libswscale/yuv2rgb.c

   1 /*
   2  * yuv2rgb.c, Software YUV to RGB coverter
   3  *
   4  *  Copyright (C) 1999, Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
   5  *  All Rights Reserved.
   6  *
   7  *  Functions broken out from display_x11.c and several new modes
   8  *  added by Håkan Hjort <d95hjort@dtek.chalmers.se>
   9  *
  10  *  15 & 16 bpp support by Franck Sicard <Franck.Sicard@solsoft.fr>
  11  *
  12  *  This file is part of mpeg2dec, a free MPEG-2 video decoder
  13  *
  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)
  17  *  any later version.
  18  *
  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.
  23  *
  24  *  You should have received a copy of the GNU General Public License
  25  *  along with mpeg2dec; if not, write to the Free Software
  26  *  Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  27  *
  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
  31  */
  32
  33 #include <stdio.h>
  34 #include <stdlib.h>
  35 #include <inttypes.h>
  36 #include <assert.h>
  37
  38 #include "config.h"
  39 #include "rgb2rgb.h"
  40 #include "swscale.h"
  41 #include "swscale_internal.h"
  42
  43 #ifdef HAVE_MLIB
  44 #include "yuv2rgb_mlib.c"
  45 #endif
  46
  47 #define DITHER1XBPP // only for mmx
  48
  49 const uint8_t  __attribute__((aligned(8))) dither_2x2_4[2][8]={
  50 {  1,   3,   1,   3,   1,   3,   1,   3, },
  51 {  2,   0,   2,   0,   2,   0,   2,   0, },
  52 };
  53
  54 const uint8_t  __attribute__((aligned(8))) dither_2x2_8[2][8]={
  55 {  6,   2,   6,   2,   6,   2,   6,   2, },
  56 {  0,   4,   0,   4,   0,   4,   0,   4, },
  57 };
  58
  59 const uint8_t  __attribute__((aligned(8))) dither_8x8_32[8][8]={
  60 { 17,   9,  23,  15,  16,   8,  22,  14, },
  61 {  5,  29,   3,  27,   4,  28,   2,  26, },
  62 { 21,  13,  19,  11,  20,  12,  18,  10, },
  63 {  0,  24,   6,  30,   1,  25,   7,  31, },
  64 { 16,   8,  22,  14,  17,   9,  23,  15, },
  65 {  4,  28,   2,  26,   5,  29,   3,  27, },
  66 { 20,  12,  18,  10,  21,  13,  19,  11, },
  67 {  1,  25,   7,  31,   0,  24,   6,  30, },
  68 };
  69
  70 #if 0
  71 const uint8_t  __attribute__((aligned(8))) dither_8x8_64[8][8]={
  72 {  0,  48,  12,  60,   3,  51,  15,  63, },
  73 { 32,  16,  44,  28,  35,  19,  47,  31, },
  74 {  8,  56,   4,  52,  11,  59,   7,  55, },
  75 { 40,  24,  36,  20,  43,  27,  39,  23, },
  76 {  2,  50,  14,  62,   1,  49,  13,  61, },
  77 { 34,  18,  46,  30,  33,  17,  45,  29, },
  78 { 10,  58,   6,  54,   9,  57,   5,  53, },
  79 { 42,  26,  38,  22,  41,  25,  37,  21, },
  80 };
  81 #endif
  82
  83 const uint8_t  __attribute__((aligned(8))) dither_8x8_73[8][8]={
  84 {  0,  55,  14,  68,   3,  58,  17,  72, },
  85 { 37,  18,  50,  32,  40,  22,  54,  35, },
  86 {  9,  64,   5,  59,  13,  67,   8,  63, },
  87 { 46,  27,  41,  23,  49,  31,  44,  26, },
  88 {  2,  57,  16,  71,   1,  56,  15,  70, },
  89 { 39,  21,  52,  34,  38,  19,  51,  33, },
  90 { 11,  66,   7,  62,  10,  65,   6,  60, },
  91 { 48,  30,  43,  25,  47,  29,  42,  24, },
  92 };
  93
  94 #if 0
  95 const uint8_t  __attribute__((aligned(8))) dither_8x8_128[8][8]={
  96 { 68,  36,  92,  60,  66,  34,  90,  58, },
  97 { 20, 116,  12, 108,  18, 114,  10, 106, },
  98 { 84,  52,  76,  44,  82,  50,  74,  42, },
  99 {  0,  96,  24, 120,   6, 102,  30, 126, },
 100 { 64,  32,  88,  56,  70,  38,  94,  62, },
 101 { 16, 112,   8, 104,  22, 118,  14, 110, },
 102 { 80,  48,  72,  40,  86,  54,  78,  46, },
 103 {  4, 100,  28, 124,   2,  98,  26, 122, },
 104 };
 105 #endif
 106
 107 #if 1
 108 const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
 109 {117,  62, 158, 103, 113,  58, 155, 100, },
 110 { 34, 199,  21, 186,  31, 196,  17, 182, },
 111 {144,  89, 131,  76, 141,  86, 127,  72, },
 112 {  0, 165,  41, 206,  10, 175,  52, 217, },
 113 {110,  55, 151,  96, 120,  65, 162, 107, },
 114 { 28, 193,  14, 179,  38, 203,  24, 189, },
 115 {138,  83, 124,  69, 148,  93, 134,  79, },
 116 {  7, 172,  48, 213,   3, 168,  45, 210, },
 117 };
 118 #elif 1
 119 // tries to correct a gamma of 1.5
 120 const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
 121 {  0, 143,  18, 200,   2, 156,  25, 215, },
 122 { 78,  28, 125,  64,  89,  36, 138,  74, },
 123 { 10, 180,   3, 161,  16, 195,   8, 175, },
 124 {109,  51,  93,  38, 121,  60, 105,  47, },
 125 {  1, 152,  23, 210,   0, 147,  20, 205, },
 126 { 85,  33, 134,  71,  81,  30, 130,  67, },
 127 { 14, 190,   6, 171,  12, 185,   5, 166, },
 128 {117,  57, 101,  44, 113,  54,  97,  41, },
 129 };
 130 #elif 1
 131 // tries to correct a gamma of 2.0
 132 const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
 133 {  0, 124,   8, 193,   0, 140,  12, 213, },
 134 { 55,  14, 104,  42,  66,  19, 119,  52, },
 135 {  3, 168,   1, 145,   6, 187,   3, 162, },
 136 { 86,  31,  70,  21,  99,  39,  82,  28, },
 137 {  0, 134,  11, 206,   0, 129,   9, 200, },
 138 { 62,  17, 114,  48,  58,  16, 109,  45, },
 139 {  5, 181,   2, 157,   4, 175,   1, 151, },
 140 { 95,  36,  78,  26,  90,  34,  74,  24, },
 141 };
 142 #else
 143 // tries to correct a gamma of 2.5
 144 const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
 145 {  0, 107,   3, 187,   0, 125,   6, 212, },
 146 { 39,   7,  86,  28,  49,  11, 102,  36, },
 147 {  1, 158,   0, 131,   3, 180,   1, 151, },
 148 { 68,  19,  52,  12,  81,  25,  64,  17, },
 149 {  0, 119,   5, 203,   0, 113,   4, 195, },
 150 { 45,   9,  96,  33,  42,   8,  91,  30, },
 151 {  2, 172,   1, 144,   2, 165,   0, 137, },
 152 { 77,  23,  60,  15,  72,  21,  56,  14, },
 153 };
 154 #endif
 155
 156 #ifdef HAVE_MMX
 157
 158 /* hope these constant values are cache line aligned */
 159 static uint64_t attribute_used __attribute__((aligned(8))) mmx_00ffw = 0x00ff00ff00ff00ffULL;
 160 static uint64_t attribute_used __attribute__((aligned(8))) mmx_redmask = 0xf8f8f8f8f8f8f8f8ULL;
 161 static uint64_t attribute_used __attribute__((aligned(8))) mmx_grnmask = 0xfcfcfcfcfcfcfcfcULL;
 162
 163 static uint64_t attribute_used __attribute__((aligned(8))) M24A=   0x00FF0000FF0000FFULL;
 164 static uint64_t attribute_used __attribute__((aligned(8))) M24B=   0xFF0000FF0000FF00ULL;
 165 static uint64_t attribute_used __attribute__((aligned(8))) M24C=   0x0000FF0000FF0000ULL;
 166
 167 // the volatile is required because gcc otherwise optimizes some writes away not knowing that these
 168 // are read in the asm block
 169 static volatile uint64_t attribute_used __attribute__((aligned(8))) b5Dither;
 170 static volatile uint64_t attribute_used __attribute__((aligned(8))) g5Dither;
 171 static volatile uint64_t attribute_used __attribute__((aligned(8))) g6Dither;
 172 static volatile uint64_t attribute_used __attribute__((aligned(8))) r5Dither;
 173
 174 static uint64_t __attribute__((aligned(8))) dither4[2]={
 175         0x0103010301030103LL,
 176         0x0200020002000200LL,};
 177
 178 static uint64_t __attribute__((aligned(8))) dither8[2]={
 179         0x0602060206020602LL,
 180         0x0004000400040004LL,};
 181
 182 #undef HAVE_MMX
 183
 184 //MMX versions
 185 #undef RENAME
 186 #define HAVE_MMX
 187 #undef HAVE_MMX2
 188 #undef HAVE_3DNOW
 189 #define RENAME(a) a ## _MMX
 190 #include "yuv2rgb_template.c"
 191
 192 //MMX2 versions
 193 #undef RENAME
 194 #define HAVE_MMX
 195 #define HAVE_MMX2
 196 #undef HAVE_3DNOW
 197 #define RENAME(a) a ## _MMX2
 198 #include "yuv2rgb_template.c"
 199
 200 #endif /* defined(ARCH_X86) */
 201
 202 const int32_t Inverse_Table_6_9[8][4] = {
 203     {117504, 138453, 13954, 34903}, /* no sequence_display_extension */
 204     {117504, 138453, 13954, 34903}, /* ITU-R Rec. 709 (1990) */
 205     {104597, 132201, 25675, 53279}, /* unspecified */
 206     {104597, 132201, 25675, 53279}, /* reserved */
 207     {104448, 132798, 24759, 53109}, /* FCC */
 208     {104597, 132201, 25675, 53279}, /* ITU-R Rec. 624-4 System B, G */
 209     {104597, 132201, 25675, 53279}, /* SMPTE 170M */
 210     {117579, 136230, 16907, 35559}  /* SMPTE 240M (1987) */
 211 };
 212
 213 #define RGB(i)                                  \
 214         U = pu[i];                              \
 215         V = pv[i];                              \
 216         r = (void *)c->table_rV[V];                     \
 217         g = (void *)(c->table_gU[U] + c->table_gV[V]);          \
 218         b = (void *)c->table_bU[U];
 219
 220 #define DST1(i)                                 \
 221         Y = py_1[2*i];                          \
 222         dst_1[2*i] = r[Y] + g[Y] + b[Y];        \
 223         Y = py_1[2*i+1];                        \
 224         dst_1[2*i+1] = r[Y] + g[Y] + b[Y];
 225
 226 #define DST2(i)                                 \
 227         Y = py_2[2*i];                          \
 228         dst_2[2*i] = r[Y] + g[Y] + b[Y];        \
 229         Y = py_2[2*i+1];                        \
 230         dst_2[2*i+1] = r[Y] + g[Y] + b[Y];
 231
 232 #define DST1RGB(i)                                                      \
 233         Y = py_1[2*i];                                                  \
 234         dst_1[6*i] = r[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = b[Y];    \
 235         Y = py_1[2*i+1];                                                \
 236         dst_1[6*i+3] = r[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = b[Y];
 237
 238 #define DST2RGB(i)                                                      \
 239         Y = py_2[2*i];                                                  \
 240         dst_2[6*i] = r[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = b[Y];    \
 241         Y = py_2[2*i+1];                                                \
 242         dst_2[6*i+3] = r[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = b[Y];
 243
 244 #define DST1BGR(i)                                                      \
 245         Y = py_1[2*i];                                                  \
 246         dst_1[6*i] = b[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = r[Y];    \
 247         Y = py_1[2*i+1];                                                \
 248         dst_1[6*i+3] = b[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = r[Y];
 249
 250 #define DST2BGR(i)                                                      \
 251         Y = py_2[2*i];                                                  \
 252         dst_2[6*i] = b[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = r[Y];    \
 253         Y = py_2[2*i+1];                                                \
 254         dst_2[6*i+3] = b[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = r[Y];
 255
 256 #define PROLOG(func_name, dst_type) \
 257 static int func_name(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, \
 258              int srcSliceH, uint8_t* dst[], int dstStride[]){\
 259     int y;\
 260 \
 261     if(c->srcFormat == PIX_FMT_YUV422P){\
 262         srcStride[1] *= 2;\
 263         srcStride[2] *= 2;\
 264     }\
 265     for(y=0; y<srcSliceH; y+=2){\
 266         dst_type *dst_1= (dst_type*)(dst[0] + (y+srcSliceY  )*dstStride[0]);\
 267         dst_type *dst_2= (dst_type*)(dst[0] + (y+srcSliceY+1)*dstStride[0]);\
 268         dst_type attribute_unused *r, *b;\
 269         dst_type *g;\
 270         uint8_t *py_1= src[0] + y*srcStride[0];\
 271         uint8_t *py_2= py_1 + srcStride[0];\
 272         uint8_t *pu= src[1] + (y>>1)*srcStride[1];\
 273         uint8_t *pv= src[2] + (y>>1)*srcStride[2];\
 274         unsigned int h_size= c->dstW>>3;\
 275         while (h_size--) {\
 276             int attribute_unused U, V;\
 277             int Y;\
 278
 279 #define EPILOG(dst_delta)\
 280             pu += 4;\
 281             pv += 4;\
 282             py_1 += 8;\
 283             py_2 += 8;\
 284             dst_1 += dst_delta;\
 285             dst_2 += dst_delta;\
 286         }\
 287     }\
 288     return srcSliceH;\
 289 }
 290
 291 PROLOG(yuv2rgb_c_32, uint32_t)
 292         RGB(0);
 293         DST1(0);
 294         DST2(0);
 295
 296         RGB(1);
 297         DST2(1);
 298         DST1(1);
 299
 300         RGB(2);
 301         DST1(2);
 302         DST2(2);
 303
 304         RGB(3);
 305         DST2(3);
 306         DST1(3);
 307 EPILOG(8)
 308
 309 PROLOG(yuv2rgb_c_24_rgb, uint8_t)
 310         RGB(0);
 311         DST1RGB(0);
 312         DST2RGB(0);
 313
 314         RGB(1);
 315         DST2RGB(1);
 316         DST1RGB(1);
 317
 318         RGB(2);
 319         DST1RGB(2);
 320         DST2RGB(2);
 321
 322         RGB(3);
 323         DST2RGB(3);
 324         DST1RGB(3);
 325 EPILOG(24)
 326
 327 // only trivial mods from yuv2rgb_c_24_rgb
 328 PROLOG(yuv2rgb_c_24_bgr, uint8_t)
 329         RGB(0);
 330         DST1BGR(0);
 331         DST2BGR(0);
 332
 333         RGB(1);
 334         DST2BGR(1);
 335         DST1BGR(1);
 336
 337         RGB(2);
 338         DST1BGR(2);
 339         DST2BGR(2);
 340
 341         RGB(3);
 342         DST2BGR(3);
 343         DST1BGR(3);
 344 EPILOG(24)
 345
 346 // This is exactly the same code as yuv2rgb_c_32 except for the types of
 347 // r, g, b, dst_1, dst_2
 348 PROLOG(yuv2rgb_c_16, uint16_t)
 349         RGB(0);
 350         DST1(0);
 351         DST2(0);
 352
 353         RGB(1);
 354         DST2(1);
 355         DST1(1);
 356
 357         RGB(2);
 358         DST1(2);
 359         DST2(2);
 360
 361         RGB(3);
 362         DST2(3);
 363         DST1(3);
 364 EPILOG(8)
 365
 366 // This is exactly the same code as yuv2rgb_c_32 except for the types of
 367 // r, g, b, dst_1, dst_2
 368 PROLOG(yuv2rgb_c_8, uint8_t)
 369         RGB(0);
 370         DST1(0);
 371         DST2(0);
 372
 373         RGB(1);
 374         DST2(1);
 375         DST1(1);
 376
 377         RGB(2);
 378         DST1(2);
 379         DST2(2);
 380
 381         RGB(3);
 382         DST2(3);
 383         DST1(3);
 384 EPILOG(8)
 385
 386 // r, g, b, dst_1, dst_2
 387 PROLOG(yuv2rgb_c_8_ordered_dither, uint8_t)
 388         const uint8_t *d32= dither_8x8_32[y&7];
 389         const uint8_t *d64= dither_8x8_73[y&7];
 390 #define DST1bpp8(i,o)                                   \
 391         Y = py_1[2*i];                          \
 392         dst_1[2*i] = r[Y+d32[0+o]] + g[Y+d32[0+o]] + b[Y+d64[0+o]];     \
 393         Y = py_1[2*i+1];                        \
 394         dst_1[2*i+1] = r[Y+d32[1+o]] + g[Y+d32[1+o]] + b[Y+d64[1+o]];
 395
 396 #define DST2bpp8(i,o)                                   \
 397         Y = py_2[2*i];                          \
 398         dst_2[2*i] =  r[Y+d32[8+o]] + g[Y+d32[8+o]] + b[Y+d64[8+o]];    \
 399         Y = py_2[2*i+1];                        \
 400         dst_2[2*i+1] =  r[Y+d32[9+o]] + g[Y+d32[9+o]] + b[Y+d64[9+o]];
 401
 402
 403         RGB(0);
 404         DST1bpp8(0,0);
 405         DST2bpp8(0,0);
 406
 407         RGB(1);
 408         DST2bpp8(1,2);
 409         DST1bpp8(1,2);
 410
 411         RGB(2);
 412         DST1bpp8(2,4);
 413         DST2bpp8(2,4);
 414
 415         RGB(3);
 416         DST2bpp8(3,6);
 417         DST1bpp8(3,6);
 418 EPILOG(8)
 419
 420
 421 // This is exactly the same code as yuv2rgb_c_32 except for the types of
 422 // r, g, b, dst_1, dst_2
 423 PROLOG(yuv2rgb_c_4, uint8_t)
 424         int acc;
 425 #define DST1_4(i)                                       \
 426         Y = py_1[2*i];                          \
 427         acc = r[Y] + g[Y] + b[Y];       \
 428         Y = py_1[2*i+1];                        \
 429         acc |= (r[Y] + g[Y] + b[Y])<<4;\
 430         dst_1[i] = acc;
 431
 432 #define DST2_4(i)                                       \
 433         Y = py_2[2*i];                          \
 434         acc = r[Y] + g[Y] + b[Y];       \
 435         Y = py_2[2*i+1];                        \
 436         acc |= (r[Y] + g[Y] + b[Y])<<4;\
 437         dst_2[i] = acc;
 438
 439         RGB(0);
 440         DST1_4(0);
 441         DST2_4(0);
 442
 443         RGB(1);
 444         DST2_4(1);
 445         DST1_4(1);
 446
 447         RGB(2);
 448         DST1_4(2);
 449         DST2_4(2);
 450
 451         RGB(3);
 452         DST2_4(3);
 453         DST1_4(3);
 454 EPILOG(4)
 455
 456 PROLOG(yuv2rgb_c_4_ordered_dither, uint8_t)
 457         const uint8_t *d64= dither_8x8_73[y&7];
 458         const uint8_t *d128=dither_8x8_220[y&7];
 459         int acc;
 460
 461 #define DST1bpp4(i,o)                                   \
 462         Y = py_1[2*i];                          \
 463         acc = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]];  \
 464         Y = py_1[2*i+1];                        \
 465         acc |= (r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]])<<4;\
 466         dst_1[i]= acc;
 467
 468 #define DST2bpp4(i,o)                                   \
 469         Y = py_2[2*i];                          \
 470         acc =  r[Y+d128[8+o]] + g[Y+d64[8+o]] + b[Y+d128[8+o]]; \
 471         Y = py_2[2*i+1];                        \
 472         acc |=  (r[Y+d128[9+o]] + g[Y+d64[9+o]] + b[Y+d128[9+o]])<<4;\
 473         dst_2[i]= acc;
 474
 475
 476         RGB(0);
 477         DST1bpp4(0,0);
 478         DST2bpp4(0,0);
 479
 480         RGB(1);
 481         DST2bpp4(1,2);
 482         DST1bpp4(1,2);
 483
 484         RGB(2);
 485         DST1bpp4(2,4);
 486         DST2bpp4(2,4);
 487
 488         RGB(3);
 489         DST2bpp4(3,6);
 490         DST1bpp4(3,6);
 491 EPILOG(4)
 492
 493 // This is exactly the same code as yuv2rgb_c_32 except for the types of
 494 // r, g, b, dst_1, dst_2
 495 PROLOG(yuv2rgb_c_4b, uint8_t)
 496         RGB(0);
 497         DST1(0);
 498         DST2(0);
 499
 500         RGB(1);
 501         DST2(1);
 502         DST1(1);
 503
 504         RGB(2);
 505         DST1(2);
 506         DST2(2);
 507
 508         RGB(3);
 509         DST2(3);
 510         DST1(3);
 511 EPILOG(8)
 512
 513 PROLOG(yuv2rgb_c_4b_ordered_dither, uint8_t)
 514         const uint8_t *d64= dither_8x8_73[y&7];
 515         const uint8_t *d128=dither_8x8_220[y&7];
 516
 517 #define DST1bpp4b(i,o)                                  \
 518         Y = py_1[2*i];                          \
 519         dst_1[2*i] = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]];   \
 520         Y = py_1[2*i+1];                        \
 521         dst_1[2*i+1] = r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]];
 522
 523 #define DST2bpp4b(i,o)                                  \
 524         Y = py_2[2*i];                          \
 525         dst_2[2*i] =  r[Y+d128[8+o]] + g[Y+d64[8+o]] + b[Y+d128[8+o]];  \
 526         Y = py_2[2*i+1];                        \
 527         dst_2[2*i+1] =  r[Y+d128[9+o]] + g[Y+d64[9+o]] + b[Y+d128[9+o]];
 528
 529
 530         RGB(0);
 531         DST1bpp4b(0,0);
 532         DST2bpp4b(0,0);
 533
 534         RGB(1);
 535         DST2bpp4b(1,2);
 536         DST1bpp4b(1,2);
 537
 538         RGB(2);
 539         DST1bpp4b(2,4);
 540         DST2bpp4b(2,4);
 541
 542         RGB(3);
 543         DST2bpp4b(3,6);
 544         DST1bpp4b(3,6);
 545 EPILOG(8)
 546
 547 PROLOG(yuv2rgb_c_1_ordered_dither, uint8_t)
 548         const uint8_t *d128=dither_8x8_220[y&7];
 549         char out_1=0, out_2=0;
 550         g= c->table_gU[128] + c->table_gV[128];
 551
 552 #define DST1bpp1(i,o)                                   \
 553         Y = py_1[2*i];                          \
 554         out_1+= out_1 + g[Y+d128[0+o]]; \
 555         Y = py_1[2*i+1];                        \
 556         out_1+= out_1 + g[Y+d128[1+o]];
 557
 558 #define DST2bpp1(i,o)                                   \
 559         Y = py_2[2*i];                          \
 560         out_2+= out_2 + g[Y+d128[8+o]]; \
 561         Y = py_2[2*i+1];                        \
 562         out_2+= out_2 + g[Y+d128[9+o]];
 563
 564         DST1bpp1(0,0);
 565         DST2bpp1(0,0);
 566
 567         DST2bpp1(1,2);
 568         DST1bpp1(1,2);
 569
 570         DST1bpp1(2,4);
 571         DST2bpp1(2,4);
 572
 573         DST2bpp1(3,6);
 574         DST1bpp1(3,6);
 575
 576         dst_1[0]= out_1;
 577         dst_2[0]= out_2;
 578 EPILOG(1)
 579
 580 SwsFunc yuv2rgb_get_func_ptr (SwsContext *c)
 581 {
 582 #if defined(HAVE_MMX2) || defined(HAVE_MMX)
 583     if(c->flags & SWS_CPU_CAPS_MMX2){
 584         switch(c->dstFormat){
 585         case PIX_FMT_RGB32: return yuv420_rgb32_MMX2;
 586         case PIX_FMT_BGR24: return yuv420_rgb24_MMX2;
 587         case PIX_FMT_BGR565: return yuv420_rgb16_MMX2;
 588         case PIX_FMT_BGR555: return yuv420_rgb15_MMX2;
 589         }
 590     }
 591     if(c->flags & SWS_CPU_CAPS_MMX){
 592         switch(c->dstFormat){
 593         case PIX_FMT_RGB32: return yuv420_rgb32_MMX;
 594         case PIX_FMT_BGR24: return yuv420_rgb24_MMX;
 595         case PIX_FMT_BGR565: return yuv420_rgb16_MMX;
 596         case PIX_FMT_BGR555: return yuv420_rgb15_MMX;
 597         }
 598     }
 599 #endif
 600 #ifdef HAVE_MLIB
 601     {
 602         SwsFunc t= yuv2rgb_init_mlib(c);
 603         if(t) return t;
 604     }
 605 #endif
 606 #ifdef HAVE_ALTIVEC
 607     if (c->flags & SWS_CPU_CAPS_ALTIVEC)
 608     {
 609         SwsFunc t = yuv2rgb_init_altivec(c);
 610         if(t) return t;
 611     }
 612 #endif
 613
 614     av_log(c, AV_LOG_WARNING, "No accelerated colorspace conversion found\n");
 615
 616     switch(c->dstFormat){
 617     case PIX_FMT_BGR32:
 618     case PIX_FMT_RGB32: return yuv2rgb_c_32;
 619     case PIX_FMT_RGB24: return yuv2rgb_c_24_rgb;
 620     case PIX_FMT_BGR24: return yuv2rgb_c_24_bgr;
 621     case PIX_FMT_RGB565:
 622     case PIX_FMT_BGR565:
 623     case PIX_FMT_RGB555:
 624     case PIX_FMT_BGR555: return yuv2rgb_c_16;
 625     case PIX_FMT_RGB8:
 626     case PIX_FMT_BGR8:  return yuv2rgb_c_8_ordered_dither;
 627     case PIX_FMT_RGB4:
 628     case PIX_FMT_BGR4:  return yuv2rgb_c_4_ordered_dither;
 629     case PIX_FMT_RGB4_BYTE:
 630     case PIX_FMT_BGR4_BYTE:  return yuv2rgb_c_4b_ordered_dither;
 631     case PIX_FMT_MONOBLACK:  return yuv2rgb_c_1_ordered_dither;
 632     default:
 633         assert(0);
 634     }
 635     return NULL;
 636 }
 637
 638 static int div_round (int dividend, int divisor)
 639 {
 640     if (dividend > 0)
 641         return (dividend + (divisor>>1)) / divisor;
 642     else
 643         return -((-dividend + (divisor>>1)) / divisor);
 644 }
 645
 646 int yuv2rgb_c_init_tables (SwsContext *c, const int inv_table[4], int fullRange, int brightness, int contrast, int saturation)
 647 {
 648     const int isRgb = isBGR(c->dstFormat);
 649     const int bpp = fmt_depth(c->dstFormat);
 650     int i;
 651     uint8_t table_Y[1024];
 652     uint32_t *table_32 = 0;
 653     uint16_t *table_16 = 0;
 654     uint8_t *table_8 = 0;
 655     uint8_t *table_332 = 0;
 656     uint8_t *table_121 = 0;
 657     uint8_t *table_1 = 0;
 658     int entry_size = 0;
 659     void *table_r = 0, *table_g = 0, *table_b = 0;
 660     void *table_start;
 661
 662     int64_t crv =  inv_table[0];
 663     int64_t cbu =  inv_table[1];
 664     int64_t cgu = -inv_table[2];
 665     int64_t cgv = -inv_table[3];
 666     int64_t cy  = 1<<16;
 667     int64_t oy  = 0;
 668
 669 //printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
 670     if(!fullRange){
 671         cy= (cy*255) / 219;
 672         oy= 16<<16;
 673     }else{
 674         crv= (crv*224) / 255;
 675         cbu= (cbu*224) / 255;
 676         cgu= (cgu*224) / 255;
 677         cgv= (cgv*224) / 255;
 678     }
 679
 680     cy = (cy *contrast             )>>16;
 681     crv= (crv*contrast * saturation)>>32;
 682     cbu= (cbu*contrast * saturation)>>32;
 683     cgu= (cgu*contrast * saturation)>>32;
 684     cgv= (cgv*contrast * saturation)>>32;
 685 //printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
 686     oy -= 256*brightness;
 687
 688     for (i = 0; i < 1024; i++) {
 689         int j;
 690
 691         j= (cy*(((i - 384)<<16) - oy) + (1<<31))>>32;
 692         j = (j < 0) ? 0 : ((j > 255) ? 255 : j);
 693         table_Y[i] = j;
 694     }
 695
 696     switch (bpp) {
 697     case 32:
 698         table_start= table_32 = av_malloc ((197 + 2*682 + 256 + 132) * sizeof (uint32_t));
 699
 700         entry_size = sizeof (uint32_t);
 701         table_r = table_32 + 197;
 702         table_b = table_32 + 197 + 685;
 703         table_g = table_32 + 197 + 2*682;
 704
 705         for (i = -197; i < 256+197; i++)
 706             ((uint32_t *)table_r)[i] = table_Y[i+384] << (isRgb ? 16 : 0);
 707         for (i = -132; i < 256+132; i++)
 708             ((uint32_t *)table_g)[i] = table_Y[i+384] << 8;
 709         for (i = -232; i < 256+232; i++)
 710             ((uint32_t *)table_b)[i] = table_Y[i+384] << (isRgb ? 0 : 16);
 711         break;
 712
 713     case 24:
 714         table_start= table_8 = av_malloc ((256 + 2*232) * sizeof (uint8_t));
 715
 716         entry_size = sizeof (uint8_t);
 717         table_r = table_g = table_b = table_8 + 232;
 718
 719         for (i = -232; i < 256+232; i++)
 720             ((uint8_t * )table_b)[i] = table_Y[i+384];
 721         break;
 722
 723     case 15:
 724     case 16:
 725         table_start= table_16 = av_malloc ((197 + 2*682 + 256 + 132) * sizeof (uint16_t));
 726
 727         entry_size = sizeof (uint16_t);
 728         table_r = table_16 + 197;
 729         table_b = table_16 + 197 + 685;
 730         table_g = table_16 + 197 + 2*682;
 731
 732         for (i = -197; i < 256+197; i++) {
 733             int j = table_Y[i+384] >> 3;
 734
 735             if (isRgb)
 736                 j <<= ((bpp==16) ? 11 : 10);
 737
 738             ((uint16_t *)table_r)[i] = j;
 739         }
 740         for (i = -132; i < 256+132; i++) {
 741             int j = table_Y[i+384] >> ((bpp==16) ? 2 : 3);
 742
 743             ((uint16_t *)table_g)[i] = j << 5;
 744         }
 745         for (i = -232; i < 256+232; i++) {
 746             int j = table_Y[i+384] >> 3;
 747
 748             if (!isRgb)
 749                 j <<= ((bpp==16) ? 11 : 10);
 750
 751             ((uint16_t *)table_b)[i] = j;
 752         }
 753         break;
 754
 755     case 8:
 756         table_start= table_332 = av_malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t));
 757
 758         entry_size = sizeof (uint8_t);
 759         table_r = table_332 + 197;
 760         table_b = table_332 + 197 + 685;
 761         table_g = table_332 + 197 + 2*682;
 762
 763         for (i = -197; i < 256+197; i++) {
 764             int j = (table_Y[i+384 - 16] + 18)/36;
 765
 766             if (isRgb)
 767                 j <<= 5;
 768
 769             ((uint8_t *)table_r)[i] = j;
 770         }
 771         for (i = -132; i < 256+132; i++) {
 772             int j = (table_Y[i+384 - 16] + 18)/36;
 773
 774             if (!isRgb)
 775                 j <<= 1;
 776
 777             ((uint8_t *)table_g)[i] = j << 2;
 778         }
 779         for (i = -232; i < 256+232; i++) {
 780             int j = (table_Y[i+384 - 37] + 43)/85;
 781
 782             if (!isRgb)
 783                 j <<= 6;
 784
 785             ((uint8_t *)table_b)[i] = j;
 786         }
 787         break;
 788     case 4:
 789     case 4|128:
 790         table_start= table_121 = av_malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t));
 791
 792         entry_size = sizeof (uint8_t);
 793         table_r = table_121 + 197;
 794         table_b = table_121 + 197 + 685;
 795         table_g = table_121 + 197 + 2*682;
 796
 797         for (i = -197; i < 256+197; i++) {
 798             int j = table_Y[i+384 - 110] >> 7;
 799
 800             if (isRgb)
 801                 j <<= 3;
 802
 803             ((uint8_t *)table_r)[i] = j;
 804         }
 805         for (i = -132; i < 256+132; i++) {
 806             int j = (table_Y[i+384 - 37]+ 43)/85;
 807
 808             ((uint8_t *)table_g)[i] = j << 1;
 809         }
 810         for (i = -232; i < 256+232; i++) {
 811             int j =table_Y[i+384 - 110] >> 7;
 812
 813             if (!isRgb)
 814                 j <<= 3;
 815
 816             ((uint8_t *)table_b)[i] = j;
 817         }
 818         break;
 819
 820     case 1:
 821         table_start= table_1 = av_malloc (256*2 * sizeof (uint8_t));
 822
 823         entry_size = sizeof (uint8_t);
 824         table_g = table_1;
 825         table_r = table_b = NULL;
 826
 827         for (i = 0; i < 256+256; i++) {
 828             int j = table_Y[i + 384 - 110]>>7;
 829
 830             ((uint8_t *)table_g)[i] = j;
 831         }
 832         break;
 833
 834     default:
 835         table_start= NULL;
 836         av_log(c, AV_LOG_ERROR, "%ibpp not supported by yuv2rgb\n", bpp);
 837         //free mem?
 838         return -1;
 839     }
 840
 841     for (i = 0; i < 256; i++) {
 842         c->table_rV[i] = (uint8_t *)table_r + entry_size * div_round (crv * (i-128), 76309);
 843         c->table_gU[i] = (uint8_t *)table_g + entry_size * div_round (cgu * (i-128), 76309);
 844         c->table_gV[i] = entry_size * div_round (cgv * (i-128), 76309);
 845         c->table_bU[i] = (uint8_t *)table_b + entry_size * div_round (cbu * (i-128), 76309);
 846     }
 847
 848     av_free(c->yuvTable);
 849     c->yuvTable= table_start;
 850     return 0;
 851 }