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)
9 * This file is part of FFmpeg.
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.
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.
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
25 * The C code (not assembly, MMX, ...) of this file can be used
26 * under the LGPL license.
30 #include "libavutil/x86_cpu.h"
31 #include "libavutil/bswap.h"
34 #include "swscale_internal.h"
36 #define FAST_BGR2YV12 // use 7-bit instead of 15-bit coefficients
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 (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
51 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
52 void (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
53 void (*rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size);
54 void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
55 //void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
56 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
57 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
59 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
60 long width, long height,
61 long lumStride, long chromStride, long dstStride);
62 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
63 long width, long height,
64 long lumStride, long chromStride, long dstStride);
65 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
66 long width, long height,
67 long lumStride, long chromStride, long dstStride);
68 void (*yuv422ptouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
69 long width, long height,
70 long lumStride, long chromStride, long dstStride);
71 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
72 long width, long height,
73 long lumStride, long chromStride, long srcStride);
74 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
75 long width, long height,
76 long lumStride, long chromStride, long srcStride);
77 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
78 long srcStride, long dstStride);
79 void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
80 long width, long height, long src1Stride,
81 long src2Stride, long dstStride);
82 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
83 uint8_t *dst1, uint8_t *dst2,
84 long width, long height,
85 long srcStride1, long srcStride2,
86 long dstStride1, long dstStride2);
87 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
89 long width, long height,
90 long srcStride1, long srcStride2,
91 long srcStride3, long dstStride);
93 #if defined(ARCH_X86) && defined(CONFIG_GPL)
94 DECLARE_ASM_CONST(8, uint64_t, mmx_null) = 0x0000000000000000ULL;
95 DECLARE_ASM_CONST(8, uint64_t, mmx_one) = 0xFFFFFFFFFFFFFFFFULL;
96 DECLARE_ASM_CONST(8, uint64_t, mask32b) = 0x000000FF000000FFULL;
97 DECLARE_ASM_CONST(8, uint64_t, mask32g) = 0x0000FF000000FF00ULL;
98 DECLARE_ASM_CONST(8, uint64_t, mask32r) = 0x00FF000000FF0000ULL;
99 DECLARE_ASM_CONST(8, uint64_t, mask32) = 0x00FFFFFF00FFFFFFULL;
100 DECLARE_ASM_CONST(8, uint64_t, mask3216br) = 0x00F800F800F800F8ULL;
101 DECLARE_ASM_CONST(8, uint64_t, mask3216g) = 0x0000FC000000FC00ULL;
102 DECLARE_ASM_CONST(8, uint64_t, mask3215g) = 0x0000F8000000F800ULL;
103 DECLARE_ASM_CONST(8, uint64_t, mul3216) = 0x2000000420000004ULL;
104 DECLARE_ASM_CONST(8, uint64_t, mul3215) = 0x2000000820000008ULL;
105 DECLARE_ASM_CONST(8, uint64_t, mask24b) = 0x00FF0000FF0000FFULL;
106 DECLARE_ASM_CONST(8, uint64_t, mask24g) = 0xFF0000FF0000FF00ULL;
107 DECLARE_ASM_CONST(8, uint64_t, mask24r) = 0x0000FF0000FF0000ULL;
108 DECLARE_ASM_CONST(8, uint64_t, mask24l) = 0x0000000000FFFFFFULL;
109 DECLARE_ASM_CONST(8, uint64_t, mask24h) = 0x0000FFFFFF000000ULL;
110 DECLARE_ASM_CONST(8, uint64_t, mask24hh) = 0xffff000000000000ULL;
111 DECLARE_ASM_CONST(8, uint64_t, mask24hhh) = 0xffffffff00000000ULL;
112 DECLARE_ASM_CONST(8, uint64_t, mask24hhhh) = 0xffffffffffff0000ULL;
113 DECLARE_ASM_CONST(8, uint64_t, mask15b) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */
114 DECLARE_ASM_CONST(8, uint64_t, mask15rg) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */
115 DECLARE_ASM_CONST(8, uint64_t, mask15s) = 0xFFE0FFE0FFE0FFE0ULL;
116 DECLARE_ASM_CONST(8, uint64_t, mask15g) = 0x03E003E003E003E0ULL;
117 DECLARE_ASM_CONST(8, uint64_t, mask15r) = 0x7C007C007C007C00ULL;
118 #define mask16b mask15b
119 DECLARE_ASM_CONST(8, uint64_t, mask16g) = 0x07E007E007E007E0ULL;
120 DECLARE_ASM_CONST(8, uint64_t, mask16r) = 0xF800F800F800F800ULL;
121 DECLARE_ASM_CONST(8, uint64_t, red_16mask) = 0x0000f8000000f800ULL;
122 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
123 DECLARE_ASM_CONST(8, uint64_t, blue_16mask) = 0x0000001f0000001fULL;
124 DECLARE_ASM_CONST(8, uint64_t, red_15mask) = 0x00007c0000007c00ULL;
125 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
126 DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;
129 static volatile uint64_t __attribute__((aligned(8))) b5Dither;
130 static volatile uint64_t __attribute__((aligned(8))) g5Dither;
131 static volatile uint64_t __attribute__((aligned(8))) g6Dither;
132 static volatile uint64_t __attribute__((aligned(8))) r5Dither;
134 static uint64_t __attribute__((aligned(8))) dither4[2]={
135 0x0103010301030103LL,
136 0x0200020002000200LL,};
138 static uint64_t __attribute__((aligned(8))) dither8[2]={
139 0x0602060206020602LL,
140 0x0004000400040004LL,};
142 #endif /* defined(ARCH_X86) */
144 #define RGB2YUV_SHIFT 8
145 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
146 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
147 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
148 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
149 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
150 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
151 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
152 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
153 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
155 //Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
161 #define RENAME(a) a ## _C
162 #include "rgb2rgb_template.c"
164 #if defined(ARCH_X86) && defined(CONFIG_GPL)
172 #define RENAME(a) a ## _MMX
173 #include "rgb2rgb_template.c"
181 #define RENAME(a) a ## _MMX2
182 #include "rgb2rgb_template.c"
190 #define RENAME(a) a ## _3DNOW
191 #include "rgb2rgb_template.c"
193 #endif //ARCH_X86 || ARCH_X86_64
196 RGB15->RGB16 original by Strepto/Astral
197 ported to gcc & bugfixed : A'rpi
198 MMX2, 3DNOW optimization by Nick Kurshev
199 32-bit C version, and and&add trick by Michael Niedermayer
202 void sws_rgb2rgb_init(int flags){
203 #if (defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX)) && defined(CONFIG_GPL)
204 if (flags & SWS_CPU_CAPS_MMX2)
206 else if (flags & SWS_CPU_CAPS_3DNOW)
207 rgb2rgb_init_3DNOW();
208 else if (flags & SWS_CPU_CAPS_MMX)
211 #endif /* defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX) */
216 * Palette is assumed to contain BGR32.
218 void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
223 for (i=0; i<num_pixels; i++)
224 ((unsigned *)dst)[i] = ((unsigned *)palette)[src[i]];
227 for (i=0; i<num_pixels; i++)
229 #ifdef WORDS_BIGENDIAN
230 dst[3]= palette[src[i]*4+2];
231 dst[2]= palette[src[i]*4+1];
232 dst[1]= palette[src[i]*4+0];
235 dst[0]= palette[src[i]*4+2];
236 dst[1]= palette[src[i]*4+1];
237 dst[2]= palette[src[i]*4+0];
238 //dst[3]= 0; /* do we need this cleansing? */
244 void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
247 for (i=0; i<num_pixels; i++)
249 #ifdef WORDS_BIGENDIAN
250 dst[3]= palette[src[i]*4+0];
251 dst[2]= palette[src[i]*4+1];
252 dst[1]= palette[src[i]*4+2];
255 dst[0]= palette[src[i]*4+0];
256 dst[1]= palette[src[i]*4+1];
257 dst[2]= palette[src[i]*4+2];
258 //dst[3]= 0; /* do we need this cleansing? */
266 * Palette is assumed to contain BGR32.
268 void palette8torgb24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
272 Writes 1 byte too much and might cause alignment issues on some architectures?
273 for (i=0; i<num_pixels; i++)
274 ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[src[i]];
276 for (i=0; i<num_pixels; i++)
279 dst[0]= palette[src[i]*4+2];
280 dst[1]= palette[src[i]*4+1];
281 dst[2]= palette[src[i]*4+0];
286 void palette8tobgr24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
290 Writes 1 byte too much and might cause alignment issues on some architectures?
291 for (i=0; i<num_pixels; i++)
292 ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[src[i]];
294 for (i=0; i<num_pixels; i++)
297 dst[0]= palette[src[i]*4+0];
298 dst[1]= palette[src[i]*4+1];
299 dst[2]= palette[src[i]*4+2];
305 * Palette is assumed to contain BGR16, see rgb32to16 to convert the palette.
307 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
310 for (i=0; i<num_pixels; i++)
311 ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
313 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
316 for (i=0; i<num_pixels; i++)
317 ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
321 * Palette is assumed to contain BGR15, see rgb32to15 to convert the palette.
323 void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
326 for (i=0; i<num_pixels; i++)
327 ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
329 void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
332 for (i=0; i<num_pixels; i++)
333 ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
336 void rgb32to24(const uint8_t *src, uint8_t *dst, long src_size)
339 long num_pixels = src_size >> 2;
340 for (i=0; i<num_pixels; i++)
342 #ifdef WORDS_BIGENDIAN
343 /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
344 dst[3*i + 0] = src[4*i + 1];
345 dst[3*i + 1] = src[4*i + 2];
346 dst[3*i + 2] = src[4*i + 3];
348 dst[3*i + 0] = src[4*i + 2];
349 dst[3*i + 1] = src[4*i + 1];
350 dst[3*i + 2] = src[4*i + 0];
355 void rgb24to32(const uint8_t *src, uint8_t *dst, long src_size)
358 for (i=0; 3*i<src_size; i++)
360 #ifdef WORDS_BIGENDIAN
361 /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
363 dst[4*i + 1] = src[3*i + 0];
364 dst[4*i + 2] = src[3*i + 1];
365 dst[4*i + 3] = src[3*i + 2];
367 dst[4*i + 0] = src[3*i + 2];
368 dst[4*i + 1] = src[3*i + 1];
369 dst[4*i + 2] = src[3*i + 0];
375 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
379 const uint16_t *s = (const uint16_t *)src;
380 end = s + src_size/2;
383 register uint16_t bgr;
385 #ifdef WORDS_BIGENDIAN
387 *d++ = (bgr&0x1F)<<3;
388 *d++ = (bgr&0x7E0)>>3;
389 *d++ = (bgr&0xF800)>>8;
391 *d++ = (bgr&0xF800)>>8;
392 *d++ = (bgr&0x7E0)>>3;
393 *d++ = (bgr&0x1F)<<3;
399 void rgb16to24(const uint8_t *src, uint8_t *dst, long src_size)
403 const uint16_t *s = (const uint16_t *)src;
404 end = s + src_size/2;
407 register uint16_t bgr;
409 *d++ = (bgr&0xF800)>>8;
410 *d++ = (bgr&0x7E0)>>3;
411 *d++ = (bgr&0x1F)<<3;
415 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
418 long num_pixels = src_size >> 1;
420 for (i=0; i<num_pixels; i++)
422 unsigned rgb = ((const uint16_t*)src)[i];
423 ((uint16_t*)dst)[i] = (rgb>>11) | (rgb&0x7E0) | (rgb<<11);
427 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
430 long num_pixels = src_size >> 1;
432 for (i=0; i<num_pixels; i++)
434 unsigned rgb = ((const uint16_t*)src)[i];
435 ((uint16_t*)dst)[i] = (rgb>>11) | ((rgb&0x7C0)>>1) | ((rgb&0x1F)<<10);
439 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
443 const uint16_t *s = (const uint16_t *)src;
444 end = s + src_size/2;
447 register uint16_t bgr;
449 #ifdef WORDS_BIGENDIAN
451 *d++ = (bgr&0x1F)<<3;
452 *d++ = (bgr&0x3E0)>>2;
453 *d++ = (bgr&0x7C00)>>7;
455 *d++ = (bgr&0x7C00)>>7;
456 *d++ = (bgr&0x3E0)>>2;
457 *d++ = (bgr&0x1F)<<3;
463 void rgb15to24(const uint8_t *src, uint8_t *dst, long src_size)
467 const uint16_t *s = (const uint16_t *)src;
468 end = s + src_size/2;
471 register uint16_t bgr;
473 *d++ = (bgr&0x7C00)>>7;
474 *d++ = (bgr&0x3E0)>>2;
475 *d++ = (bgr&0x1F)<<3;
479 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
482 long num_pixels = src_size >> 1;
484 for (i=0; i<num_pixels; i++)
486 unsigned rgb = ((const uint16_t*)src)[i];
487 ((uint16_t*)dst)[i] = ((rgb&0x7C00)>>10) | ((rgb&0x3E0)<<1) | (rgb<<11);
491 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
494 long num_pixels = src_size >> 1;
496 for (i=0; i<num_pixels; i++)
499 unsigned rgb = ((const uint16_t*)src)[i];
501 ((uint16_t*)dst)[i] = (br>>10) | (rgb&0x3E0) | (br<<10);
505 void bgr8torgb8(const uint8_t *src, uint8_t *dst, long src_size)
508 long num_pixels = src_size;
509 for (i=0; i<num_pixels; i++)
512 register uint8_t rgb;
517 dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);