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"
43 #include "../cpudetect.h"
44 #include "../mangle.h"
45 #include "../mp_msg.h"
46 #include "../libvo/img_format.h" //FIXME try to reduce dependency of such stuff
49 #include "yuv2rgb_mlib.c"
52 #define DITHER1XBPP // only for mmx
54 const uint8_t __attribute__((aligned(8))) dither_2x2_4[2][8]={
55 { 1, 3, 1, 3, 1, 3, 1, 3, },
56 { 2, 0, 2, 0, 2, 0, 2, 0, },
59 const uint8_t __attribute__((aligned(8))) dither_2x2_8[2][8]={
60 { 6, 2, 6, 2, 6, 2, 6, 2, },
61 { 0, 4, 0, 4, 0, 4, 0, 4, },
64 const uint8_t __attribute__((aligned(8))) dither_8x8_32[8][8]={
65 { 17, 9, 23, 15, 16, 8, 22, 14, },
66 { 5, 29, 3, 27, 4, 28, 2, 26, },
67 { 21, 13, 19, 11, 20, 12, 18, 10, },
68 { 0, 24, 6, 30, 1, 25, 7, 31, },
69 { 16, 8, 22, 14, 17, 9, 23, 15, },
70 { 4, 28, 2, 26, 5, 29, 3, 27, },
71 { 20, 12, 18, 10, 21, 13, 19, 11, },
72 { 1, 25, 7, 31, 0, 24, 6, 30, },
76 const uint8_t __attribute__((aligned(8))) dither_8x8_64[8][8]={
77 { 0, 48, 12, 60, 3, 51, 15, 63, },
78 { 32, 16, 44, 28, 35, 19, 47, 31, },
79 { 8, 56, 4, 52, 11, 59, 7, 55, },
80 { 40, 24, 36, 20, 43, 27, 39, 23, },
81 { 2, 50, 14, 62, 1, 49, 13, 61, },
82 { 34, 18, 46, 30, 33, 17, 45, 29, },
83 { 10, 58, 6, 54, 9, 57, 5, 53, },
84 { 42, 26, 38, 22, 41, 25, 37, 21, },
88 const uint8_t __attribute__((aligned(8))) dither_8x8_73[8][8]={
89 { 0, 55, 14, 68, 3, 58, 17, 72, },
90 { 37, 18, 50, 32, 40, 22, 54, 35, },
91 { 9, 64, 5, 59, 13, 67, 8, 63, },
92 { 46, 27, 41, 23, 49, 31, 44, 26, },
93 { 2, 57, 16, 71, 1, 56, 15, 70, },
94 { 39, 21, 52, 34, 38, 19, 51, 33, },
95 { 11, 66, 7, 62, 10, 65, 6, 60, },
96 { 48, 30, 43, 25, 47, 29, 42, 24, },
100 const uint8_t __attribute__((aligned(8))) dither_8x8_128[8][8]={
101 { 68, 36, 92, 60, 66, 34, 90, 58, },
102 { 20, 116, 12, 108, 18, 114, 10, 106, },
103 { 84, 52, 76, 44, 82, 50, 74, 42, },
104 { 0, 96, 24, 120, 6, 102, 30, 126, },
105 { 64, 32, 88, 56, 70, 38, 94, 62, },
106 { 16, 112, 8, 104, 22, 118, 14, 110, },
107 { 80, 48, 72, 40, 86, 54, 78, 46, },
108 { 4, 100, 28, 124, 2, 98, 26, 122, },
113 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
114 {117, 62, 158, 103, 113, 58, 155, 100, },
115 { 34, 199, 21, 186, 31, 196, 17, 182, },
116 {144, 89, 131, 76, 141, 86, 127, 72, },
117 { 0, 165, 41, 206, 10, 175, 52, 217, },
118 {110, 55, 151, 96, 120, 65, 162, 107, },
119 { 28, 193, 14, 179, 38, 203, 24, 189, },
120 {138, 83, 124, 69, 148, 93, 134, 79, },
121 { 7, 172, 48, 213, 3, 168, 45, 210, },
124 // tries to correct a gamma of 1.5
125 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
126 { 0, 143, 18, 200, 2, 156, 25, 215, },
127 { 78, 28, 125, 64, 89, 36, 138, 74, },
128 { 10, 180, 3, 161, 16, 195, 8, 175, },
129 {109, 51, 93, 38, 121, 60, 105, 47, },
130 { 1, 152, 23, 210, 0, 147, 20, 205, },
131 { 85, 33, 134, 71, 81, 30, 130, 67, },
132 { 14, 190, 6, 171, 12, 185, 5, 166, },
133 {117, 57, 101, 44, 113, 54, 97, 41, },
136 // tries to correct a gamma of 2.0
137 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
138 { 0, 124, 8, 193, 0, 140, 12, 213, },
139 { 55, 14, 104, 42, 66, 19, 119, 52, },
140 { 3, 168, 1, 145, 6, 187, 3, 162, },
141 { 86, 31, 70, 21, 99, 39, 82, 28, },
142 { 0, 134, 11, 206, 0, 129, 9, 200, },
143 { 62, 17, 114, 48, 58, 16, 109, 45, },
144 { 5, 181, 2, 157, 4, 175, 1, 151, },
145 { 95, 36, 78, 26, 90, 34, 74, 24, },
148 // tries to correct a gamma of 2.5
149 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
150 { 0, 107, 3, 187, 0, 125, 6, 212, },
151 { 39, 7, 86, 28, 49, 11, 102, 36, },
152 { 1, 158, 0, 131, 3, 180, 1, 151, },
153 { 68, 19, 52, 12, 81, 25, 64, 17, },
154 { 0, 119, 5, 203, 0, 113, 4, 195, },
155 { 45, 9, 96, 33, 42, 8, 91, 30, },
156 { 2, 172, 1, 144, 2, 165, 0, 137, },
157 { 77, 23, 60, 15, 72, 21, 56, 14, },
163 /* hope these constant values are cache line aligned */
164 uint64_t __attribute__((aligned(8))) mmx_00ffw = 0x00ff00ff00ff00ff;
165 uint64_t __attribute__((aligned(8))) mmx_redmask = 0xf8f8f8f8f8f8f8f8;
166 uint64_t __attribute__((aligned(8))) mmx_grnmask = 0xfcfcfcfcfcfcfcfc;
168 uint64_t __attribute__((aligned(8))) M24A= 0x00FF0000FF0000FFLL;
169 uint64_t __attribute__((aligned(8))) M24B= 0xFF0000FF0000FF00LL;
170 uint64_t __attribute__((aligned(8))) M24C= 0x0000FF0000FF0000LL;
172 // the volatile is required because gcc otherwise optimizes some writes away not knowing that these
173 // are read in the asm block
174 volatile uint64_t __attribute__((aligned(8))) b5Dither;
175 volatile uint64_t __attribute__((aligned(8))) g5Dither;
176 volatile uint64_t __attribute__((aligned(8))) g6Dither;
177 volatile uint64_t __attribute__((aligned(8))) r5Dither;
179 uint64_t __attribute__((aligned(8))) dither4[2]={
180 0x0103010301030103LL,
181 0x0200020002000200LL,};
183 uint64_t __attribute__((aligned(8))) dither8[2]={
184 0x0602060206020602LL,
185 0x0004000400040004LL,};
196 #define RENAME(a) a ## _MMX
197 #include "yuv2rgb_template.c"
205 #define RENAME(a) a ## _MMX2
206 #include "yuv2rgb_template.c"
208 #endif // CAN_COMPILE_X86_ASM
210 const int32_t Inverse_Table_6_9[8][4] = {
211 {117504, 138453, 13954, 34903}, /* no sequence_display_extension */
212 {117504, 138453, 13954, 34903}, /* ITU-R Rec. 709 (1990) */
213 {104597, 132201, 25675, 53279}, /* unspecified */
214 {104597, 132201, 25675, 53279}, /* reserved */
215 {104448, 132798, 24759, 53109}, /* FCC */
216 {104597, 132201, 25675, 53279}, /* ITU-R Rec. 624-4 System B, G */
217 {104597, 132201, 25675, 53279}, /* SMPTE 170M */
218 {117579, 136230, 16907, 35559} /* SMPTE 240M (1987) */
224 r = c->table_rV[V]; \
225 g = c->table_gU[U] + c->table_gV[V]; \
230 dst_1[2*i] = r[Y] + g[Y] + b[Y]; \
232 dst_1[2*i+1] = r[Y] + g[Y] + b[Y];
236 dst_2[2*i] = r[Y] + g[Y] + b[Y]; \
238 dst_2[2*i+1] = r[Y] + g[Y] + b[Y];
242 dst_1[6*i] = r[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = b[Y]; \
244 dst_1[6*i+3] = r[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = b[Y];
248 dst_2[6*i] = r[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = b[Y]; \
250 dst_2[6*i+3] = r[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = b[Y];
254 dst_1[6*i] = b[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = r[Y]; \
256 dst_1[6*i+3] = b[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = r[Y];
260 dst_2[6*i] = b[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = r[Y]; \
262 dst_2[6*i+3] = b[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = r[Y];
264 #define PROLOG(func_name, dst_type) \
265 static int func_name(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, \
266 int srcSliceH, uint8_t* dst[], int dstStride[]){\
269 if(c->srcFormat == IMGFMT_422P){\
273 for(y=0; y<srcSliceH; y+=2){\
274 dst_type *dst_1= (dst_type*)(dst[0] + (y+srcSliceY )*dstStride[0]);\
275 dst_type *dst_2= (dst_type*)(dst[0] + (y+srcSliceY+1)*dstStride[0]);\
276 dst_type *r, *g, *b;\
277 uint8_t *py_1= src[0] + y*srcStride[0];\
278 uint8_t *py_2= py_1 + srcStride[0];\
279 uint8_t *pu= src[1] + (y>>1)*srcStride[1];\
280 uint8_t *pv= src[2] + (y>>1)*srcStride[2];\
281 unsigned int h_size= c->dstW>>3;\
285 #define EPILOG(dst_delta)\
297 PROLOG(yuv2rgb_c_32, uint32_t)
315 PROLOG(yuv2rgb_c_24_rgb, uint8_t)
333 // only trivial mods from yuv2rgb_c_24_rgb
334 PROLOG(yuv2rgb_c_24_bgr, uint8_t)
352 // This is exactly the same code as yuv2rgb_c_32 except for the types of
353 // r, g, b, dst_1, dst_2
354 PROLOG(yuv2rgb_c_16, uint16_t)
372 // This is exactly the same code as yuv2rgb_c_32 except for the types of
373 // r, g, b, dst_1, dst_2
374 PROLOG(yuv2rgb_c_8, uint8_t)
392 // r, g, b, dst_1, dst_2
393 PROLOG(yuv2rgb_c_8_ordered_dither, uint8_t)
394 const uint8_t *d32= dither_8x8_32[y&7];
395 const uint8_t *d64= dither_8x8_73[y&7];
396 #define DST1bpp8(i,o) \
398 dst_1[2*i] = r[Y+d32[0+o]] + g[Y+d32[0+o]] + b[Y+d64[0+o]]; \
400 dst_1[2*i+1] = r[Y+d32[1+o]] + g[Y+d32[1+o]] + b[Y+d64[1+o]];
402 #define DST2bpp8(i,o) \
404 dst_2[2*i] = r[Y+d32[8+o]] + g[Y+d32[8+o]] + b[Y+d64[8+o]]; \
406 dst_2[2*i+1] = r[Y+d32[9+o]] + g[Y+d32[9+o]] + b[Y+d64[9+o]];
427 // This is exactly the same code as yuv2rgb_c_32 except for the types of
428 // r, g, b, dst_1, dst_2
429 PROLOG(yuv2rgb_c_4, uint8_t)
433 acc = r[Y] + g[Y] + b[Y]; \
435 acc |= (r[Y] + g[Y] + b[Y])<<4;\
440 acc = r[Y] + g[Y] + b[Y]; \
442 acc |= (r[Y] + g[Y] + b[Y])<<4;\
462 PROLOG(yuv2rgb_c_4_ordered_dither, uint8_t)
463 const uint8_t *d64= dither_8x8_73[y&7];
464 const uint8_t *d128=dither_8x8_220[y&7];
467 #define DST1bpp4(i,o) \
469 acc = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]]; \
471 acc |= (r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]])<<4;\
474 #define DST2bpp4(i,o) \
476 acc = r[Y+d128[8+o]] + g[Y+d64[8+o]] + b[Y+d128[8+o]]; \
478 acc |= (r[Y+d128[9+o]] + g[Y+d64[9+o]] + b[Y+d128[9+o]])<<4;\
499 // This is exactly the same code as yuv2rgb_c_32 except for the types of
500 // r, g, b, dst_1, dst_2
501 PROLOG(yuv2rgb_c_4b, uint8_t)
519 PROLOG(yuv2rgb_c_4b_ordered_dither, uint8_t)
520 const uint8_t *d64= dither_8x8_73[y&7];
521 const uint8_t *d128=dither_8x8_220[y&7];
523 #define DST1bpp4b(i,o) \
525 dst_1[2*i] = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]]; \
527 dst_1[2*i+1] = r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]];
529 #define DST2bpp4b(i,o) \
531 dst_2[2*i] = r[Y+d128[8+o]] + g[Y+d64[8+o]] + b[Y+d128[8+o]]; \
533 dst_2[2*i+1] = r[Y+d128[9+o]] + g[Y+d64[9+o]] + b[Y+d128[9+o]];
553 PROLOG(yuv2rgb_c_1_ordered_dither, uint8_t)
554 const uint8_t *d128=dither_8x8_220[y&7];
555 char out_1=0, out_2=0;
556 g= c->table_gU[128] + c->table_gV[128];
558 #define DST1bpp1(i,o) \
560 out_1+= out_1 + g[Y+d128[0+o]]; \
562 out_1+= out_1 + g[Y+d128[1+o]];
564 #define DST2bpp1(i,o) \
566 out_2+= out_2 + g[Y+d128[8+o]]; \
568 out_2+= out_2 + g[Y+d128[9+o]];
586 SwsFunc yuv2rgb_get_func_ptr (SwsContext *c)
589 if(gCpuCaps.hasMMX2){
590 switch(c->dstFormat){
591 case IMGFMT_BGR32: return yuv420_rgb32_MMX2;
592 case IMGFMT_BGR24: return yuv420_rgb24_MMX2;
593 case IMGFMT_BGR16: return yuv420_rgb16_MMX2;
594 case IMGFMT_BGR15: return yuv420_rgb15_MMX2;
598 switch(c->dstFormat){
599 case IMGFMT_BGR32: return yuv420_rgb32_MMX;
600 case IMGFMT_BGR24: return yuv420_rgb24_MMX;
601 case IMGFMT_BGR16: return yuv420_rgb16_MMX;
602 case IMGFMT_BGR15: return yuv420_rgb15_MMX;
608 SwsFunc t= yuv2rgb_init_mlib(c);
612 mp_msg(MSGT_SWS,MSGL_WARN,"No accelerated colorspace conversion found\n");
614 switch(c->dstFormat){
616 case IMGFMT_BGR32: return yuv2rgb_c_32;
617 case IMGFMT_RGB24: return yuv2rgb_c_24_rgb;
618 case IMGFMT_BGR24: return yuv2rgb_c_24_bgr;
622 case IMGFMT_BGR15: return yuv2rgb_c_16;
624 case IMGFMT_BGR8: return yuv2rgb_c_8_ordered_dither;
626 case IMGFMT_BGR4: return yuv2rgb_c_4_ordered_dither;
628 case IMGFMT_BG4B: return yuv2rgb_c_4b_ordered_dither;
630 case IMGFMT_BGR1: return yuv2rgb_c_1_ordered_dither;
637 static int div_round (int dividend, int divisor)
640 return (dividend + (divisor>>1)) / divisor;
642 return -((-dividend + (divisor>>1)) / divisor);
645 int yuv2rgb_c_init_tables (SwsContext *c, const int inv_table[4], int fullRange, int brightness, int contrast, int saturation)
647 const int bpp= c->dstFormat&0xFF;
648 const int isRgb= (c->dstFormat>>24) != 'R';
650 uint8_t table_Y[1024];
651 uint32_t *table_32 = 0;
652 uint16_t *table_16 = 0;
653 uint8_t *table_8 = 0;
654 uint8_t *table_332 = 0;
655 uint8_t *table_121 = 0;
656 uint8_t *table_1 = 0;
658 void *table_r = 0, *table_g = 0, *table_b = 0;
661 int64_t crv = inv_table[0];
662 int64_t cbu = inv_table[1];
663 int64_t cgu = -inv_table[2];
664 int64_t cgv = -inv_table[3];
668 //printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
674 cy = (cy *contrast )>>16;
675 crv= (crv*contrast * saturation)>>32;
676 cbu= (cbu*contrast * saturation)>>32;
677 cgu= (cgu*contrast * saturation)>>32;
678 cgv= (cgv*contrast * saturation)>>32;
679 //printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
680 oy -= 256*brightness;
682 for (i = 0; i < 1024; i++) {
685 j= (cy*(((i - 384)<<16) - oy) + (1<<31))>>32;
686 j = (j < 0) ? 0 : ((j > 255) ? 255 : j);
692 table_start= table_32 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint32_t));
694 entry_size = sizeof (uint32_t);
695 table_r = table_32 + 197;
696 table_b = table_32 + 197 + 685;
697 table_g = table_32 + 197 + 2*682;
699 for (i = -197; i < 256+197; i++)
700 ((uint32_t *)table_r)[i] = table_Y[i+384] << (isRgb ? 16 : 0);
701 for (i = -132; i < 256+132; i++)
702 ((uint32_t *)table_g)[i] = table_Y[i+384] << 8;
703 for (i = -232; i < 256+232; i++)
704 ((uint32_t *)table_b)[i] = table_Y[i+384] << (isRgb ? 0 : 16);
708 table_start= table_8 = malloc ((256 + 2*232) * sizeof (uint8_t));
710 entry_size = sizeof (uint8_t);
711 table_r = table_g = table_b = table_8 + 232;
713 for (i = -232; i < 256+232; i++)
714 ((uint8_t * )table_b)[i] = table_Y[i+384];
719 table_start= table_16 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint16_t));
721 entry_size = sizeof (uint16_t);
722 table_r = table_16 + 197;
723 table_b = table_16 + 197 + 685;
724 table_g = table_16 + 197 + 2*682;
726 for (i = -197; i < 256+197; i++) {
727 int j = table_Y[i+384] >> 3;
730 j <<= ((bpp==16) ? 11 : 10);
732 ((uint16_t *)table_r)[i] = j;
734 for (i = -132; i < 256+132; i++) {
735 int j = table_Y[i+384] >> ((bpp==16) ? 2 : 3);
737 ((uint16_t *)table_g)[i] = j << 5;
739 for (i = -232; i < 256+232; i++) {
740 int j = table_Y[i+384] >> 3;
743 j <<= ((bpp==16) ? 11 : 10);
745 ((uint16_t *)table_b)[i] = j;
750 table_start= table_332 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t));
752 entry_size = sizeof (uint8_t);
753 table_r = table_332 + 197;
754 table_b = table_332 + 197 + 685;
755 table_g = table_332 + 197 + 2*682;
757 for (i = -197; i < 256+197; i++) {
758 int j = (table_Y[i+384 - 16] + 18)/36;
763 ((uint8_t *)table_r)[i] = j;
765 for (i = -132; i < 256+132; i++) {
766 int j = (table_Y[i+384 - 16] + 18)/36;
771 ((uint8_t *)table_g)[i] = j << 2;
773 for (i = -232; i < 256+232; i++) {
774 int j = (table_Y[i+384 - 37] + 43)/85;
779 ((uint8_t *)table_b)[i] = j;
784 table_start= table_121 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t));
786 entry_size = sizeof (uint8_t);
787 table_r = table_121 + 197;
788 table_b = table_121 + 197 + 685;
789 table_g = table_121 + 197 + 2*682;
791 for (i = -197; i < 256+197; i++) {
792 int j = table_Y[i+384 - 110] >> 7;
797 ((uint8_t *)table_r)[i] = j;
799 for (i = -132; i < 256+132; i++) {
800 int j = (table_Y[i+384 - 37]+ 43)/85;
802 ((uint8_t *)table_g)[i] = j << 1;
804 for (i = -232; i < 256+232; i++) {
805 int j =table_Y[i+384 - 110] >> 7;
810 ((uint8_t *)table_b)[i] = j;
815 table_start= table_1 = malloc (256*2 * sizeof (uint8_t));
817 entry_size = sizeof (uint8_t);
819 table_r = table_b = NULL;
821 for (i = 0; i < 256+256; i++) {
822 int j = table_Y[i + 384 - 110]>>7;
824 ((uint8_t *)table_g)[i] = j;
830 mp_msg(MSGT_SWS,MSGL_ERR,"%ibpp not supported by yuv2rgb\n", bpp);
835 for (i = 0; i < 256; i++) {
836 c->table_rV[i] = table_r + entry_size * div_round (crv * (i-128), 76309);
837 c->table_gU[i] = table_g + entry_size * div_round (cgu * (i-128), 76309);
838 c->table_gV[i] = entry_size * div_round (cgv * (i-128), 76309);
839 c->table_bU[i] = table_b + entry_size * div_round (cbu * (i-128), 76309);
842 if(c->yuvTable) free(c->yuvTable);
843 c->yuvTable= table_start;