2 * yuv2rgb.c, Software YUV to RGB coverter
4 * Copyright (C) 1999, Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
7 * Functions broken out from display_x11.c and several new modes
8 * added by HÃ¥kan Hjort <d95hjort@dtek.chalmers.se>
10 * 15 & 16 bpp support by Franck Sicard <Franck.Sicard@solsoft.fr>
12 * This file is part of mpeg2dec, a free MPEG-2 video decoder
14 * mpeg2dec is free software; you can redistribute it and/or modify
15 * it under the terms of the GNU General Public License as published by
16 * the Free Software Foundation; either version 2, or (at your option)
19 * mpeg2dec is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22 * GNU General Public License for more details.
24 * You should have received a copy of the GNU General Public License
25 * along with GNU Make; see the file COPYING. If not, write to
26 * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
28 * MMX/MMX2 Template stuff from Michael Niedermayer (michaelni@gmx.at) (needed for fast movntq support)
29 * 1,4,8bpp support by Michael Niedermayer (michaelni@gmx.at)
30 * context / deglobalize stuff by Michael Niedermayer
39 //#include "video_out.h"
42 #include "swscale_internal.h"
46 #include "yuv2rgb_mlib.c"
49 #define DITHER1XBPP // only for mmx
51 const uint8_t __attribute__((aligned(8))) dither_2x2_4[2][8]={
52 { 1, 3, 1, 3, 1, 3, 1, 3, },
53 { 2, 0, 2, 0, 2, 0, 2, 0, },
56 const uint8_t __attribute__((aligned(8))) dither_2x2_8[2][8]={
57 { 6, 2, 6, 2, 6, 2, 6, 2, },
58 { 0, 4, 0, 4, 0, 4, 0, 4, },
61 const uint8_t __attribute__((aligned(8))) dither_8x8_32[8][8]={
62 { 17, 9, 23, 15, 16, 8, 22, 14, },
63 { 5, 29, 3, 27, 4, 28, 2, 26, },
64 { 21, 13, 19, 11, 20, 12, 18, 10, },
65 { 0, 24, 6, 30, 1, 25, 7, 31, },
66 { 16, 8, 22, 14, 17, 9, 23, 15, },
67 { 4, 28, 2, 26, 5, 29, 3, 27, },
68 { 20, 12, 18, 10, 21, 13, 19, 11, },
69 { 1, 25, 7, 31, 0, 24, 6, 30, },
73 const uint8_t __attribute__((aligned(8))) dither_8x8_64[8][8]={
74 { 0, 48, 12, 60, 3, 51, 15, 63, },
75 { 32, 16, 44, 28, 35, 19, 47, 31, },
76 { 8, 56, 4, 52, 11, 59, 7, 55, },
77 { 40, 24, 36, 20, 43, 27, 39, 23, },
78 { 2, 50, 14, 62, 1, 49, 13, 61, },
79 { 34, 18, 46, 30, 33, 17, 45, 29, },
80 { 10, 58, 6, 54, 9, 57, 5, 53, },
81 { 42, 26, 38, 22, 41, 25, 37, 21, },
85 const uint8_t __attribute__((aligned(8))) dither_8x8_73[8][8]={
86 { 0, 55, 14, 68, 3, 58, 17, 72, },
87 { 37, 18, 50, 32, 40, 22, 54, 35, },
88 { 9, 64, 5, 59, 13, 67, 8, 63, },
89 { 46, 27, 41, 23, 49, 31, 44, 26, },
90 { 2, 57, 16, 71, 1, 56, 15, 70, },
91 { 39, 21, 52, 34, 38, 19, 51, 33, },
92 { 11, 66, 7, 62, 10, 65, 6, 60, },
93 { 48, 30, 43, 25, 47, 29, 42, 24, },
97 const uint8_t __attribute__((aligned(8))) dither_8x8_128[8][8]={
98 { 68, 36, 92, 60, 66, 34, 90, 58, },
99 { 20, 116, 12, 108, 18, 114, 10, 106, },
100 { 84, 52, 76, 44, 82, 50, 74, 42, },
101 { 0, 96, 24, 120, 6, 102, 30, 126, },
102 { 64, 32, 88, 56, 70, 38, 94, 62, },
103 { 16, 112, 8, 104, 22, 118, 14, 110, },
104 { 80, 48, 72, 40, 86, 54, 78, 46, },
105 { 4, 100, 28, 124, 2, 98, 26, 122, },
110 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
111 {117, 62, 158, 103, 113, 58, 155, 100, },
112 { 34, 199, 21, 186, 31, 196, 17, 182, },
113 {144, 89, 131, 76, 141, 86, 127, 72, },
114 { 0, 165, 41, 206, 10, 175, 52, 217, },
115 {110, 55, 151, 96, 120, 65, 162, 107, },
116 { 28, 193, 14, 179, 38, 203, 24, 189, },
117 {138, 83, 124, 69, 148, 93, 134, 79, },
118 { 7, 172, 48, 213, 3, 168, 45, 210, },
121 // tries to correct a gamma of 1.5
122 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
123 { 0, 143, 18, 200, 2, 156, 25, 215, },
124 { 78, 28, 125, 64, 89, 36, 138, 74, },
125 { 10, 180, 3, 161, 16, 195, 8, 175, },
126 {109, 51, 93, 38, 121, 60, 105, 47, },
127 { 1, 152, 23, 210, 0, 147, 20, 205, },
128 { 85, 33, 134, 71, 81, 30, 130, 67, },
129 { 14, 190, 6, 171, 12, 185, 5, 166, },
130 {117, 57, 101, 44, 113, 54, 97, 41, },
133 // tries to correct a gamma of 2.0
134 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
135 { 0, 124, 8, 193, 0, 140, 12, 213, },
136 { 55, 14, 104, 42, 66, 19, 119, 52, },
137 { 3, 168, 1, 145, 6, 187, 3, 162, },
138 { 86, 31, 70, 21, 99, 39, 82, 28, },
139 { 0, 134, 11, 206, 0, 129, 9, 200, },
140 { 62, 17, 114, 48, 58, 16, 109, 45, },
141 { 5, 181, 2, 157, 4, 175, 1, 151, },
142 { 95, 36, 78, 26, 90, 34, 74, 24, },
145 // tries to correct a gamma of 2.5
146 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
147 { 0, 107, 3, 187, 0, 125, 6, 212, },
148 { 39, 7, 86, 28, 49, 11, 102, 36, },
149 { 1, 158, 0, 131, 3, 180, 1, 151, },
150 { 68, 19, 52, 12, 81, 25, 64, 17, },
151 { 0, 119, 5, 203, 0, 113, 4, 195, },
152 { 45, 9, 96, 33, 42, 8, 91, 30, },
153 { 2, 172, 1, 144, 2, 165, 0, 137, },
154 { 77, 23, 60, 15, 72, 21, 56, 14, },
160 /* hope these constant values are cache line aligned */
161 uint64_t attribute_used __attribute__((aligned(8))) mmx_00ffw = 0x00ff00ff00ff00ffULL;
162 uint64_t attribute_used __attribute__((aligned(8))) mmx_redmask = 0xf8f8f8f8f8f8f8f8ULL;
163 uint64_t attribute_used __attribute__((aligned(8))) mmx_grnmask = 0xfcfcfcfcfcfcfcfcULL;
165 uint64_t attribute_used __attribute__((aligned(8))) M24A= 0x00FF0000FF0000FFULL;
166 uint64_t attribute_used __attribute__((aligned(8))) M24B= 0xFF0000FF0000FF00ULL;
167 uint64_t attribute_used __attribute__((aligned(8))) M24C= 0x0000FF0000FF0000ULL;
169 // the volatile is required because gcc otherwise optimizes some writes away not knowing that these
170 // are read in the asm block
171 volatile uint64_t attribute_used __attribute__((aligned(8))) b5Dither;
172 volatile uint64_t attribute_used __attribute__((aligned(8))) g5Dither;
173 volatile uint64_t attribute_used __attribute__((aligned(8))) g6Dither;
174 volatile uint64_t attribute_used __attribute__((aligned(8))) r5Dither;
176 uint64_t __attribute__((aligned(8))) dither4[2]={
177 0x0103010301030103LL,
178 0x0200020002000200LL,};
180 uint64_t __attribute__((aligned(8))) dither8[2]={
181 0x0602060206020602LL,
182 0x0004000400040004LL,};
193 #define RENAME(a) a ## _MMX
194 #include "yuv2rgb_template.c"
202 #define RENAME(a) a ## _MMX2
203 #include "yuv2rgb_template.c"
205 #endif // CAN_COMPILE_X86_ASM
207 const int32_t Inverse_Table_6_9[8][4] = {
208 {117504, 138453, 13954, 34903}, /* no sequence_display_extension */
209 {117504, 138453, 13954, 34903}, /* ITU-R Rec. 709 (1990) */
210 {104597, 132201, 25675, 53279}, /* unspecified */
211 {104597, 132201, 25675, 53279}, /* reserved */
212 {104448, 132798, 24759, 53109}, /* FCC */
213 {104597, 132201, 25675, 53279}, /* ITU-R Rec. 624-4 System B, G */
214 {104597, 132201, 25675, 53279}, /* SMPTE 170M */
215 {117579, 136230, 16907, 35559} /* SMPTE 240M (1987) */
221 r = c->table_rV[V]; \
222 g = c->table_gU[U] + c->table_gV[V]; \
227 dst_1[2*i] = r[Y] + g[Y] + b[Y]; \
229 dst_1[2*i+1] = r[Y] + g[Y] + b[Y];
233 dst_2[2*i] = r[Y] + g[Y] + b[Y]; \
235 dst_2[2*i+1] = r[Y] + g[Y] + b[Y];
239 dst_1[6*i] = r[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = b[Y]; \
241 dst_1[6*i+3] = r[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = b[Y];
245 dst_2[6*i] = r[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = b[Y]; \
247 dst_2[6*i+3] = r[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = b[Y];
251 dst_1[6*i] = b[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = r[Y]; \
253 dst_1[6*i+3] = b[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = r[Y];
257 dst_2[6*i] = b[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = r[Y]; \
259 dst_2[6*i+3] = b[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = r[Y];
261 #define PROLOG(func_name, dst_type) \
262 static int func_name(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, \
263 int srcSliceH, uint8_t* dst[], int dstStride[]){\
266 if(c->srcFormat == IMGFMT_422P){\
270 for(y=0; y<srcSliceH; y+=2){\
271 dst_type *dst_1= (dst_type*)(dst[0] + (y+srcSliceY )*dstStride[0]);\
272 dst_type *dst_2= (dst_type*)(dst[0] + (y+srcSliceY+1)*dstStride[0]);\
273 dst_type *r, *g, *b;\
274 uint8_t *py_1= src[0] + y*srcStride[0];\
275 uint8_t *py_2= py_1 + srcStride[0];\
276 uint8_t *pu= src[1] + (y>>1)*srcStride[1];\
277 uint8_t *pv= src[2] + (y>>1)*srcStride[2];\
278 unsigned int h_size= c->dstW>>3;\
282 #define EPILOG(dst_delta)\
294 PROLOG(yuv2rgb_c_32, uint32_t)
312 PROLOG(yuv2rgb_c_24_rgb, uint8_t)
330 // only trivial mods from yuv2rgb_c_24_rgb
331 PROLOG(yuv2rgb_c_24_bgr, uint8_t)
349 // This is exactly the same code as yuv2rgb_c_32 except for the types of
350 // r, g, b, dst_1, dst_2
351 PROLOG(yuv2rgb_c_16, uint16_t)
369 // This is exactly the same code as yuv2rgb_c_32 except for the types of
370 // r, g, b, dst_1, dst_2
371 PROLOG(yuv2rgb_c_8, uint8_t)
389 // r, g, b, dst_1, dst_2
390 PROLOG(yuv2rgb_c_8_ordered_dither, uint8_t)
391 const uint8_t *d32= dither_8x8_32[y&7];
392 const uint8_t *d64= dither_8x8_73[y&7];
393 #define DST1bpp8(i,o) \
395 dst_1[2*i] = r[Y+d32[0+o]] + g[Y+d32[0+o]] + b[Y+d64[0+o]]; \
397 dst_1[2*i+1] = r[Y+d32[1+o]] + g[Y+d32[1+o]] + b[Y+d64[1+o]];
399 #define DST2bpp8(i,o) \
401 dst_2[2*i] = r[Y+d32[8+o]] + g[Y+d32[8+o]] + b[Y+d64[8+o]]; \
403 dst_2[2*i+1] = r[Y+d32[9+o]] + g[Y+d32[9+o]] + b[Y+d64[9+o]];
424 // This is exactly the same code as yuv2rgb_c_32 except for the types of
425 // r, g, b, dst_1, dst_2
426 PROLOG(yuv2rgb_c_4, uint8_t)
430 acc = r[Y] + g[Y] + b[Y]; \
432 acc |= (r[Y] + g[Y] + b[Y])<<4;\
437 acc = r[Y] + g[Y] + b[Y]; \
439 acc |= (r[Y] + g[Y] + b[Y])<<4;\
459 PROLOG(yuv2rgb_c_4_ordered_dither, uint8_t)
460 const uint8_t *d64= dither_8x8_73[y&7];
461 const uint8_t *d128=dither_8x8_220[y&7];
464 #define DST1bpp4(i,o) \
466 acc = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]]; \
468 acc |= (r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]])<<4;\
471 #define DST2bpp4(i,o) \
473 acc = r[Y+d128[8+o]] + g[Y+d64[8+o]] + b[Y+d128[8+o]]; \
475 acc |= (r[Y+d128[9+o]] + g[Y+d64[9+o]] + b[Y+d128[9+o]])<<4;\
496 // This is exactly the same code as yuv2rgb_c_32 except for the types of
497 // r, g, b, dst_1, dst_2
498 PROLOG(yuv2rgb_c_4b, uint8_t)
516 PROLOG(yuv2rgb_c_4b_ordered_dither, uint8_t)
517 const uint8_t *d64= dither_8x8_73[y&7];
518 const uint8_t *d128=dither_8x8_220[y&7];
520 #define DST1bpp4b(i,o) \
522 dst_1[2*i] = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]]; \
524 dst_1[2*i+1] = r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]];
526 #define DST2bpp4b(i,o) \
528 dst_2[2*i] = r[Y+d128[8+o]] + g[Y+d64[8+o]] + b[Y+d128[8+o]]; \
530 dst_2[2*i+1] = r[Y+d128[9+o]] + g[Y+d64[9+o]] + b[Y+d128[9+o]];
550 PROLOG(yuv2rgb_c_1_ordered_dither, uint8_t)
551 const uint8_t *d128=dither_8x8_220[y&7];
552 char out_1=0, out_2=0;
553 g= c->table_gU[128] + c->table_gV[128];
555 #define DST1bpp1(i,o) \
557 out_1+= out_1 + g[Y+d128[0+o]]; \
559 out_1+= out_1 + g[Y+d128[1+o]];
561 #define DST2bpp1(i,o) \
563 out_2+= out_2 + g[Y+d128[8+o]]; \
565 out_2+= out_2 + g[Y+d128[9+o]];
583 SwsFunc yuv2rgb_get_func_ptr (SwsContext *c)
586 if(c->flags & SWS_CPU_CAPS_MMX2){
587 switch(c->dstFormat){
588 case IMGFMT_BGR32: return yuv420_rgb32_MMX2;
589 case IMGFMT_BGR24: return yuv420_rgb24_MMX2;
590 case IMGFMT_BGR16: return yuv420_rgb16_MMX2;
591 case IMGFMT_BGR15: return yuv420_rgb15_MMX2;
594 if(c->flags & SWS_CPU_CAPS_MMX){
595 switch(c->dstFormat){
596 case IMGFMT_BGR32: return yuv420_rgb32_MMX;
597 case IMGFMT_BGR24: return yuv420_rgb24_MMX;
598 case IMGFMT_BGR16: return yuv420_rgb16_MMX;
599 case IMGFMT_BGR15: return yuv420_rgb15_MMX;
605 SwsFunc t= yuv2rgb_init_mlib(c);
610 if (c->flags & SWS_CPU_CAPS_ALTIVEC)
612 SwsFunc t = yuv2rgb_init_altivec(c);
617 MSG_WARN("No accelerated colorspace conversion found\n");
619 switch(c->dstFormat){
621 case IMGFMT_BGR32: return yuv2rgb_c_32;
622 case IMGFMT_RGB24: return yuv2rgb_c_24_rgb;
623 case IMGFMT_BGR24: return yuv2rgb_c_24_bgr;
627 case IMGFMT_BGR15: return yuv2rgb_c_16;
629 case IMGFMT_BGR8: return yuv2rgb_c_8_ordered_dither;
631 case IMGFMT_BGR4: return yuv2rgb_c_4_ordered_dither;
633 case IMGFMT_BG4B: return yuv2rgb_c_4b_ordered_dither;
635 case IMGFMT_BGR1: return yuv2rgb_c_1_ordered_dither;
642 static int div_round (int dividend, int divisor)
645 return (dividend + (divisor>>1)) / divisor;
647 return -((-dividend + (divisor>>1)) / divisor);
650 int yuv2rgb_c_init_tables (SwsContext *c, const int inv_table[4], int fullRange, int brightness, int contrast, int saturation)
652 const int isRgb = IMGFMT_IS_BGR(c->dstFormat);
653 const int bpp = isRgb?IMGFMT_RGB_DEPTH(c->dstFormat):IMGFMT_BGR_DEPTH(c->dstFormat);
655 uint8_t table_Y[1024];
656 uint32_t *table_32 = 0;
657 uint16_t *table_16 = 0;
658 uint8_t *table_8 = 0;
659 uint8_t *table_332 = 0;
660 uint8_t *table_121 = 0;
661 uint8_t *table_1 = 0;
663 void *table_r = 0, *table_g = 0, *table_b = 0;
666 int64_t crv = inv_table[0];
667 int64_t cbu = inv_table[1];
668 int64_t cgu = -inv_table[2];
669 int64_t cgv = -inv_table[3];
673 //printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
679 cy = (cy *contrast )>>16;
680 crv= (crv*contrast * saturation)>>32;
681 cbu= (cbu*contrast * saturation)>>32;
682 cgu= (cgu*contrast * saturation)>>32;
683 cgv= (cgv*contrast * saturation)>>32;
684 //printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
685 oy -= 256*brightness;
687 for (i = 0; i < 1024; i++) {
690 j= (cy*(((i - 384)<<16) - oy) + (1<<31))>>32;
691 j = (j < 0) ? 0 : ((j > 255) ? 255 : j);
697 table_start= table_32 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint32_t));
699 entry_size = sizeof (uint32_t);
700 table_r = table_32 + 197;
701 table_b = table_32 + 197 + 685;
702 table_g = table_32 + 197 + 2*682;
704 for (i = -197; i < 256+197; i++)
705 ((uint32_t *)table_r)[i] = table_Y[i+384] << (isRgb ? 16 : 0);
706 for (i = -132; i < 256+132; i++)
707 ((uint32_t *)table_g)[i] = table_Y[i+384] << 8;
708 for (i = -232; i < 256+232; i++)
709 ((uint32_t *)table_b)[i] = table_Y[i+384] << (isRgb ? 0 : 16);
713 table_start= table_8 = malloc ((256 + 2*232) * sizeof (uint8_t));
715 entry_size = sizeof (uint8_t);
716 table_r = table_g = table_b = table_8 + 232;
718 for (i = -232; i < 256+232; i++)
719 ((uint8_t * )table_b)[i] = table_Y[i+384];
724 table_start= table_16 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint16_t));
726 entry_size = sizeof (uint16_t);
727 table_r = table_16 + 197;
728 table_b = table_16 + 197 + 685;
729 table_g = table_16 + 197 + 2*682;
731 for (i = -197; i < 256+197; i++) {
732 int j = table_Y[i+384] >> 3;
735 j <<= ((bpp==16) ? 11 : 10);
737 ((uint16_t *)table_r)[i] = j;
739 for (i = -132; i < 256+132; i++) {
740 int j = table_Y[i+384] >> ((bpp==16) ? 2 : 3);
742 ((uint16_t *)table_g)[i] = j << 5;
744 for (i = -232; i < 256+232; i++) {
745 int j = table_Y[i+384] >> 3;
748 j <<= ((bpp==16) ? 11 : 10);
750 ((uint16_t *)table_b)[i] = j;
755 table_start= table_332 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t));
757 entry_size = sizeof (uint8_t);
758 table_r = table_332 + 197;
759 table_b = table_332 + 197 + 685;
760 table_g = table_332 + 197 + 2*682;
762 for (i = -197; i < 256+197; i++) {
763 int j = (table_Y[i+384 - 16] + 18)/36;
768 ((uint8_t *)table_r)[i] = j;
770 for (i = -132; i < 256+132; i++) {
771 int j = (table_Y[i+384 - 16] + 18)/36;
776 ((uint8_t *)table_g)[i] = j << 2;
778 for (i = -232; i < 256+232; i++) {
779 int j = (table_Y[i+384 - 37] + 43)/85;
784 ((uint8_t *)table_b)[i] = j;
789 table_start= table_121 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t));
791 entry_size = sizeof (uint8_t);
792 table_r = table_121 + 197;
793 table_b = table_121 + 197 + 685;
794 table_g = table_121 + 197 + 2*682;
796 for (i = -197; i < 256+197; i++) {
797 int j = table_Y[i+384 - 110] >> 7;
802 ((uint8_t *)table_r)[i] = j;
804 for (i = -132; i < 256+132; i++) {
805 int j = (table_Y[i+384 - 37]+ 43)/85;
807 ((uint8_t *)table_g)[i] = j << 1;
809 for (i = -232; i < 256+232; i++) {
810 int j =table_Y[i+384 - 110] >> 7;
815 ((uint8_t *)table_b)[i] = j;
820 table_start= table_1 = malloc (256*2 * sizeof (uint8_t));
822 entry_size = sizeof (uint8_t);
824 table_r = table_b = NULL;
826 for (i = 0; i < 256+256; i++) {
827 int j = table_Y[i + 384 - 110]>>7;
829 ((uint8_t *)table_g)[i] = j;
835 MSG_ERR("%ibpp not supported by yuv2rgb\n", bpp);
840 for (i = 0; i < 256; i++) {
841 c->table_rV[i] = table_r + entry_size * div_round (crv * (i-128), 76309);
842 c->table_gU[i] = table_g + entry_size * div_round (cgu * (i-128), 76309);
843 c->table_gV[i] = entry_size * div_round (cgv * (i-128), 76309);
844 c->table_bU[i] = table_b + entry_size * div_round (cbu * (i-128), 76309);
847 if(c->yuvTable) free(c->yuvTable);
848 c->yuvTable= table_start;