libswscale/yuv2rgb.c

   1 /*
   2  * yuv2rgb.c, Software YUV to RGB converter
   3  *
   4  *  Copyright (C) 1999, Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
   5  *
   6  *  Functions broken out from display_x11.c and several new modes
   7  *  added by Håkan Hjort <d95hjort@dtek.chalmers.se>
   8  *
   9  *  15 & 16 bpp support by Franck Sicard <Franck.Sicard@solsoft.fr>
  10  *
  11  *  MMX/MMX2 template stuff (needed for fast movntq support),
  12  *  1,4,8bpp support and context / deglobalize stuff
  13  *  by Michael Niedermayer (michaelni@gmx.at)
  14  *
  15  *  This file is part of mpeg2dec, a free MPEG-2 video decoder
  16  *
  17  *  mpeg2dec is free software; you can redistribute it and/or modify
  18  *  it under the terms of the GNU General Public License as published by
  19  *  the Free Software Foundation; either version 2, or (at your option)
  20  *  any later version.
  21  *
  22  *  mpeg2dec is distributed in the hope that it will be useful,
  23  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  24  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  25  *  GNU General Public License for more details.
  26  *
  27  *  You should have received a copy of the GNU General Public License
  28  *  along with mpeg2dec; if not, write to the Free Software
  29  *  Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  30  */
  31
  32 #include <stdio.h>
  33 #include <stdlib.h>
  34 #include <inttypes.h>
  35 #include <assert.h>
  36
  37 #include "config.h"
  38 #include "rgb2rgb.h"
  39 #include "swscale.h"
  40 #include "swscale_internal.h"
  41
  42 #ifdef HAVE_VIS
  43 #include "yuv2rgb_vis.c"
  44 #endif
  45
  46 #ifdef HAVE_MLIB
  47 #include "yuv2rgb_mlib.c"
  48 #endif
  49
  50 #define DITHER1XBPP // only for mmx
  51
  52 const uint8_t  __attribute__((aligned(8))) dither_2x2_4[2][8]={
  53 {  1,   3,   1,   3,   1,   3,   1,   3, },
  54 {  2,   0,   2,   0,   2,   0,   2,   0, },
  55 };
  56
  57 const uint8_t  __attribute__((aligned(8))) dither_2x2_8[2][8]={
  58 {  6,   2,   6,   2,   6,   2,   6,   2, },
  59 {  0,   4,   0,   4,   0,   4,   0,   4, },
  60 };
  61
  62 const uint8_t  __attribute__((aligned(8))) dither_8x8_32[8][8]={
  63 { 17,   9,  23,  15,  16,   8,  22,  14, },
  64 {  5,  29,   3,  27,   4,  28,   2,  26, },
  65 { 21,  13,  19,  11,  20,  12,  18,  10, },
  66 {  0,  24,   6,  30,   1,  25,   7,  31, },
  67 { 16,   8,  22,  14,  17,   9,  23,  15, },
  68 {  4,  28,   2,  26,   5,  29,   3,  27, },
  69 { 20,  12,  18,  10,  21,  13,  19,  11, },
  70 {  1,  25,   7,  31,   0,  24,   6,  30, },
  71 };
  72
  73 #if 0
  74 const uint8_t  __attribute__((aligned(8))) dither_8x8_64[8][8]={
  75 {  0,  48,  12,  60,   3,  51,  15,  63, },
  76 { 32,  16,  44,  28,  35,  19,  47,  31, },
  77 {  8,  56,   4,  52,  11,  59,   7,  55, },
  78 { 40,  24,  36,  20,  43,  27,  39,  23, },
  79 {  2,  50,  14,  62,   1,  49,  13,  61, },
  80 { 34,  18,  46,  30,  33,  17,  45,  29, },
  81 { 10,  58,   6,  54,   9,  57,   5,  53, },
  82 { 42,  26,  38,  22,  41,  25,  37,  21, },
  83 };
  84 #endif
  85
  86 const uint8_t  __attribute__((aligned(8))) dither_8x8_73[8][8]={
  87 {  0,  55,  14,  68,   3,  58,  17,  72, },
  88 { 37,  18,  50,  32,  40,  22,  54,  35, },
  89 {  9,  64,   5,  59,  13,  67,   8,  63, },
  90 { 46,  27,  41,  23,  49,  31,  44,  26, },
  91 {  2,  57,  16,  71,   1,  56,  15,  70, },
  92 { 39,  21,  52,  34,  38,  19,  51,  33, },
  93 { 11,  66,   7,  62,  10,  65,   6,  60, },
  94 { 48,  30,  43,  25,  47,  29,  42,  24, },
  95 };
  96
  97 #if 0
  98 const uint8_t  __attribute__((aligned(8))) dither_8x8_128[8][8]={
  99 { 68,  36,  92,  60,  66,  34,  90,  58, },
 100 { 20, 116,  12, 108,  18, 114,  10, 106, },
 101 { 84,  52,  76,  44,  82,  50,  74,  42, },
 102 {  0,  96,  24, 120,   6, 102,  30, 126, },
 103 { 64,  32,  88,  56,  70,  38,  94,  62, },
 104 { 16, 112,   8, 104,  22, 118,  14, 110, },
 105 { 80,  48,  72,  40,  86,  54,  78,  46, },
 106 {  4, 100,  28, 124,   2,  98,  26, 122, },
 107 };
 108 #endif
 109
 110 #if 1
 111 const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
 112 {117,  62, 158, 103, 113,  58, 155, 100, },
 113 { 34, 199,  21, 186,  31, 196,  17, 182, },
 114 {144,  89, 131,  76, 141,  86, 127,  72, },
 115 {  0, 165,  41, 206,  10, 175,  52, 217, },
 116 {110,  55, 151,  96, 120,  65, 162, 107, },
 117 { 28, 193,  14, 179,  38, 203,  24, 189, },
 118 {138,  83, 124,  69, 148,  93, 134,  79, },
 119 {  7, 172,  48, 213,   3, 168,  45, 210, },
 120 };
 121 #elif 1
 122 // tries to correct a gamma of 1.5
 123 const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
 124 {  0, 143,  18, 200,   2, 156,  25, 215, },
 125 { 78,  28, 125,  64,  89,  36, 138,  74, },
 126 { 10, 180,   3, 161,  16, 195,   8, 175, },
 127 {109,  51,  93,  38, 121,  60, 105,  47, },
 128 {  1, 152,  23, 210,   0, 147,  20, 205, },
 129 { 85,  33, 134,  71,  81,  30, 130,  67, },
 130 { 14, 190,   6, 171,  12, 185,   5, 166, },
 131 {117,  57, 101,  44, 113,  54,  97,  41, },
 132 };
 133 #elif 1
 134 // tries to correct a gamma of 2.0
 135 const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
 136 {  0, 124,   8, 193,   0, 140,  12, 213, },
 137 { 55,  14, 104,  42,  66,  19, 119,  52, },
 138 {  3, 168,   1, 145,   6, 187,   3, 162, },
 139 { 86,  31,  70,  21,  99,  39,  82,  28, },
 140 {  0, 134,  11, 206,   0, 129,   9, 200, },
 141 { 62,  17, 114,  48,  58,  16, 109,  45, },
 142 {  5, 181,   2, 157,   4, 175,   1, 151, },
 143 { 95,  36,  78,  26,  90,  34,  74,  24, },
 144 };
 145 #else
 146 // tries to correct a gamma of 2.5
 147 const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
 148 {  0, 107,   3, 187,   0, 125,   6, 212, },
 149 { 39,   7,  86,  28,  49,  11, 102,  36, },
 150 {  1, 158,   0, 131,   3, 180,   1, 151, },
 151 { 68,  19,  52,  12,  81,  25,  64,  17, },
 152 {  0, 119,   5, 203,   0, 113,   4, 195, },
 153 { 45,   9,  96,  33,  42,   8,  91,  30, },
 154 {  2, 172,   1, 144,   2, 165,   0, 137, },
 155 { 77,  23,  60,  15,  72,  21,  56,  14, },
 156 };
 157 #endif
 158
 159 #ifdef HAVE_MMX
 160
 161 /* hope these constant values are cache line aligned */
 162 static uint64_t attribute_used __attribute__((aligned(8))) mmx_00ffw   = 0x00ff00ff00ff00ffULL;
 163 static uint64_t attribute_used __attribute__((aligned(8))) mmx_redmask = 0xf8f8f8f8f8f8f8f8ULL;
 164 static uint64_t attribute_used __attribute__((aligned(8))) mmx_grnmask = 0xfcfcfcfcfcfcfcfcULL;
 165
 166 static uint64_t attribute_used __attribute__((aligned(8))) M24A=   0x00FF0000FF0000FFULL;
 167 static uint64_t attribute_used __attribute__((aligned(8))) M24B=   0xFF0000FF0000FF00ULL;
 168 static uint64_t attribute_used __attribute__((aligned(8))) M24C=   0x0000FF0000FF0000ULL;
 169
 170 // the volatile is required because gcc otherwise optimizes some writes away not knowing that these
 171 // are read in the asm block
 172 static volatile uint64_t attribute_used __attribute__((aligned(8))) b5Dither;
 173 static volatile uint64_t attribute_used __attribute__((aligned(8))) g5Dither;
 174 static volatile uint64_t attribute_used __attribute__((aligned(8))) g6Dither;
 175 static volatile uint64_t attribute_used __attribute__((aligned(8))) r5Dither;
 176
 177 static uint64_t __attribute__((aligned(8))) dither4[2]={
 178     0x0103010301030103LL,
 179     0x0200020002000200LL,};
 180
 181 static uint64_t __attribute__((aligned(8))) dither8[2]={
 182     0x0602060206020602LL,
 183     0x0004000400040004LL,};
 184
 185 #undef HAVE_MMX
 186
 187 //MMX versions
 188 #undef RENAME
 189 #define HAVE_MMX
 190 #undef HAVE_MMX2
 191 #undef HAVE_3DNOW
 192 #define RENAME(a) a ## _MMX
 193 #include "yuv2rgb_template.c"
 194
 195 //MMX2 versions
 196 #undef RENAME
 197 #define HAVE_MMX
 198 #define HAVE_MMX2
 199 #undef HAVE_3DNOW
 200 #define RENAME(a) a ## _MMX2
 201 #include "yuv2rgb_template.c"
 202
 203 #endif /* defined(ARCH_X86) */
 204
 205 const int32_t Inverse_Table_6_9[8][4] = {
 206     {117504, 138453, 13954, 34903}, /* no sequence_display_extension */
 207     {117504, 138453, 13954, 34903}, /* ITU-R Rec. 709 (1990) */
 208     {104597, 132201, 25675, 53279}, /* unspecified */
 209     {104597, 132201, 25675, 53279}, /* reserved */
 210     {104448, 132798, 24759, 53109}, /* FCC */
 211     {104597, 132201, 25675, 53279}, /* ITU-R Rec. 624-4 System B, G */
 212     {104597, 132201, 25675, 53279}, /* SMPTE 170M */
 213     {117579, 136230, 16907, 35559}  /* SMPTE 240M (1987) */
 214 };
 215
 216 #define RGB(i)                                      \
 217     U = pu[i];                                      \
 218     V = pv[i];                                      \
 219     r = (void *)c->table_rV[V];                     \
 220     g = (void *)(c->table_gU[U] + c->table_gV[V]);  \
 221     b = (void *)c->table_bU[U];
 222
 223 #define DST1(i)                         \
 224     Y = py_1[2*i];                      \
 225     dst_1[2*i] = r[Y] + g[Y] + b[Y];    \
 226     Y = py_1[2*i+1];                    \
 227     dst_1[2*i+1] = r[Y] + g[Y] + b[Y];
 228
 229 #define DST2(i)                         \
 230     Y = py_2[2*i];                      \
 231     dst_2[2*i] = r[Y] + g[Y] + b[Y];    \
 232     Y = py_2[2*i+1];                    \
 233     dst_2[2*i+1] = r[Y] + g[Y] + b[Y];
 234
 235 #define DST1RGB(i)                                                \
 236     Y = py_1[2*i];                                                \
 237     dst_1[6*i] = r[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = b[Y];  \
 238     Y = py_1[2*i+1];                                              \
 239     dst_1[6*i+3] = r[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = b[Y];
 240
 241 #define DST2RGB(i)                                                \
 242     Y = py_2[2*i];                                                \
 243     dst_2[6*i] = r[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = b[Y];  \
 244     Y = py_2[2*i+1];                                              \
 245     dst_2[6*i+3] = r[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = b[Y];
 246
 247 #define DST1BGR(i)                                                \
 248     Y = py_1[2*i];                                                \
 249     dst_1[6*i] = b[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = r[Y];  \
 250     Y = py_1[2*i+1];                                              \
 251     dst_1[6*i+3] = b[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = r[Y];
 252
 253 #define DST2BGR(i)                                                \
 254     Y = py_2[2*i];                                                \
 255     dst_2[6*i] = b[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = r[Y];  \
 256     Y = py_2[2*i+1];                                              \
 257     dst_2[6*i+3] = b[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = r[Y];
 258
 259 #define PROLOG(func_name, dst_type) \
 260 static int func_name(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, \
 261                      int srcSliceH, uint8_t* dst[], int dstStride[]){\
 262     int y;\
 263 \
 264     if (c->srcFormat == PIX_FMT_YUV422P){\
 265         srcStride[1] *= 2;\
 266         srcStride[2] *= 2;\
 267     }\
 268     for (y=0; y<srcSliceH; y+=2){\
 269         dst_type *dst_1= (dst_type*)(dst[0] + (y+srcSliceY  )*dstStride[0]);\
 270         dst_type *dst_2= (dst_type*)(dst[0] + (y+srcSliceY+1)*dstStride[0]);\
 271         dst_type av_unused *r, *b;\
 272         dst_type *g;\
 273         uint8_t *py_1= src[0] + y*srcStride[0];\
 274         uint8_t *py_2= py_1 + srcStride[0];\
 275         uint8_t *pu= src[1] + (y>>1)*srcStride[1];\
 276         uint8_t *pv= src[2] + (y>>1)*srcStride[2];\
 277         unsigned int h_size= c->dstW>>3;\
 278         while (h_size--) {\
 279             int av_unused U, V;\
 280             int Y;\
 281
 282 #define EPILOG1(dst_delta)\
 283             pu += 4;\
 284             pv += 4;\
 285             py_1 += 8;\
 286             py_2 += 8;\
 287             dst_1 += dst_delta;\
 288             dst_2 += dst_delta;\
 289         }\
 290         if (c->dstW & 4) {\
 291             int av_unused U, V;\
 292             int Y;\
 293
 294 #define EPILOG2()\
 295         }\
 296     }\
 297     return srcSliceH;\
 298 }
 299
 300 #define EPILOG(dst_delta)\
 301     EPILOG1(dst_delta)\
 302     EPILOG2()
 303
 304 PROLOG(yuv2rgb_c_32, uint32_t)
 305     RGB(0);
 306     DST1(0);
 307     DST2(0);
 308
 309     RGB(1);
 310     DST2(1);
 311     DST1(1);
 312
 313     RGB(2);
 314     DST1(2);
 315     DST2(2);
 316
 317     RGB(3);
 318     DST2(3);
 319     DST1(3);
 320 EPILOG1(8)
 321     RGB(0);
 322     DST1(0);
 323     DST2(0);
 324
 325     RGB(1);
 326     DST2(1);
 327     DST1(1);
 328 EPILOG2()
 329
 330 PROLOG(yuv2rgb_c_24_rgb, uint8_t)
 331     RGB(0);
 332     DST1RGB(0);
 333     DST2RGB(0);
 334
 335     RGB(1);
 336     DST2RGB(1);
 337     DST1RGB(1);
 338
 339     RGB(2);
 340     DST1RGB(2);
 341     DST2RGB(2);
 342
 343     RGB(3);
 344     DST2RGB(3);
 345     DST1RGB(3);
 346 EPILOG1(24)
 347     RGB(0);
 348     DST1RGB(0);
 349     DST2RGB(0);
 350
 351     RGB(1);
 352     DST2RGB(1);
 353     DST1RGB(1);
 354 EPILOG2()
 355
 356 // only trivial mods from yuv2rgb_c_24_rgb
 357 PROLOG(yuv2rgb_c_24_bgr, uint8_t)
 358     RGB(0);
 359     DST1BGR(0);
 360     DST2BGR(0);
 361
 362     RGB(1);
 363     DST2BGR(1);
 364     DST1BGR(1);
 365
 366     RGB(2);
 367     DST1BGR(2);
 368     DST2BGR(2);
 369
 370     RGB(3);
 371     DST2BGR(3);
 372     DST1BGR(3);
 373 EPILOG1(24)
 374     RGB(0);
 375     DST1BGR(0);
 376     DST2BGR(0);
 377
 378     RGB(1);
 379     DST2BGR(1);
 380     DST1BGR(1);
 381 EPILOG2()
 382
 383 // This is exactly the same code as yuv2rgb_c_32 except for the types of
 384 // r, g, b, dst_1, dst_2
 385 PROLOG(yuv2rgb_c_16, uint16_t)
 386     RGB(0);
 387     DST1(0);
 388     DST2(0);
 389
 390     RGB(1);
 391     DST2(1);
 392     DST1(1);
 393
 394     RGB(2);
 395     DST1(2);
 396     DST2(2);
 397
 398     RGB(3);
 399     DST2(3);
 400     DST1(3);
 401 EPILOG(8)
 402
 403 // This is exactly the same code as yuv2rgb_c_32 except for the types of
 404 // r, g, b, dst_1, dst_2
 405 PROLOG(yuv2rgb_c_8, uint8_t)
 406     RGB(0);
 407     DST1(0);
 408     DST2(0);
 409
 410     RGB(1);
 411     DST2(1);
 412     DST1(1);
 413
 414     RGB(2);
 415     DST1(2);
 416     DST2(2);
 417
 418     RGB(3);
 419     DST2(3);
 420     DST1(3);
 421 EPILOG(8)
 422
 423 // r, g, b, dst_1, dst_2
 424 PROLOG(yuv2rgb_c_8_ordered_dither, uint8_t)
 425     const uint8_t *d32= dither_8x8_32[y&7];
 426     const uint8_t *d64= dither_8x8_73[y&7];
 427 #define DST1bpp8(i,o)                                               \
 428     Y = py_1[2*i];                                                  \
 429     dst_1[2*i]   = r[Y+d32[0+o]] + g[Y+d32[0+o]] + b[Y+d64[0+o]];   \
 430     Y = py_1[2*i+1];                                                \
 431     dst_1[2*i+1] = r[Y+d32[1+o]] + g[Y+d32[1+o]] + b[Y+d64[1+o]];
 432
 433 #define DST2bpp8(i,o)                                               \
 434     Y = py_2[2*i];                                                  \
 435     dst_2[2*i]   =  r[Y+d32[8+o]] + g[Y+d32[8+o]] + b[Y+d64[8+o]];  \
 436     Y = py_2[2*i+1];                                                \
 437     dst_2[2*i+1] =  r[Y+d32[9+o]] + g[Y+d32[9+o]] + b[Y+d64[9+o]];
 438
 439
 440     RGB(0);
 441     DST1bpp8(0,0);
 442     DST2bpp8(0,0);
 443
 444     RGB(1);
 445     DST2bpp8(1,2);
 446     DST1bpp8(1,2);
 447
 448     RGB(2);
 449     DST1bpp8(2,4);
 450     DST2bpp8(2,4);
 451
 452     RGB(3);
 453     DST2bpp8(3,6);
 454     DST1bpp8(3,6);
 455 EPILOG(8)
 456
 457
 458 // This is exactly the same code as yuv2rgb_c_32 except for the types of
 459 // r, g, b, dst_1, dst_2
 460 PROLOG(yuv2rgb_c_4, uint8_t)
 461     int acc;
 462 #define DST1_4(i)                   \
 463     Y = py_1[2*i];                  \
 464     acc = r[Y] + g[Y] + b[Y];       \
 465     Y = py_1[2*i+1];                \
 466     acc |= (r[Y] + g[Y] + b[Y])<<4; \
 467     dst_1[i] = acc;
 468
 469 #define DST2_4(i)                   \
 470     Y = py_2[2*i];                  \
 471     acc = r[Y] + g[Y] + b[Y];       \
 472     Y = py_2[2*i+1];                \
 473     acc |= (r[Y] + g[Y] + b[Y])<<4; \
 474     dst_2[i] = acc;
 475
 476     RGB(0);
 477     DST1_4(0);
 478     DST2_4(0);
 479
 480     RGB(1);
 481     DST2_4(1);
 482     DST1_4(1);
 483
 484     RGB(2);
 485     DST1_4(2);
 486     DST2_4(2);
 487
 488     RGB(3);
 489     DST2_4(3);
 490     DST1_4(3);
 491 EPILOG(4)
 492
 493 PROLOG(yuv2rgb_c_4_ordered_dither, uint8_t)
 494     const uint8_t *d64= dither_8x8_73[y&7];
 495     const uint8_t *d128=dither_8x8_220[y&7];
 496     int acc;
 497
 498 #define DST1bpp4(i,o)                                             \
 499     Y = py_1[2*i];                                                \
 500     acc = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]];        \
 501     Y = py_1[2*i+1];                                              \
 502     acc |= (r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]])<<4;  \
 503     dst_1[i]= acc;
 504
 505 #define DST2bpp4(i,o)                                             \
 506     Y = py_2[2*i];                                                \
 507     acc =  r[Y+d128[8+o]] + g[Y+d64[8+o]] + b[Y+d128[8+o]];       \
 508     Y = py_2[2*i+1];                                              \
 509     acc |=  (r[Y+d128[9+o]] + g[Y+d64[9+o]] + b[Y+d128[9+o]])<<4; \
 510     dst_2[i]= acc;
 511
 512
 513     RGB(0);
 514     DST1bpp4(0,0);
 515     DST2bpp4(0,0);
 516
 517     RGB(1);
 518     DST2bpp4(1,2);
 519     DST1bpp4(1,2);
 520
 521     RGB(2);
 522     DST1bpp4(2,4);
 523     DST2bpp4(2,4);
 524
 525     RGB(3);
 526     DST2bpp4(3,6);
 527     DST1bpp4(3,6);
 528 EPILOG(4)
 529
 530 // This is exactly the same code as yuv2rgb_c_32 except for the types of
 531 // r, g, b, dst_1, dst_2
 532 PROLOG(yuv2rgb_c_4b, uint8_t)
 533     RGB(0);
 534     DST1(0);
 535     DST2(0);
 536
 537     RGB(1);
 538     DST2(1);
 539     DST1(1);
 540
 541     RGB(2);
 542     DST1(2);
 543     DST2(2);
 544
 545     RGB(3);
 546     DST2(3);
 547     DST1(3);
 548 EPILOG(8)
 549
 550 PROLOG(yuv2rgb_c_4b_ordered_dither, uint8_t)
 551     const uint8_t *d64= dither_8x8_73[y&7];
 552     const uint8_t *d128=dither_8x8_220[y&7];
 553
 554 #define DST1bpp4b(i,o)                                                \
 555     Y = py_1[2*i];                                                    \
 556     dst_1[2*i]   = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]];   \
 557     Y = py_1[2*i+1];                                                  \
 558     dst_1[2*i+1] = r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]];
 559
 560 #define DST2bpp4b(i,o)                                                \
 561     Y = py_2[2*i];                                                    \
 562     dst_2[2*i]   =  r[Y+d128[8+o]] + g[Y+d64[8+o]] + b[Y+d128[8+o]];  \
 563     Y = py_2[2*i+1];                                                  \
 564     dst_2[2*i+1] =  r[Y+d128[9+o]] + g[Y+d64[9+o]] + b[Y+d128[9+o]];
 565
 566
 567     RGB(0);
 568     DST1bpp4b(0,0);
 569     DST2bpp4b(0,0);
 570
 571     RGB(1);
 572     DST2bpp4b(1,2);
 573     DST1bpp4b(1,2);
 574
 575     RGB(2);
 576     DST1bpp4b(2,4);
 577     DST2bpp4b(2,4);
 578
 579     RGB(3);
 580     DST2bpp4b(3,6);
 581     DST1bpp4b(3,6);
 582 EPILOG(8)
 583
 584 PROLOG(yuv2rgb_c_1_ordered_dither, uint8_t)
 585         const uint8_t *d128=dither_8x8_220[y&7];
 586         char out_1=0, out_2=0;
 587         g= c->table_gU[128] + c->table_gV[128];
 588
 589 #define DST1bpp1(i,o)               \
 590     Y = py_1[2*i];                  \
 591     out_1+= out_1 + g[Y+d128[0+o]]; \
 592     Y = py_1[2*i+1];                \
 593     out_1+= out_1 + g[Y+d128[1+o]];
 594
 595 #define DST2bpp1(i,o)               \
 596     Y = py_2[2*i];                  \
 597     out_2+= out_2 + g[Y+d128[8+o]]; \
 598     Y = py_2[2*i+1];                \
 599     out_2+= out_2 + g[Y+d128[9+o]];
 600
 601     DST1bpp1(0,0);
 602     DST2bpp1(0,0);
 603
 604     DST2bpp1(1,2);
 605     DST1bpp1(1,2);
 606
 607     DST1bpp1(2,4);
 608     DST2bpp1(2,4);
 609
 610     DST2bpp1(3,6);
 611     DST1bpp1(3,6);
 612
 613     dst_1[0]= out_1;
 614     dst_2[0]= out_2;
 615 EPILOG(1)
 616
 617 SwsFunc yuv2rgb_get_func_ptr (SwsContext *c)
 618 {
 619 #if defined(HAVE_MMX2) || defined(HAVE_MMX)
 620     if (c->flags & SWS_CPU_CAPS_MMX2){
 621         switch(c->dstFormat){
 622         case PIX_FMT_RGB32:  return yuv420_rgb32_MMX2;
 623         case PIX_FMT_BGR24:  return yuv420_rgb24_MMX2;
 624         case PIX_FMT_BGR565: return yuv420_rgb16_MMX2;
 625         case PIX_FMT_BGR555: return yuv420_rgb15_MMX2;
 626         }
 627     }
 628     if (c->flags & SWS_CPU_CAPS_MMX){
 629         switch(c->dstFormat){
 630         case PIX_FMT_RGB32:  return yuv420_rgb32_MMX;
 631         case PIX_FMT_BGR24:  return yuv420_rgb24_MMX;
 632         case PIX_FMT_BGR565: return yuv420_rgb16_MMX;
 633         case PIX_FMT_BGR555: return yuv420_rgb15_MMX;
 634         }
 635     }
 636 #endif
 637 #ifdef HAVE_VIS
 638     {
 639         SwsFunc t= yuv2rgb_init_vis(c);
 640         if (t) return t;
 641     }
 642 #endif
 643 #ifdef HAVE_MLIB
 644     {
 645         SwsFunc t= yuv2rgb_init_mlib(c);
 646         if (t) return t;
 647     }
 648 #endif
 649 #ifdef HAVE_ALTIVEC
 650     if (c->flags & SWS_CPU_CAPS_ALTIVEC)
 651     {
 652         SwsFunc t = yuv2rgb_init_altivec(c);
 653         if (t) return t;
 654     }
 655 #endif
 656
 657 #ifdef ARCH_BFIN
 658     if (c->flags & SWS_CPU_CAPS_BFIN)
 659     {
 660         SwsFunc t = ff_bfin_yuv2rgb_get_func_ptr (c);
 661         if (t) return t;
 662     }
 663 #endif
 664
 665     av_log(c, AV_LOG_WARNING, "No accelerated colorspace conversion found\n");
 666
 667     switch(c->dstFormat){
 668     case PIX_FMT_BGR32:
 669     case PIX_FMT_RGB32: return yuv2rgb_c_32;
 670     case PIX_FMT_RGB24: return yuv2rgb_c_24_rgb;
 671     case PIX_FMT_BGR24: return yuv2rgb_c_24_bgr;
 672     case PIX_FMT_RGB565:
 673     case PIX_FMT_BGR565:
 674     case PIX_FMT_RGB555:
 675     case PIX_FMT_BGR555: return yuv2rgb_c_16;
 676     case PIX_FMT_RGB8:
 677     case PIX_FMT_BGR8:  return yuv2rgb_c_8_ordered_dither;
 678     case PIX_FMT_RGB4:
 679     case PIX_FMT_BGR4:  return yuv2rgb_c_4_ordered_dither;
 680     case PIX_FMT_RGB4_BYTE:
 681     case PIX_FMT_BGR4_BYTE:  return yuv2rgb_c_4b_ordered_dither;
 682     case PIX_FMT_MONOBLACK:  return yuv2rgb_c_1_ordered_dither;
 683     default:
 684         assert(0);
 685     }
 686     return NULL;
 687 }
 688
 689 static int div_round (int dividend, int divisor)
 690 {
 691     if (dividend > 0)
 692         return (dividend + (divisor>>1)) / divisor;
 693     else
 694         return -((-dividend + (divisor>>1)) / divisor);
 695 }
 696
 697 int yuv2rgb_c_init_tables (SwsContext *c, const int inv_table[4], int fullRange, int brightness, int contrast, int saturation)
 698 {
 699     const int isRgb = isBGR(c->dstFormat);
 700     const int bpp = fmt_depth(c->dstFormat);
 701     int i;
 702     uint8_t table_Y[1024];
 703     uint32_t *table_32 = 0;
 704     uint16_t *table_16 = 0;
 705     uint8_t *table_8 = 0;
 706     uint8_t *table_332 = 0;
 707     uint8_t *table_121 = 0;
 708     uint8_t *table_1 = 0;
 709     int entry_size = 0;
 710     void *table_r = 0, *table_g = 0, *table_b = 0;
 711     void *table_start;
 712
 713     int64_t crv =  inv_table[0];
 714     int64_t cbu =  inv_table[1];
 715     int64_t cgu = -inv_table[2];
 716     int64_t cgv = -inv_table[3];
 717     int64_t cy  = 1<<16;
 718     int64_t oy  = 0;
 719
 720 //printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
 721     if (!fullRange){
 722         cy= (cy*255) / 219;
 723         oy= 16<<16;
 724     }else{
 725         crv= (crv*224) / 255;
 726         cbu= (cbu*224) / 255;
 727         cgu= (cgu*224) / 255;
 728         cgv= (cgv*224) / 255;
 729     }
 730
 731     cy = (cy *contrast             )>>16;
 732     crv= (crv*contrast * saturation)>>32;
 733     cbu= (cbu*contrast * saturation)>>32;
 734     cgu= (cgu*contrast * saturation)>>32;
 735     cgv= (cgv*contrast * saturation)>>32;
 736 //printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
 737     oy -= 256*brightness;
 738
 739     for (i = 0; i < 1024; i++) {
 740         int j;
 741
 742         j= (cy*(((i - 384)<<16) - oy) + (1<<31))>>32;
 743         j = (j < 0) ? 0 : ((j > 255) ? 255 : j);
 744         table_Y[i] = j;
 745     }
 746
 747     switch (bpp) {
 748     case 32:
 749         table_start= table_32 = av_malloc ((197 + 2*682 + 256 + 132) * sizeof (uint32_t));
 750
 751         entry_size = sizeof (uint32_t);
 752         table_r = table_32 + 197;
 753         table_b = table_32 + 197 + 685;
 754         table_g = table_32 + 197 + 2*682;
 755
 756         for (i = -197; i < 256+197; i++)
 757             ((uint32_t *)table_r)[i] = table_Y[i+384] << (isRgb ? 16 : 0);
 758         for (i = -132; i < 256+132; i++)
 759             ((uint32_t *)table_g)[i] = table_Y[i+384] << 8;
 760         for (i = -232; i < 256+232; i++)
 761             ((uint32_t *)table_b)[i] = table_Y[i+384] << (isRgb ? 0 : 16);
 762         break;
 763
 764     case 24:
 765         table_start= table_8 = av_malloc ((256 + 2*232) * sizeof (uint8_t));
 766
 767         entry_size = sizeof (uint8_t);
 768         table_r = table_g = table_b = table_8 + 232;
 769
 770         for (i = -232; i < 256+232; i++)
 771             ((uint8_t * )table_b)[i] = table_Y[i+384];
 772         break;
 773
 774     case 15:
 775     case 16:
 776         table_start= table_16 = av_malloc ((197 + 2*682 + 256 + 132) * sizeof (uint16_t));
 777
 778         entry_size = sizeof (uint16_t);
 779         table_r = table_16 + 197;
 780         table_b = table_16 + 197 + 685;
 781         table_g = table_16 + 197 + 2*682;
 782
 783         for (i = -197; i < 256+197; i++) {
 784             int j = table_Y[i+384] >> 3;
 785
 786             if (isRgb)
 787                 j <<= ((bpp==16) ? 11 : 10);
 788
 789             ((uint16_t *)table_r)[i] = j;
 790         }
 791         for (i = -132; i < 256+132; i++) {
 792             int j = table_Y[i+384] >> ((bpp==16) ? 2 : 3);
 793
 794             ((uint16_t *)table_g)[i] = j << 5;
 795         }
 796         for (i = -232; i < 256+232; i++) {
 797             int j = table_Y[i+384] >> 3;
 798
 799             if (!isRgb)
 800                 j <<= ((bpp==16) ? 11 : 10);
 801
 802             ((uint16_t *)table_b)[i] = j;
 803         }
 804         break;
 805
 806     case 8:
 807         table_start= table_332 = av_malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t));
 808
 809         entry_size = sizeof (uint8_t);
 810         table_r = table_332 + 197;
 811         table_b = table_332 + 197 + 685;
 812         table_g = table_332 + 197 + 2*682;
 813
 814         for (i = -197; i < 256+197; i++) {
 815             int j = (table_Y[i+384 - 16] + 18)/36;
 816
 817             if (isRgb)
 818                 j <<= 5;
 819
 820             ((uint8_t *)table_r)[i] = j;
 821         }
 822         for (i = -132; i < 256+132; i++) {
 823             int j = (table_Y[i+384 - 16] + 18)/36;
 824
 825             if (!isRgb)
 826                 j <<= 1;
 827
 828             ((uint8_t *)table_g)[i] = j << 2;
 829         }
 830         for (i = -232; i < 256+232; i++) {
 831             int j = (table_Y[i+384 - 37] + 43)/85;
 832
 833             if (!isRgb)
 834                 j <<= 6;
 835
 836             ((uint8_t *)table_b)[i] = j;
 837         }
 838         break;
 839     case 4:
 840     case 4|128:
 841         table_start= table_121 = av_malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t));
 842
 843         entry_size = sizeof (uint8_t);
 844         table_r = table_121 + 197;
 845         table_b = table_121 + 197 + 685;
 846         table_g = table_121 + 197 + 2*682;
 847
 848         for (i = -197; i < 256+197; i++) {
 849             int j = table_Y[i+384 - 110] >> 7;
 850
 851             if (isRgb)
 852                 j <<= 3;
 853
 854             ((uint8_t *)table_r)[i] = j;
 855         }
 856         for (i = -132; i < 256+132; i++) {
 857             int j = (table_Y[i+384 - 37]+ 43)/85;
 858
 859             ((uint8_t *)table_g)[i] = j << 1;
 860         }
 861         for (i = -232; i < 256+232; i++) {
 862             int j =table_Y[i+384 - 110] >> 7;
 863
 864             if (!isRgb)
 865                 j <<= 3;
 866
 867             ((uint8_t *)table_b)[i] = j;
 868         }
 869         break;
 870
 871     case 1:
 872         table_start= table_1 = av_malloc (256*2 * sizeof (uint8_t));
 873
 874         entry_size = sizeof (uint8_t);
 875         table_g = table_1;
 876         table_r = table_b = NULL;
 877
 878         for (i = 0; i < 256+256; i++) {
 879             int j = table_Y[i + 384 - 110]>>7;
 880
 881             ((uint8_t *)table_g)[i] = j;
 882         }
 883         break;
 884
 885     default:
 886         table_start= NULL;
 887         av_log(c, AV_LOG_ERROR, "%ibpp not supported by yuv2rgb\n", bpp);
 888         //free mem?
 889         return -1;
 890     }
 891
 892     for (i = 0; i < 256; i++) {
 893         c->table_rV[i] = (uint8_t *)table_r + entry_size * div_round (crv * (i-128), 76309);
 894         c->table_gU[i] = (uint8_t *)table_g + entry_size * div_round (cgu * (i-128), 76309);
 895         c->table_gV[i] = entry_size * div_round (cgv * (i-128), 76309);
 896         c->table_bU[i] = (uint8_t *)table_b + entry_size * div_round (cbu * (i-128), 76309);
 897     }
 898
 899     av_free(c->yuvTable);
 900     c->yuvTable= table_start;
 901     return 0;
 902 }