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 * lot of big-endian byte order fixes by Alex Beregszaszi
10 * This file is part of FFmpeg.
12 * FFmpeg is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
15 * version 2.1 of the License, or (at your option) any later version.
17 * FFmpeg is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with FFmpeg; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 static inline void rgb24tobgr32_c(const uint8_t *src, uint8_t *dst, long src_size)
32 const uint8_t *s = src;
38 /* RGB24 (= R,G,B) -> RGB32 (= A,B,G,R) */
53 static inline void rgb32tobgr24_c(const uint8_t *src, uint8_t *dst, long src_size)
56 const uint8_t *s = src;
63 /* RGB32 (= A,B,G,R) -> RGB24 (= R,G,B) */
79 original by Strepto/Astral
80 ported to gcc & bugfixed: A'rpi
81 MMX2, 3DNOW optimization by Nick Kurshev
82 32-bit C version, and and&add trick by Michael Niedermayer
84 static inline void rgb15to16_c(const uint8_t *src, uint8_t *dst, long src_size)
86 register const uint8_t* s=src;
87 register uint8_t* d=dst;
88 register const uint8_t *end;
89 const uint8_t *mm_end;
93 register unsigned x= *((const uint32_t *)s);
94 *((uint32_t *)d) = (x&0x7FFF7FFF) + (x&0x7FE07FE0);
99 register unsigned short x= *((const uint16_t *)s);
100 *((uint16_t *)d) = (x&0x7FFF) + (x&0x7FE0);
104 static inline void rgb16to15_c(const uint8_t *src, uint8_t *dst, long src_size)
106 register const uint8_t* s=src;
107 register uint8_t* d=dst;
108 register const uint8_t *end;
109 const uint8_t *mm_end;
114 register uint32_t x= *((const uint32_t*)s);
115 *((uint32_t *)d) = ((x>>1)&0x7FE07FE0) | (x&0x001F001F);
120 register uint16_t x= *((const uint16_t*)s);
121 *((uint16_t *)d) = ((x>>1)&0x7FE0) | (x&0x001F);
125 static inline void rgb32to16_c(const uint8_t *src, uint8_t *dst, long src_size)
127 const uint8_t *s = src;
129 uint16_t *d = (uint16_t *)dst;
133 register int rgb = *(const uint32_t*)s; s += 4;
134 *d++ = ((rgb&0xFF)>>3) + ((rgb&0xFC00)>>5) + ((rgb&0xF80000)>>8);
138 static inline void rgb32tobgr16_c(const uint8_t *src, uint8_t *dst, long src_size)
140 const uint8_t *s = src;
142 uint16_t *d = (uint16_t *)dst;
145 register int rgb = *(const uint32_t*)s; s += 4;
146 *d++ = ((rgb&0xF8)<<8) + ((rgb&0xFC00)>>5) + ((rgb&0xF80000)>>19);
150 static inline void rgb32to15_c(const uint8_t *src, uint8_t *dst, long src_size)
152 const uint8_t *s = src;
154 uint16_t *d = (uint16_t *)dst;
157 register int rgb = *(const uint32_t*)s; s += 4;
158 *d++ = ((rgb&0xFF)>>3) + ((rgb&0xF800)>>6) + ((rgb&0xF80000)>>9);
162 static inline void rgb32tobgr15_c(const uint8_t *src, uint8_t *dst, long src_size)
164 const uint8_t *s = src;
166 uint16_t *d = (uint16_t *)dst;
169 register int rgb = *(const uint32_t*)s; s += 4;
170 *d++ = ((rgb&0xF8)<<7) + ((rgb&0xF800)>>6) + ((rgb&0xF80000)>>19);
174 static inline void rgb24tobgr16_c(const uint8_t *src, uint8_t *dst, long src_size)
176 const uint8_t *s = src;
178 uint16_t *d = (uint16_t *)dst;
184 *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);
188 static inline void rgb24to16_c(const uint8_t *src, uint8_t *dst, long src_size)
190 const uint8_t *s = src;
192 uint16_t *d = (uint16_t *)dst;
198 *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);
202 static inline void rgb24tobgr15_c(const uint8_t *src, uint8_t *dst, long src_size)
204 const uint8_t *s = src;
206 uint16_t *d = (uint16_t *)dst;
212 *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
216 static inline void rgb24to15_c(const uint8_t *src, uint8_t *dst, long src_size)
218 const uint8_t *s = src;
220 uint16_t *d = (uint16_t *)dst;
226 *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
231 I use less accurate approximation here by simply left-shifting the input
232 value and filling the low order bits with zeroes. This method improves PNG
233 compression but this scheme cannot reproduce white exactly, since it does
234 not generate an all-ones maximum value; the net effect is to darken the
237 The better method should be "left bit replication":
247 | leftmost bits repeated to fill open bits
251 static inline void rgb15tobgr24_c(const uint8_t *src, uint8_t *dst, long src_size)
255 const uint16_t *s = (const uint16_t*)src;
256 end = s + src_size/2;
258 register uint16_t bgr;
260 *d++ = (bgr&0x1F)<<3;
261 *d++ = (bgr&0x3E0)>>2;
262 *d++ = (bgr&0x7C00)>>7;
266 static inline void rgb16tobgr24_c(const uint8_t *src, uint8_t *dst, long src_size)
269 uint8_t *d = (uint8_t *)dst;
270 const uint16_t *s = (const uint16_t *)src;
271 end = s + src_size/2;
273 register uint16_t bgr;
275 *d++ = (bgr&0x1F)<<3;
276 *d++ = (bgr&0x7E0)>>3;
277 *d++ = (bgr&0xF800)>>8;
281 static inline void rgb15to32_c(const uint8_t *src, uint8_t *dst, long src_size)
285 const uint16_t *s = (const uint16_t *)src;
286 end = s + src_size/2;
288 register uint16_t bgr;
292 *d++ = (bgr&0x7C00)>>7;
293 *d++ = (bgr&0x3E0)>>2;
294 *d++ = (bgr&0x1F)<<3;
296 *d++ = (bgr&0x1F)<<3;
297 *d++ = (bgr&0x3E0)>>2;
298 *d++ = (bgr&0x7C00)>>7;
304 static inline void rgb16to32_c(const uint8_t *src, uint8_t *dst, long src_size)
308 const uint16_t *s = (const uint16_t*)src;
309 end = s + src_size/2;
311 register uint16_t bgr;
315 *d++ = (bgr&0xF800)>>8;
316 *d++ = (bgr&0x7E0)>>3;
317 *d++ = (bgr&0x1F)<<3;
319 *d++ = (bgr&0x1F)<<3;
320 *d++ = (bgr&0x7E0)>>3;
321 *d++ = (bgr&0xF800)>>8;
327 static inline void shuffle_bytes_2103_c(const uint8_t *src, uint8_t *dst, long src_size)
329 int idx = 15 - src_size;
330 const uint8_t *s = src-idx;
331 uint8_t *d = dst-idx;
332 for (; idx<15; idx+=4) {
333 register int v = *(const uint32_t *)&s[idx], g = v & 0xff00ff00;
335 *(uint32_t *)&d[idx] = (v>>16) + g + (v<<16);
339 static inline void rgb24tobgr24_c(const uint8_t *src, uint8_t *dst, long src_size)
342 for (i=0; i<src_size; i+=3) {
345 dst[i + 1] = src[i + 1];
346 dst[i + 2] = src[i + 0];
351 static inline void yuvPlanartoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
352 const uint8_t *vsrc, uint8_t *dst,
353 long width, long height,
354 long lumStride, long chromStride,
355 long dstStride, long vertLumPerChroma)
358 const int chromWidth = width >> 1;
359 for (y=0; y<height; y++) {
362 uint64_t *ldst = (uint64_t *) dst;
363 const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
364 for (i = 0; i < chromWidth; i += 2) {
366 k = yc[0] + (uc[0] << 8) +
367 (yc[1] << 16) + (vc[0] << 24);
368 l = yc[2] + (uc[1] << 8) +
369 (yc[3] << 16) + (vc[1] << 24);
370 *ldst++ = k + (l << 32);
377 int i, *idst = (int32_t *) dst;
378 const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
379 for (i = 0; i < chromWidth; i++) {
381 *idst++ = (yc[0] << 24)+ (uc[0] << 16) +
382 (yc[1] << 8) + (vc[0] << 0);
384 *idst++ = yc[0] + (uc[0] << 8) +
385 (yc[1] << 16) + (vc[0] << 24);
392 if ((y&(vertLumPerChroma-1)) == vertLumPerChroma-1) {
402 * Height should be a multiple of 2 and width should be a multiple of 16.
403 * (If this is a problem for anyone then tell me, and I will fix it.)
405 static inline void yv12toyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
406 const uint8_t *vsrc, uint8_t *dst,
407 long width, long height,
408 long lumStride, long chromStride,
411 //FIXME interpolate chroma
412 yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
413 chromStride, dstStride, 2);
416 static inline void yuvPlanartouyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
417 const uint8_t *vsrc, uint8_t *dst,
418 long width, long height,
419 long lumStride, long chromStride,
420 long dstStride, long vertLumPerChroma)
423 const int chromWidth = width >> 1;
424 for (y=0; y<height; y++) {
427 uint64_t *ldst = (uint64_t *) dst;
428 const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
429 for (i = 0; i < chromWidth; i += 2) {
431 k = uc[0] + (yc[0] << 8) +
432 (vc[0] << 16) + (yc[1] << 24);
433 l = uc[1] + (yc[2] << 8) +
434 (vc[1] << 16) + (yc[3] << 24);
435 *ldst++ = k + (l << 32);
442 int i, *idst = (int32_t *) dst;
443 const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
444 for (i = 0; i < chromWidth; i++) {
446 *idst++ = (uc[0] << 24)+ (yc[0] << 16) +
447 (vc[0] << 8) + (yc[1] << 0);
449 *idst++ = uc[0] + (yc[0] << 8) +
450 (vc[0] << 16) + (yc[1] << 24);
457 if ((y&(vertLumPerChroma-1)) == vertLumPerChroma-1) {
467 * Height should be a multiple of 2 and width should be a multiple of 16
468 * (If this is a problem for anyone then tell me, and I will fix it.)
470 static inline void yv12touyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
471 const uint8_t *vsrc, uint8_t *dst,
472 long width, long height,
473 long lumStride, long chromStride,
476 //FIXME interpolate chroma
477 yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
478 chromStride, dstStride, 2);
482 * Width should be a multiple of 16.
484 static inline void yuv422ptouyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
485 const uint8_t *vsrc, uint8_t *dst,
486 long width, long height,
487 long lumStride, long chromStride,
490 yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
491 chromStride, dstStride, 1);
495 * Width should be a multiple of 16.
497 static inline void yuv422ptoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
498 const uint8_t *vsrc, uint8_t *dst,
499 long width, long height,
500 long lumStride, long chromStride,
503 yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
504 chromStride, dstStride, 1);
508 * Height should be a multiple of 2 and width should be a multiple of 16.
509 * (If this is a problem for anyone then tell me, and I will fix it.)
511 static inline void yuy2toyv12_c(const uint8_t *src, uint8_t *ydst,
512 uint8_t *udst, uint8_t *vdst,
513 long width, long height,
514 long lumStride, long chromStride,
518 const int chromWidth = width >> 1;
519 for (y=0; y<height; y+=2) {
521 for (i=0; i<chromWidth; i++) {
522 ydst[2*i+0] = src[4*i+0];
523 udst[i] = src[4*i+1];
524 ydst[2*i+1] = src[4*i+2];
525 vdst[i] = src[4*i+3];
530 for (i=0; i<chromWidth; i++) {
531 ydst[2*i+0] = src[4*i+0];
532 ydst[2*i+1] = src[4*i+2];
541 static inline void planar2x_c(const uint8_t *src, uint8_t *dst, long srcWidth,
542 long srcHeight, long srcStride, long dstStride)
549 for (x=0; x<srcWidth-1; x++) {
550 dst[2*x+1]= (3*src[x] + src[x+1])>>2;
551 dst[2*x+2]= ( src[x] + 3*src[x+1])>>2;
553 dst[2*srcWidth-1]= src[srcWidth-1];
557 for (y=1; y<srcHeight; y++) {
558 const int mmxSize = 1;
560 dst[0 ]= (3*src[0] + src[srcStride])>>2;
561 dst[dstStride]= ( src[0] + 3*src[srcStride])>>2;
563 for (x=mmxSize-1; x<srcWidth-1; x++) {
564 dst[2*x +1]= (3*src[x+0] + src[x+srcStride+1])>>2;
565 dst[2*x+dstStride+2]= ( src[x+0] + 3*src[x+srcStride+1])>>2;
566 dst[2*x+dstStride+1]= ( src[x+1] + 3*src[x+srcStride ])>>2;
567 dst[2*x +2]= (3*src[x+1] + src[x+srcStride ])>>2;
569 dst[srcWidth*2 -1 ]= (3*src[srcWidth-1] + src[srcWidth-1 + srcStride])>>2;
570 dst[srcWidth*2 -1 + dstStride]= ( src[srcWidth-1] + 3*src[srcWidth-1 + srcStride])>>2;
580 for (x=0; x<srcWidth-1; x++) {
581 dst[2*x+1]= (3*src[x] + src[x+1])>>2;
582 dst[2*x+2]= ( src[x] + 3*src[x+1])>>2;
584 dst[2*srcWidth-1]= src[srcWidth-1];
586 for (x=0; x<srcWidth; x++) {
594 * Height should be a multiple of 2 and width should be a multiple of 16.
595 * (If this is a problem for anyone then tell me, and I will fix it.)
596 * Chrominance data is only taken from every second line, others are ignored.
597 * FIXME: Write HQ version.
599 static inline void uyvytoyv12_c(const uint8_t *src, uint8_t *ydst,
600 uint8_t *udst, uint8_t *vdst,
601 long width, long height,
602 long lumStride, long chromStride,
606 const int chromWidth = width >> 1;
607 for (y=0; y<height; y+=2) {
609 for (i=0; i<chromWidth; i++) {
610 udst[i] = src[4*i+0];
611 ydst[2*i+0] = src[4*i+1];
612 vdst[i] = src[4*i+2];
613 ydst[2*i+1] = src[4*i+3];
618 for (i=0; i<chromWidth; i++) {
619 ydst[2*i+0] = src[4*i+1];
620 ydst[2*i+1] = src[4*i+3];
630 * Height should be a multiple of 2 and width should be a multiple of 2.
631 * (If this is a problem for anyone then tell me, and I will fix it.)
632 * Chrominance data is only taken from every second line,
633 * others are ignored in the C version.
634 * FIXME: Write HQ version.
636 void rgb24toyv12_c(const uint8_t *src, uint8_t *ydst, uint8_t *udst,
637 uint8_t *vdst, long width, long height, long lumStride,
638 long chromStride, long srcStride)
641 const int chromWidth = width >> 1;
643 for (; y<height; y+=2) {
645 for (i=0; i<chromWidth; i++) {
646 unsigned int b = src[6*i+0];
647 unsigned int g = src[6*i+1];
648 unsigned int r = src[6*i+2];
650 unsigned int Y = ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;
651 unsigned int V = ((RV*r + GV*g + BV*b)>>RGB2YUV_SHIFT) + 128;
652 unsigned int U = ((RU*r + GU*g + BU*b)>>RGB2YUV_SHIFT) + 128;
662 Y = ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;
671 for (i=0; i<chromWidth; i++) {
672 unsigned int b = src[6*i+0];
673 unsigned int g = src[6*i+1];
674 unsigned int r = src[6*i+2];
676 unsigned int Y = ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;
684 Y = ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;
694 static void interleaveBytes_c(const uint8_t *src1, const uint8_t *src2,
695 uint8_t *dest, long width,
696 long height, long src1Stride,
697 long src2Stride, long dstStride)
701 for (h=0; h < height; h++) {
703 for (w=0; w < width; w++) {
704 dest[2*w+0] = src1[w];
705 dest[2*w+1] = src2[w];
713 static inline void vu9_to_vu12_c(const uint8_t *src1, const uint8_t *src2,
714 uint8_t *dst1, uint8_t *dst2,
715 long width, long height,
716 long srcStride1, long srcStride2,
717 long dstStride1, long dstStride2)
721 w=width/2; h=height/2;
723 const uint8_t* s1=src1+srcStride1*(y>>1);
724 uint8_t* d=dst1+dstStride1*y;
726 for (;x<w;x++) d[2*x]=d[2*x+1]=s1[x];
729 const uint8_t* s2=src2+srcStride2*(y>>1);
730 uint8_t* d=dst2+dstStride2*y;
732 for (;x<w;x++) d[2*x]=d[2*x+1]=s2[x];
736 static inline void yvu9_to_yuy2_c(const uint8_t *src1, const uint8_t *src2,
737 const uint8_t *src3, uint8_t *dst,
738 long width, long height,
739 long srcStride1, long srcStride2,
740 long srcStride3, long dstStride)
746 const uint8_t* yp=src1+srcStride1*y;
747 const uint8_t* up=src2+srcStride2*(y>>2);
748 const uint8_t* vp=src3+srcStride3*(y>>2);
749 uint8_t* d=dst+dstStride*y;
752 const long x2 = x<<2;
765 static void extract_even_c(const uint8_t *src, uint8_t *dst, int count)
772 dst[count]= src[2*count];
777 static void extract_even2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1,
785 dst0[count]= src[4*count+0];
786 dst1[count]= src[4*count+2];
791 static void extract_even2avg_c(const uint8_t *src0, const uint8_t *src1,
792 uint8_t *dst0, uint8_t *dst1, int count)
800 dst0[count]= (src0[4*count+0]+src1[4*count+0])>>1;
801 dst1[count]= (src0[4*count+2]+src1[4*count+2])>>1;
806 static void extract_odd2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1,
815 dst0[count]= src[4*count+0];
816 dst1[count]= src[4*count+2];
821 static void extract_odd2avg_c(const uint8_t *src0, const uint8_t *src1,
822 uint8_t *dst0, uint8_t *dst1, int count)
832 dst0[count]= (src0[4*count+0]+src1[4*count+0])>>1;
833 dst1[count]= (src0[4*count+2]+src1[4*count+2])>>1;
838 static void yuyvtoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
839 const uint8_t *src, long width, long height,
840 long lumStride, long chromStride, long srcStride)
843 const long chromWidth= -((-width)>>1);
845 for (y=0; y<height; y++) {
846 extract_even_c(src, ydst, width);
848 extract_odd2avg_c(src - srcStride, src, udst, vdst, chromWidth);
858 static void yuyvtoyuv422_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
859 const uint8_t *src, long width, long height,
860 long lumStride, long chromStride, long srcStride)
863 const long chromWidth= -((-width)>>1);
865 for (y=0; y<height; y++) {
866 extract_even_c(src, ydst, width);
867 extract_odd2_c(src, udst, vdst, chromWidth);
876 static void uyvytoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
877 const uint8_t *src, long width, long height,
878 long lumStride, long chromStride, long srcStride)
881 const long chromWidth= -((-width)>>1);
883 for (y=0; y<height; y++) {
884 extract_even_c(src + 1, ydst, width);
886 extract_even2avg_c(src - srcStride, src, udst, vdst, chromWidth);
896 static void uyvytoyuv422_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
897 const uint8_t *src, long width, long height,
898 long lumStride, long chromStride, long srcStride)
901 const long chromWidth= -((-width)>>1);
903 for (y=0; y<height; y++) {
904 extract_even_c(src + 1, ydst, width);
905 extract_even2_c(src, udst, vdst, chromWidth);
914 static inline void rgb2rgb_init_c(void)
916 rgb15to16 = rgb15to16_c;
917 rgb15tobgr24 = rgb15tobgr24_c;
918 rgb15to32 = rgb15to32_c;
919 rgb16tobgr24 = rgb16tobgr24_c;
920 rgb16to32 = rgb16to32_c;
921 rgb16to15 = rgb16to15_c;
922 rgb24tobgr16 = rgb24tobgr16_c;
923 rgb24tobgr15 = rgb24tobgr15_c;
924 rgb24tobgr32 = rgb24tobgr32_c;
925 rgb32to16 = rgb32to16_c;
926 rgb32to15 = rgb32to15_c;
927 rgb32tobgr24 = rgb32tobgr24_c;
928 rgb24to15 = rgb24to15_c;
929 rgb24to16 = rgb24to16_c;
930 rgb24tobgr24 = rgb24tobgr24_c;
931 shuffle_bytes_2103 = shuffle_bytes_2103_c;
932 rgb32tobgr16 = rgb32tobgr16_c;
933 rgb32tobgr15 = rgb32tobgr15_c;
934 yv12toyuy2 = yv12toyuy2_c;
935 yv12touyvy = yv12touyvy_c;
936 yuv422ptoyuy2 = yuv422ptoyuy2_c;
937 yuv422ptouyvy = yuv422ptouyvy_c;
938 yuy2toyv12 = yuy2toyv12_c;
939 planar2x = planar2x_c;
940 rgb24toyv12 = rgb24toyv12_c;
941 interleaveBytes = interleaveBytes_c;
942 vu9_to_vu12 = vu9_to_vu12_c;
943 yvu9_to_yuy2 = yvu9_to_yuy2_c;
945 uyvytoyuv420 = uyvytoyuv420_c;
946 uyvytoyuv422 = uyvytoyuv422_c;
947 yuyvtoyuv420 = yuyvtoyuv420_c;
948 yuyvtoyuv422 = yuyvtoyuv422_c;