long chromStride, long dstStride);
55 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
56 long width, long height,
57 long lumStride, long chromStride, long dstStride);
58 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
59 long width, long height,
60 long lumStride, long chromStride, long dstStride);
61 void (*yuv422ptouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
62 long width, long height,
63 long lumStride, long chromStride, long dstStride);
64 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
65 long width, long height,
66 long lumStride, long chromStride, long srcStride);
67 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
68 long width, long height,
69 long lumStride, long chromStride, long srcStride);
70 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
71 long srcStride, long dstStride);
72 void (*interleaveBytes)(const uint8_t *src1, const uint8_t *src2, uint8_t *dst,
73 long width, long height, long src1Stride,
74 long src2Stride, long dstStride);
75 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
76 uint8_t *dst1, uint8_t *dst2,
77 long width, long height,
78 long srcStride1, long srcStride2,
79 long dstStride1, long dstStride2);
80 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
81 uint8_t *dst,
82 long width, long height,
83 long srcStride1, long srcStride2,
84 long srcStride3, long dstStride);
85 void (*uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
86 long width, long height,
87 long lumStride, long chromStride, long srcStride);
88 void (*uyvytoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
89 long width, long height,
90 long lumStride, long chromStride, long srcStride);
91 void (*yuyvtoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
92 long width, long height,
93 long lumStride, long chromStride, long srcStride);
94 void (*yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
95 long width, long height,
96 long lumStride, long chromStride, long srcStride);
97
98
99 #if ARCH_X86
100 DECLARE_ASM_CONST(8, uint64_t, mmx_ff) = 0x00000000000000FFULL;
101 DECLARE_ASM_CONST(8, uint64_t, mmx_null) = 0x0000000000000000ULL;
102 DECLARE_ASM_CONST(8, uint64_t, mmx_one) = 0xFFFFFFFFFFFFFFFFULL;
103 DECLARE_ASM_CONST(8, uint64_t, mask32b) = 0x000000FF000000FFULL;
104 DECLARE_ASM_CONST(8, uint64_t, mask32g) = 0x0000FF000000FF00ULL;
105 DECLARE_ASM_CONST(8, uint64_t, mask32r) = 0x00FF000000FF0000ULL;
106 DECLARE_ASM_CONST(8, uint64_t, mask32a) = 0xFF000000FF000000ULL;
107 DECLARE_ASM_CONST(8, uint64_t, mask32) = 0x00FFFFFF00FFFFFFULL;
108 DECLARE_ASM_CONST(8, uint64_t, mask3216br) = 0x00F800F800F800F8ULL;
109 DECLARE_ASM_CONST(8, uint64_t, mask3216g) = 0x0000FC000000FC00ULL;
110 DECLARE_ASM_CONST(8, uint64_t, mask3215g) = 0x0000F8000000F800ULL;
111 DECLARE_ASM_CONST(8, uint64_t, mul3216) = 0x2000000420000004ULL;
112 DECLARE_ASM_CONST(8, uint64_t, mul3215) = 0x2000000820000008ULL;
113 DECLARE_ASM_CONST(8, uint64_t, mask24b) = 0x00FF0000FF0000FFULL;
114 DECLARE_ASM_CONST(8, uint64_t, mask24g) = 0xFF0000FF0000FF00ULL;
115 DECLARE_ASM_CONST(8, uint64_t, mask24r) = 0x0000FF0000FF0000ULL;
116 DECLARE_ASM_CONST(8, uint64_t, mask24l) = 0x0000000000FFFFFFULL;
117 DECLARE_ASM_CONST(8, uint64_t, mask24h) = 0x0000FFFFFF000000ULL;
118 DECLARE_ASM_CONST(8, uint64_t, mask24hh) = 0xffff000000000000ULL;
119 DECLARE_ASM_CONST(8, uint64_t, mask24hhh) = 0xffffffff00000000ULL;
120 DECLARE_ASM_CONST(8, uint64_t, mask24hhhh) = 0xffffffffffff0000ULL;
121 DECLARE_ASM_CONST(8, uint64_t, mask15b) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */
122 DECLARE_ASM_CONST(8, uint64_t, mask15rg) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */
123 DECLARE_ASM_CONST(8, uint64_t, mask15s) = 0xFFE0FFE0FFE0FFE0ULL;
124 DECLARE_ASM_CONST(8, uint64_t, mask15g) = 0x03E003E003E003E0ULL;
125 DECLARE_ASM_CONST(8, uint64_t, mask15r) = 0x7C007C007C007C00ULL;
126 #define mask16b mask15b
127 DECLARE_ASM_CONST(8, uint64_t, mask16g) = 0x07E007E007E007E0ULL;
128 DECLARE_ASM_CONST(8, uint64_t, mask16r) = 0xF800F800F800F800ULL;
129 DECLARE_ASM_CONST(8, uint64_t, red_16mask) = 0x0000f8000000f800ULL;
130 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
131 DECLARE_ASM_CONST(8, uint64_t, blue_16mask) = 0x0000001f0000001fULL;
132 DECLARE_ASM_CONST(8, uint64_t, red_15mask) = 0x00007c0000007c00ULL;
133 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
134 DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;
135 #endif /* ARCH_X86 */
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 versions, there is no 3DNOW + MMX2 one.
149 //plain C versions
150 #define COMPILE_TEMPLATE_MMX 0
151 #define COMPILE_TEMPLATE_MMX2 0
152 #define COMPILE_TEMPLATE_AMD3DNOW 0
153 #define COMPILE_TEMPLATE_SSE2 0
154 #define RENAME(a) a ## _C
155 #include "rgb2rgb_template.c"
156
157 #if ARCH_X86
158
159 //MMX versions
160 #undef RENAME
161 #undef COMPILE_TEMPLATE_MMX
162 #define COMPILE_TEMPLATE_MMX 1
163 #define RENAME(a) a ## _MMX
164 #include "rgb2rgb_template.c"
165
166 //MMX2 versions
167 #undef RENAME
168 #undef COMPILE_TEMPLATE_MMX2
169 #define COMPILE_TEMPLATE_MMX2 1
170 #define RENAME(a) a ## _MMX2
171 #include "rgb2rgb_template.c"
172
173 //SSE2 versions
174 #undef RENAME
175 #undef COMPILE_TEMPLATE_SSE2
176 #define COMPILE_TEMPLATE_SSE2 1
177 #define RENAME(a) a ## _SSE2
178 #include "rgb2rgb_template.c"
179
180 //3DNOW versions
181 #undef RENAME
182 #undef COMPILE_TEMPLATE_MMX2
183 #undef COMPILE_TEMPLATE_SSE2
184 #undef COMPILE_TEMPLATE_AMD3DNOW
185 #define COMPILE_TEMPLATE_MMX2 0
186 #define COMPILE_TEMPLATE_SSE2 1
187 #define COMPILE_TEMPLATE_AMD3DNOW 1
188 #define RENAME(a) a ## _3DNOW
189 #include "rgb2rgb_template.c"
190
191 #endif //ARCH_X86 || ARCH_X86_64
192
193 /*
194 RGB15->RGB16 original by Strepto/Astral
195 ported to gcc & bugfixed : A'rpi
196 MMX2, 3DNOW optimization by Nick Kurshev
197 32-bit C version, and and&add trick by Michael Niedermayer
198 */
199
200 void sws_rgb2rgb_init(int flags)
201 {
202 #if HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX
203 if (flags & SWS_CPU_CAPS_SSE2)
204 rgb2rgb_init_SSE2();
205 else if (flags & SWS_CPU_CAPS_MMX2)
206 rgb2rgb_init_MMX2();
207 else if (flags & SWS_CPU_CAPS_3DNOW)
208 rgb2rgb_init_3DNOW();
209 else if (flags & SWS_CPU_CAPS_MMX)
210 rgb2rgb_init_MMX();
211 else
212 #endif /* HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX */
213 rgb2rgb_init_C();
214 }
215
216 #if LIBSWSCALE_VERSION_MAJOR < 1
217 void palette8topacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
218 {
219 sws_convertPalette8ToPacked32(src, dst, num_pixels, palette);
220 }
221
222 void palette8topacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
223 {
224 sws_convertPalette8ToPacked24(src, dst, num_pixels, palette);
225 }
226
227 /**
228 * Palette is assumed to contain BGR16, see rgb32to16 to convert the palette.
229 */
230 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
231 {
232 long i;
233 for (i=0; i<num_pixels; i++)
234 ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
235 }
236 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
237 {
238 long i;
239 for (i=0; i<num_pixels; i++)
240 ((uint16_t *)dst)[i] = av_bswap16(((const uint16_t *)palette)[src[i]]);
241 }
242 #endif
243
244 void rgb32to24(const uint8_t *src, uint8_t *dst, long src_size)
245 {
246 long i;
247 long num_pixels = src_size >> 2;
248 for (i=0; i<num_pixels; i++) {
249 #if HAVE_BIGENDIAN
250 /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
251 dst[3*i + 0] = src[4*i + 1];
252 dst[3*i + 1] = src[4*i + 2];
253 dst[3*i + 2] = src[4*i + 3];
254 #else
255 dst[3*i + 0] = src[4*i + 2];
256 dst[3*i + 1] = src[4*i + 1];
257 dst[3*i + 2] = src[4*i + 0];
258 #endif
259 }
260 }
261
262 void rgb24to32(const uint8_t *src, uint8_t *dst, long src_size)
263 {
264 long i;
265 for (i=0; 3*i<src_size; i++) {
266 #if HAVE_BIGENDIAN
267 /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
268 dst[4*i + 0] = 255;
269 dst[4*i + 1] = src[3*i + 0];
270 dst[4*i + 2] = src[3*i + 1];
271 dst[4*i + 3] = src[3*i + 2];
272 #else
273 dst[4*i + 0] = src[3*i + 2];
274 dst[4*i + 1] = src[3*i + 1];
275 dst[4*i + 2] = src[3*i + 0];
276 dst[4*i + 3] = 255;
277 #endif
278 }
279 }
280
281 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
282 {
283 const uint16_t *end;
284 uint8_t *d = dst;
285 const uint16_t *s = (const uint16_t *)src;
286 end = s + src_size/2;
287 while (s < end) {
288 register uint16_t bgr;
289 bgr = *s++;
290 #if HAVE_BIGENDIAN
291 *d++ = 255;
292 *d++ = (bgr&0x1F)<<3;
293 *d++ = (bgr&0x7E0)>>3;
294 *d++ = (bgr&0xF800)>>8;
295 #else
296 *d++ = (bgr&0xF800)>>8;
297 *d++ = (bgr&0x7E0)>>3;
298 *d++ = (bgr&0x1F)<<3;
299 *d++ = 255;
300 #endif
301 }
302 }
303
304 void rgb16to24(const uint8_t *src, uint8_t *dst, long src_size)
305 {
306 const uint16_t *end;
307 uint8_t *d = dst;
308 const uint16_t *s = (const uint16_t *)src;
309 end = s + src_size/2;
310 while (s < end) {
311 register uint16_t bgr;
312 bgr = *s++;
313 *d++ = (bgr&0xF800)>>8;
314 *d++ = (bgr&0x7E0)>>3;
315 *d++ = (bgr&0x1F)<<3;
316 }
317 }
318
319 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
320 {
321 long i;
322 long num_pixels = src_size >> 1;
323
324 for (i=0; i<num_pixels; i++) {
325 unsigned rgb = ((const uint16_t*)src)[i];
326 ((uint16_t*)dst)[i] = (rgb>>11) | (rgb&0x7E0) | (rgb<<11);
327 }
328 }
329
330 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
331 {
332 long i;
333 long num_pixels = src_size >> 1;
334
335 for (i=0; i<num_pixels; i++) {
336 unsigned rgb = ((const uint16_t*)src)[i];
337 ((uint16_t*)dst)[i] = (rgb>>11) | ((rgb&0x7C0)>>1) | ((rgb&0x1F)<<10);
338 }
339 }
340
341 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
342 {
343 const uint16_t *end;
344 uint8_t *d = dst;
345 const uint16_t *s = (const uint16_t *)src;
346 end = s + src_size/2;
347 while (s < end) {
348 register uint16_t bgr;
349 bgr = *s++;
350 #if HAVE_BIGENDIAN
351 *d++ = 255;
352 *d++ = (bgr&0x1F)<<3;
353 *d++ = (bgr&0x3E0)>>2;
354 *d++ = (bgr&0x7C00)>>7;
355 #else
356 *d++ = (bgr&0x7C00)>>7;
357 *d++ = (bgr&0x3E0)>>2;
358 *d++ = (bgr&0x1F)<<3;
359 *d++ = 255;
360 #endif
361 }
362 }
363
364 void rgb15to24(const uint8_t *src, uint8_t *dst, long src_size)
365 {
366 const uint16_t *end;
367 uint8_t *d = dst;
368 const uint16_t *s = (const uint16_t *)src;
369 end = s + src_size/2;
370 while (s < end) {
371 register uint16_t bgr;
372 bgr = *s++;
373 *d++ = (bgr&0x7C00)>>7;
374 *d++ = (bgr&0x3E0)>>2;
375 *d++ = (bgr&0x1F)<<3;
376 }
377 }
378
379 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
380 {
381 long i;
382 long num_pixels = src_size >> 1;
383
384 for (i=0; i<num_pixels; i++) {
385 unsigned rgb = ((const uint16_t*)src)[i];
386 ((uint16_t*)dst)[i] = ((rgb&0x7C00)>>10) | ((rgb&0x3E0)<<1) | (rgb<<11);
387 }
388 }
389
390 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
391 {
392 long i;
393 long num_pixels = src_size >> 1;
394
395 for (i=0; i<num_pixels; i++) {
396 unsigned br;
397 unsigned rgb = ((const uint16_t*)src)[i];
398 br = rgb&0x7c1F;
399 ((uint16_t*)dst)[i] = (br>>10) | (rgb&0x3E0) | (br<<10);
400 }
401 }
402
403 void bgr8torgb8(const uint8_t *src, uint8_t *dst, long src_size)
404 {
405 long i;
406 long num_pixels = src_size;
407 for (i=0; i<num_pixels; i++) {
408 unsigned b,g,r;
409 register uint8_t rgb;
410 rgb = src[i];
411 r = (rgb&0x07);
412 g = (rgb&0x38)>>3;
413 b = (rgb&0xC0)>>6;
414 dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
415 }
416 }
417
418 #define DEFINE_SHUFFLE_BYTES(a, b, c, d) \
419 void shuffle_bytes_##a##b##c##d(const uint8_t *src, uint8_t *dst, long src_size) \
420 { \
421 long i; \
422 \
423 for (i = 0; i < src_size; i+=4) { \
424 dst[i + 0] = src[i + a]; \
425 dst[i + 1] = src[i + b]; \
426 dst[i + 2] = src[i + c]; \
427 dst[i + 3] = src[i + d]; \
428 } \
429 }
430
431 DEFINE_SHUFFLE_BYTES(0, 3, 2, 1);
432 DEFINE_SHUFFLE_BYTES(1, 2, 3, 0);
433 DEFINE_SHUFFLE_BYTES(3, 0, 1, 2);
434 DEFINE_SHUFFLE_BYTES(3, 2, 1, 0);
435