libavcodec/h264pred.c

   1 /*
   2  * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
   3  * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 /**
  23  * @file h264pred.c
  24  * H.264 / AVC / MPEG4 part10 prediction functions.
  25  * @author Michael Niedermayer <michaelni@gmx.at>
  26  */
  27
  28 #include "avcodec.h"
  29 #include "mpegvideo.h"
  30 #include "h264pred.h"
  31
  32 static void pred4x4_vertical_c(uint8_t *src, uint8_t *topright, int stride){
  33     const uint32_t a= ((uint32_t*)(src-stride))[0];
  34     ((uint32_t*)(src+0*stride))[0]= a;
  35     ((uint32_t*)(src+1*stride))[0]= a;
  36     ((uint32_t*)(src+2*stride))[0]= a;
  37     ((uint32_t*)(src+3*stride))[0]= a;
  38 }
  39
  40 static void pred4x4_horizontal_c(uint8_t *src, uint8_t *topright, int stride){
  41     ((uint32_t*)(src+0*stride))[0]= src[-1+0*stride]*0x01010101;
  42     ((uint32_t*)(src+1*stride))[0]= src[-1+1*stride]*0x01010101;
  43     ((uint32_t*)(src+2*stride))[0]= src[-1+2*stride]*0x01010101;
  44     ((uint32_t*)(src+3*stride))[0]= src[-1+3*stride]*0x01010101;
  45 }
  46
  47 static void pred4x4_dc_c(uint8_t *src, uint8_t *topright, int stride){
  48     const int dc= (  src[-stride] + src[1-stride] + src[2-stride] + src[3-stride]
  49                    + src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
  50
  51     ((uint32_t*)(src+0*stride))[0]=
  52     ((uint32_t*)(src+1*stride))[0]=
  53     ((uint32_t*)(src+2*stride))[0]=
  54     ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
  55 }
  56
  57 static void pred4x4_left_dc_c(uint8_t *src, uint8_t *topright, int stride){
  58     const int dc= (  src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 2) >>2;
  59
  60     ((uint32_t*)(src+0*stride))[0]=
  61     ((uint32_t*)(src+1*stride))[0]=
  62     ((uint32_t*)(src+2*stride))[0]=
  63     ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
  64 }
  65
  66 static void pred4x4_top_dc_c(uint8_t *src, uint8_t *topright, int stride){
  67     const int dc= (  src[-stride] + src[1-stride] + src[2-stride] + src[3-stride] + 2) >>2;
  68
  69     ((uint32_t*)(src+0*stride))[0]=
  70     ((uint32_t*)(src+1*stride))[0]=
  71     ((uint32_t*)(src+2*stride))[0]=
  72     ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
  73 }
  74
  75 static void pred4x4_128_dc_c(uint8_t *src, uint8_t *topright, int stride){
  76     ((uint32_t*)(src+0*stride))[0]=
  77     ((uint32_t*)(src+1*stride))[0]=
  78     ((uint32_t*)(src+2*stride))[0]=
  79     ((uint32_t*)(src+3*stride))[0]= 128U*0x01010101U;
  80 }
  81
  82
  83 #define LOAD_TOP_RIGHT_EDGE\
  84     const int av_unused t4= topright[0];\
  85     const int av_unused t5= topright[1];\
  86     const int av_unused t6= topright[2];\
  87     const int av_unused t7= topright[3];\
  88
  89 #define LOAD_DOWN_LEFT_EDGE\
  90     const int av_unused l4= src[-1+4*stride];\
  91     const int av_unused l5= src[-1+5*stride];\
  92     const int av_unused l6= src[-1+6*stride];\
  93     const int av_unused l7= src[-1+7*stride];\
  94
  95 #define LOAD_LEFT_EDGE\
  96     const int av_unused l0= src[-1+0*stride];\
  97     const int av_unused l1= src[-1+1*stride];\
  98     const int av_unused l2= src[-1+2*stride];\
  99     const int av_unused l3= src[-1+3*stride];\
 100
 101 #define LOAD_TOP_EDGE\
 102     const int av_unused t0= src[ 0-1*stride];\
 103     const int av_unused t1= src[ 1-1*stride];\
 104     const int av_unused t2= src[ 2-1*stride];\
 105     const int av_unused t3= src[ 3-1*stride];\
 106
 107 static void pred4x4_down_right_c(uint8_t *src, uint8_t *topright, int stride){
 108     const int lt= src[-1-1*stride];
 109     LOAD_TOP_EDGE
 110     LOAD_LEFT_EDGE
 111
 112     src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2;
 113     src[0+2*stride]=
 114     src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2;
 115     src[0+1*stride]=
 116     src[1+2*stride]=
 117     src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2;
 118     src[0+0*stride]=
 119     src[1+1*stride]=
 120     src[2+2*stride]=
 121     src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
 122     src[1+0*stride]=
 123     src[2+1*stride]=
 124     src[3+2*stride]=(lt + 2*t0 + t1 + 2)>>2;
 125     src[2+0*stride]=
 126     src[3+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
 127     src[3+0*stride]=(t1 + 2*t2 + t3 + 2)>>2;
 128 }
 129
 130 static void pred4x4_down_left_c(uint8_t *src, uint8_t *topright, int stride){
 131     LOAD_TOP_EDGE
 132     LOAD_TOP_RIGHT_EDGE
 133 //    LOAD_LEFT_EDGE
 134
 135     src[0+0*stride]=(t0 + t2 + 2*t1 + 2)>>2;
 136     src[1+0*stride]=
 137     src[0+1*stride]=(t1 + t3 + 2*t2 + 2)>>2;
 138     src[2+0*stride]=
 139     src[1+1*stride]=
 140     src[0+2*stride]=(t2 + t4 + 2*t3 + 2)>>2;
 141     src[3+0*stride]=
 142     src[2+1*stride]=
 143     src[1+2*stride]=
 144     src[0+3*stride]=(t3 + t5 + 2*t4 + 2)>>2;
 145     src[3+1*stride]=
 146     src[2+2*stride]=
 147     src[1+3*stride]=(t4 + t6 + 2*t5 + 2)>>2;
 148     src[3+2*stride]=
 149     src[2+3*stride]=(t5 + t7 + 2*t6 + 2)>>2;
 150     src[3+3*stride]=(t6 + 3*t7 + 2)>>2;
 151 }
 152
 153 static void pred4x4_down_left_svq3_c(uint8_t *src, uint8_t *topright, int stride){
 154     LOAD_TOP_EDGE
 155     LOAD_LEFT_EDGE
 156     const av_unused int unu0= t0;
 157     const av_unused int unu1= l0;
 158
 159     src[0+0*stride]=(l1 + t1)>>1;
 160     src[1+0*stride]=
 161     src[0+1*stride]=(l2 + t2)>>1;
 162     src[2+0*stride]=
 163     src[1+1*stride]=
 164     src[0+2*stride]=
 165     src[3+0*stride]=
 166     src[2+1*stride]=
 167     src[1+2*stride]=
 168     src[0+3*stride]=
 169     src[3+1*stride]=
 170     src[2+2*stride]=
 171     src[1+3*stride]=
 172     src[3+2*stride]=
 173     src[2+3*stride]=
 174     src[3+3*stride]=(l3 + t3)>>1;
 175 }
 176
 177 static void pred4x4_down_left_rv40_c(uint8_t *src, uint8_t *topright, int stride){
 178     LOAD_TOP_EDGE
 179     LOAD_TOP_RIGHT_EDGE
 180     LOAD_LEFT_EDGE
 181     LOAD_DOWN_LEFT_EDGE
 182
 183     src[0+0*stride]=(t0 + t2 + 2*t1 + 2 + l0 + l2 + 2*l1 + 2)>>3;
 184     src[1+0*stride]=
 185     src[0+1*stride]=(t1 + t3 + 2*t2 + 2 + l1 + l3 + 2*l2 + 2)>>3;
 186     src[2+0*stride]=
 187     src[1+1*stride]=
 188     src[0+2*stride]=(t2 + t4 + 2*t3 + 2 + l2 + l4 + 2*l3 + 2)>>3;
 189     src[3+0*stride]=
 190     src[2+1*stride]=
 191     src[1+2*stride]=
 192     src[0+3*stride]=(t3 + t5 + 2*t4 + 2 + l3 + l5 + 2*l4 + 2)>>3;
 193     src[3+1*stride]=
 194     src[2+2*stride]=
 195     src[1+3*stride]=(t4 + t6 + 2*t5 + 2 + l4 + l6 + 2*l5 + 2)>>3;
 196     src[3+2*stride]=
 197     src[2+3*stride]=(t5 + t7 + 2*t6 + 2 + l5 + l7 + 2*l6 + 2)>>3;
 198     src[3+3*stride]=(t6 + t7 + 1 + l6 + l7 + 1)>>2;
 199 }
 200
 201 static void pred4x4_down_left_rv40_notop_c(uint8_t *src, uint8_t *topright, int stride){
 202     LOAD_LEFT_EDGE
 203     LOAD_DOWN_LEFT_EDGE
 204
 205     src[0+0*stride]=(l0 + l2 + 2*l1 + 2)>>2;
 206     src[1+0*stride]=
 207     src[0+1*stride]=(l1 + l3 + 2*l2 + 2)>>2;
 208     src[2+0*stride]=
 209     src[1+1*stride]=
 210     src[0+2*stride]=(l2 + l4 + 2*l3 + 2)>>2;
 211     src[3+0*stride]=
 212     src[2+1*stride]=
 213     src[1+2*stride]=
 214     src[0+3*stride]=(l3 + l5 + 2*l4 + 2)>>2;
 215     src[3+1*stride]=
 216     src[2+2*stride]=
 217     src[1+3*stride]=(l4 + l6 + 2*l5 + 2)>>2;
 218     src[3+2*stride]=
 219     src[2+3*stride]=(l5 + l7 + 2*l6 + 2)>>2;
 220     src[3+3*stride]=(l6 + l7 + 1)>>1;
 221 }
 222
 223 static void pred4x4_down_left_rv40_nodown_c(uint8_t *src, uint8_t *topright, int stride){
 224     LOAD_TOP_EDGE
 225     LOAD_TOP_RIGHT_EDGE
 226     LOAD_LEFT_EDGE
 227
 228     src[0+0*stride]=(t0 + t2 + 2*t1 + 2 + l0 + l2 + 2*l1 + 2)>>3;
 229     src[1+0*stride]=
 230     src[0+1*stride]=(t1 + t3 + 2*t2 + 2 + l1 + l3 + 2*l2 + 2)>>3;
 231     src[2+0*stride]=
 232     src[1+1*stride]=
 233     src[0+2*stride]=(t2 + t4 + 2*t3 + 2 + l2 + 3*l3 + 2)>>3;
 234     src[3+0*stride]=
 235     src[2+1*stride]=
 236     src[1+2*stride]=
 237     src[0+3*stride]=(t3 + t5 + 2*t4 + 2 + l3*4 + 2)>>3;
 238     src[3+1*stride]=
 239     src[2+2*stride]=
 240     src[1+3*stride]=(t4 + t6 + 2*t5 + 2 + l3*4 + 2)>>3;
 241     src[3+2*stride]=
 242     src[2+3*stride]=(t5 + t7 + 2*t6 + 2 + l3*4 + 2)>>3;
 243     src[3+3*stride]=(t6 + t7 + 1 + 2*l3 + 1)>>2;
 244 }
 245
 246 static void pred4x4_vertical_right_c(uint8_t *src, uint8_t *topright, int stride){
 247     const int lt= src[-1-1*stride];
 248     LOAD_TOP_EDGE
 249     LOAD_LEFT_EDGE
 250
 251     src[0+0*stride]=
 252     src[1+2*stride]=(lt + t0 + 1)>>1;
 253     src[1+0*stride]=
 254     src[2+2*stride]=(t0 + t1 + 1)>>1;
 255     src[2+0*stride]=
 256     src[3+2*stride]=(t1 + t2 + 1)>>1;
 257     src[3+0*stride]=(t2 + t3 + 1)>>1;
 258     src[0+1*stride]=
 259     src[1+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
 260     src[1+1*stride]=
 261     src[2+3*stride]=(lt + 2*t0 + t1 + 2)>>2;
 262     src[2+1*stride]=
 263     src[3+3*stride]=(t0 + 2*t1 + t2 + 2)>>2;
 264     src[3+1*stride]=(t1 + 2*t2 + t3 + 2)>>2;
 265     src[0+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
 266     src[0+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
 267 }
 268
 269 static void pred4x4_vertical_left_c(uint8_t *src, uint8_t *topright, int stride){
 270     LOAD_TOP_EDGE
 271     LOAD_TOP_RIGHT_EDGE
 272
 273     src[0+0*stride]=(t0 + t1 + 1)>>1;
 274     src[1+0*stride]=
 275     src[0+2*stride]=(t1 + t2 + 1)>>1;
 276     src[2+0*stride]=
 277     src[1+2*stride]=(t2 + t3 + 1)>>1;
 278     src[3+0*stride]=
 279     src[2+2*stride]=(t3 + t4+ 1)>>1;
 280     src[3+2*stride]=(t4 + t5+ 1)>>1;
 281     src[0+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
 282     src[1+1*stride]=
 283     src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
 284     src[2+1*stride]=
 285     src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
 286     src[3+1*stride]=
 287     src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
 288     src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
 289 }
 290
 291 static void pred4x4_vertical_left_rv40(uint8_t *src, uint8_t *topright, int stride,
 292                                       const int l0, const int l1, const int l2, const int l3, const int l4){
 293     LOAD_TOP_EDGE
 294     LOAD_TOP_RIGHT_EDGE
 295
 296     src[0+0*stride]=(2*t0 + 2*t1 + l1 + 2*l2 + l3 + 4)>>3;
 297     src[1+0*stride]=
 298     src[0+2*stride]=(t1 + t2 + 1)>>1;
 299     src[2+0*stride]=
 300     src[1+2*stride]=(t2 + t3 + 1)>>1;
 301     src[3+0*stride]=
 302     src[2+2*stride]=(t3 + t4+ 1)>>1;
 303     src[3+2*stride]=(t4 + t5+ 1)>>1;
 304     src[0+1*stride]=(t0 + 2*t1 + t2 + l2 + 2*l3 + l4 + 4)>>3;
 305     src[1+1*stride]=
 306     src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
 307     src[2+1*stride]=
 308     src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
 309     src[3+1*stride]=
 310     src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
 311     src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
 312 }
 313
 314 static void pred4x4_vertical_left_rv40_c(uint8_t *src, uint8_t *topright, int stride){
 315     LOAD_LEFT_EDGE
 316     LOAD_DOWN_LEFT_EDGE
 317
 318     pred4x4_vertical_left_rv40(src, topright, stride, l0, l1, l2, l3, l4);
 319 }
 320
 321 static void pred4x4_vertical_left_rv40_nodown_c(uint8_t *src, uint8_t *topright, int stride){
 322     LOAD_LEFT_EDGE
 323
 324     pred4x4_vertical_left_rv40(src, topright, stride, l0, l1, l2, l3, l3);
 325 }
 326
 327 static void pred4x4_horizontal_up_c(uint8_t *src, uint8_t *topright, int stride){
 328     LOAD_LEFT_EDGE
 329
 330     src[0+0*stride]=(l0 + l1 + 1)>>1;
 331     src[1+0*stride]=(l0 + 2*l1 + l2 + 2)>>2;
 332     src[2+0*stride]=
 333     src[0+1*stride]=(l1 + l2 + 1)>>1;
 334     src[3+0*stride]=
 335     src[1+1*stride]=(l1 + 2*l2 + l3 + 2)>>2;
 336     src[2+1*stride]=
 337     src[0+2*stride]=(l2 + l3 + 1)>>1;
 338     src[3+1*stride]=
 339     src[1+2*stride]=(l2 + 2*l3 + l3 + 2)>>2;
 340     src[3+2*stride]=
 341     src[1+3*stride]=
 342     src[0+3*stride]=
 343     src[2+2*stride]=
 344     src[2+3*stride]=
 345     src[3+3*stride]=l3;
 346 }
 347
 348 static void pred4x4_horizontal_up_rv40_c(uint8_t *src, uint8_t *topright, int stride){
 349     LOAD_LEFT_EDGE
 350     LOAD_DOWN_LEFT_EDGE
 351     LOAD_TOP_EDGE
 352     LOAD_TOP_RIGHT_EDGE
 353
 354     src[0+0*stride]=(t1 + 2*t2 + t3 + 2*l0 + 2*l1 + 4)>>3;
 355     src[1+0*stride]=(t2 + 2*t3 + t4 + l0 + 2*l1 + l2 + 4)>>3;
 356     src[2+0*stride]=
 357     src[0+1*stride]=(t3 + 2*t4 + t5 + 2*l1 + 2*l2 + 4)>>3;
 358     src[3+0*stride]=
 359     src[1+1*stride]=(t4 + 2*t5 + t6 + l1 + 2*l2 + l3 + 4)>>3;
 360     src[2+1*stride]=
 361     src[0+2*stride]=(t5 + 2*t6 + t7 + 2*l2 + 2*l3 + 4)>>3;
 362     src[3+1*stride]=
 363     src[1+2*stride]=(t6 + 3*t7 + l2 + 3*l3 + 4)>>3;
 364     src[3+2*stride]=
 365     src[1+3*stride]=(l3 + 2*l4 + l5 + 2)>>2;
 366     src[0+3*stride]=
 367     src[2+2*stride]=(t6 + t7 + l3 + l4 + 2)>>2;
 368     src[2+3*stride]=(l4 + l5 + 1)>>1;
 369     src[3+3*stride]=(l4 + 2*l5 + l6 + 2)>>2;
 370 }
 371
 372 static void pred4x4_horizontal_up_rv40_nodown_c(uint8_t *src, uint8_t *topright, int stride){
 373     LOAD_LEFT_EDGE
 374     LOAD_TOP_EDGE
 375     LOAD_TOP_RIGHT_EDGE
 376
 377     src[0+0*stride]=(t1 + 2*t2 + t3 + 2*l0 + 2*l1 + 4)>>3;
 378     src[1+0*stride]=(t2 + 2*t3 + t4 + l0 + 2*l1 + l2 + 4)>>3;
 379     src[2+0*stride]=
 380     src[0+1*stride]=(t3 + 2*t4 + t5 + 2*l1 + 2*l2 + 4)>>3;
 381     src[3+0*stride]=
 382     src[1+1*stride]=(t4 + 2*t5 + t6 + l1 + 2*l2 + l3 + 4)>>3;
 383     src[2+1*stride]=
 384     src[0+2*stride]=(t5 + 2*t6 + t7 + 2*l2 + 2*l3 + 4)>>3;
 385     src[3+1*stride]=
 386     src[1+2*stride]=(t6 + 3*t7 + l2 + 3*l3 + 4)>>3;
 387     src[3+2*stride]=
 388     src[1+3*stride]=l3;
 389     src[0+3*stride]=
 390     src[2+2*stride]=(t6 + t7 + 2*l3 + 2)>>2;
 391     src[2+3*stride]=
 392     src[3+3*stride]=l3;
 393 }
 394
 395 static void pred4x4_horizontal_down_c(uint8_t *src, uint8_t *topright, int stride){
 396     const int lt= src[-1-1*stride];
 397     LOAD_TOP_EDGE
 398     LOAD_LEFT_EDGE
 399
 400     src[0+0*stride]=
 401     src[2+1*stride]=(lt + l0 + 1)>>1;
 402     src[1+0*stride]=
 403     src[3+1*stride]=(l0 + 2*lt + t0 + 2)>>2;
 404     src[2+0*stride]=(lt + 2*t0 + t1 + 2)>>2;
 405     src[3+0*stride]=(t0 + 2*t1 + t2 + 2)>>2;
 406     src[0+1*stride]=
 407     src[2+2*stride]=(l0 + l1 + 1)>>1;
 408     src[1+1*stride]=
 409     src[3+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
 410     src[0+2*stride]=
 411     src[2+3*stride]=(l1 + l2+ 1)>>1;
 412     src[1+2*stride]=
 413     src[3+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
 414     src[0+3*stride]=(l2 + l3 + 1)>>1;
 415     src[1+3*stride]=(l1 + 2*l2 + l3 + 2)>>2;
 416 }
 417
 418 static void pred16x16_vertical_c(uint8_t *src, int stride){
 419     int i;
 420     const uint32_t a= ((uint32_t*)(src-stride))[0];
 421     const uint32_t b= ((uint32_t*)(src-stride))[1];
 422     const uint32_t c= ((uint32_t*)(src-stride))[2];
 423     const uint32_t d= ((uint32_t*)(src-stride))[3];
 424
 425     for(i=0; i<16; i++){
 426         ((uint32_t*)(src+i*stride))[0]= a;
 427         ((uint32_t*)(src+i*stride))[1]= b;
 428         ((uint32_t*)(src+i*stride))[2]= c;
 429         ((uint32_t*)(src+i*stride))[3]= d;
 430     }
 431 }
 432
 433 static void pred16x16_horizontal_c(uint8_t *src, int stride){
 434     int i;
 435
 436     for(i=0; i<16; i++){
 437         ((uint32_t*)(src+i*stride))[0]=
 438         ((uint32_t*)(src+i*stride))[1]=
 439         ((uint32_t*)(src+i*stride))[2]=
 440         ((uint32_t*)(src+i*stride))[3]= src[-1+i*stride]*0x01010101;
 441     }
 442 }
 443
 444 static void pred16x16_dc_c(uint8_t *src, int stride){
 445     int i, dc=0;
 446
 447     for(i=0;i<16; i++){
 448         dc+= src[-1+i*stride];
 449     }
 450
 451     for(i=0;i<16; i++){
 452         dc+= src[i-stride];
 453     }
 454
 455     dc= 0x01010101*((dc + 16)>>5);
 456
 457     for(i=0; i<16; i++){
 458         ((uint32_t*)(src+i*stride))[0]=
 459         ((uint32_t*)(src+i*stride))[1]=
 460         ((uint32_t*)(src+i*stride))[2]=
 461         ((uint32_t*)(src+i*stride))[3]= dc;
 462     }
 463 }
 464
 465 static void pred16x16_left_dc_c(uint8_t *src, int stride){
 466     int i, dc=0;
 467
 468     for(i=0;i<16; i++){
 469         dc+= src[-1+i*stride];
 470     }
 471
 472     dc= 0x01010101*((dc + 8)>>4);
 473
 474     for(i=0; i<16; i++){
 475         ((uint32_t*)(src+i*stride))[0]=
 476         ((uint32_t*)(src+i*stride))[1]=
 477         ((uint32_t*)(src+i*stride))[2]=
 478         ((uint32_t*)(src+i*stride))[3]= dc;
 479     }
 480 }
 481
 482 static void pred16x16_top_dc_c(uint8_t *src, int stride){
 483     int i, dc=0;
 484
 485     for(i=0;i<16; i++){
 486         dc+= src[i-stride];
 487     }
 488     dc= 0x01010101*((dc + 8)>>4);
 489
 490     for(i=0; i<16; i++){
 491         ((uint32_t*)(src+i*stride))[0]=
 492         ((uint32_t*)(src+i*stride))[1]=
 493         ((uint32_t*)(src+i*stride))[2]=
 494         ((uint32_t*)(src+i*stride))[3]= dc;
 495     }
 496 }
 497
 498 static void pred16x16_128_dc_c(uint8_t *src, int stride){
 499     int i;
 500
 501     for(i=0; i<16; i++){
 502         ((uint32_t*)(src+i*stride))[0]=
 503         ((uint32_t*)(src+i*stride))[1]=
 504         ((uint32_t*)(src+i*stride))[2]=
 505         ((uint32_t*)(src+i*stride))[3]= 0x01010101U*128U;
 506     }
 507 }
 508
 509 static inline void pred16x16_plane_compat_c(uint8_t *src, int stride, const int svq3, const int rv40){
 510   int i, j, k;
 511   int a;
 512   uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 513   const uint8_t * const src0 = src+7-stride;
 514   const uint8_t *src1 = src+8*stride-1;
 515   const uint8_t *src2 = src1-2*stride;      // == src+6*stride-1;
 516   int H = src0[1] - src0[-1];
 517   int V = src1[0] - src2[ 0];
 518   for(k=2; k<=8; ++k) {
 519     src1 += stride; src2 -= stride;
 520     H += k*(src0[k] - src0[-k]);
 521     V += k*(src1[0] - src2[ 0]);
 522   }
 523   if(svq3){
 524     H = ( 5*(H/4) ) / 16;
 525     V = ( 5*(V/4) ) / 16;
 526
 527     /* required for 100% accuracy */
 528     i = H; H = V; V = i;
 529   }else if(rv40){
 530     H = ( H + (H>>2) ) >> 4;
 531     V = ( V + (V>>2) ) >> 4;
 532   }else{
 533     H = ( 5*H+32 ) >> 6;
 534     V = ( 5*V+32 ) >> 6;
 535   }
 536
 537   a = 16*(src1[0] + src2[16] + 1) - 7*(V+H);
 538   for(j=16; j>0; --j) {
 539     int b = a;
 540     a += V;
 541     for(i=-16; i<0; i+=4) {
 542       src[16+i] = cm[ (b    ) >> 5 ];
 543       src[17+i] = cm[ (b+  H) >> 5 ];
 544       src[18+i] = cm[ (b+2*H) >> 5 ];
 545       src[19+i] = cm[ (b+3*H) >> 5 ];
 546       b += 4*H;
 547     }
 548     src += stride;
 549   }
 550 }
 551
 552 static void pred16x16_plane_c(uint8_t *src, int stride){
 553     pred16x16_plane_compat_c(src, stride, 0, 0);
 554 }
 555
 556 static void pred16x16_plane_svq3_c(uint8_t *src, int stride){
 557     pred16x16_plane_compat_c(src, stride, 1, 0);
 558 }
 559
 560 static void pred16x16_plane_rv40_c(uint8_t *src, int stride){
 561     pred16x16_plane_compat_c(src, stride, 0, 1);
 562 }
 563
 564 static void pred8x8_vertical_c(uint8_t *src, int stride){
 565     int i;
 566     const uint32_t a= ((uint32_t*)(src-stride))[0];
 567     const uint32_t b= ((uint32_t*)(src-stride))[1];
 568
 569     for(i=0; i<8; i++){
 570         ((uint32_t*)(src+i*stride))[0]= a;
 571         ((uint32_t*)(src+i*stride))[1]= b;
 572     }
 573 }
 574
 575 static void pred8x8_horizontal_c(uint8_t *src, int stride){
 576     int i;
 577
 578     for(i=0; i<8; i++){
 579         ((uint32_t*)(src+i*stride))[0]=
 580         ((uint32_t*)(src+i*stride))[1]= src[-1+i*stride]*0x01010101;
 581     }
 582 }
 583
 584 static void pred8x8_128_dc_c(uint8_t *src, int stride){
 585     int i;
 586
 587     for(i=0; i<8; i++){
 588         ((uint32_t*)(src+i*stride))[0]=
 589         ((uint32_t*)(src+i*stride))[1]= 0x01010101U*128U;
 590     }
 591 }
 592
 593 static void pred8x8_left_dc_c(uint8_t *src, int stride){
 594     int i;
 595     int dc0, dc2;
 596
 597     dc0=dc2=0;
 598     for(i=0;i<4; i++){
 599         dc0+= src[-1+i*stride];
 600         dc2+= src[-1+(i+4)*stride];
 601     }
 602     dc0= 0x01010101*((dc0 + 2)>>2);
 603     dc2= 0x01010101*((dc2 + 2)>>2);
 604
 605     for(i=0; i<4; i++){
 606         ((uint32_t*)(src+i*stride))[0]=
 607         ((uint32_t*)(src+i*stride))[1]= dc0;
 608     }
 609     for(i=4; i<8; i++){
 610         ((uint32_t*)(src+i*stride))[0]=
 611         ((uint32_t*)(src+i*stride))[1]= dc2;
 612     }
 613 }
 614
 615 static void pred8x8_left_dc_rv40_c(uint8_t *src, int stride){
 616     int i;
 617     int dc0;
 618
 619     dc0=0;
 620     for(i=0;i<8; i++)
 621         dc0+= src[-1+i*stride];
 622     dc0= 0x01010101*((dc0 + 4)>>3);
 623
 624     for(i=0; i<8; i++){
 625         ((uint32_t*)(src+i*stride))[0]=
 626         ((uint32_t*)(src+i*stride))[1]= dc0;
 627     }
 628 }
 629
 630 static void pred8x8_top_dc_c(uint8_t *src, int stride){
 631     int i;
 632     int dc0, dc1;
 633
 634     dc0=dc1=0;
 635     for(i=0;i<4; i++){
 636         dc0+= src[i-stride];
 637         dc1+= src[4+i-stride];
 638     }
 639     dc0= 0x01010101*((dc0 + 2)>>2);
 640     dc1= 0x01010101*((dc1 + 2)>>2);
 641
 642     for(i=0; i<4; i++){
 643         ((uint32_t*)(src+i*stride))[0]= dc0;
 644         ((uint32_t*)(src+i*stride))[1]= dc1;
 645     }
 646     for(i=4; i<8; i++){
 647         ((uint32_t*)(src+i*stride))[0]= dc0;
 648         ((uint32_t*)(src+i*stride))[1]= dc1;
 649     }
 650 }
 651
 652 static void pred8x8_top_dc_rv40_c(uint8_t *src, int stride){
 653     int i;
 654     int dc0;
 655
 656     dc0=0;
 657     for(i=0;i<8; i++)
 658         dc0+= src[i-stride];
 659     dc0= 0x01010101*((dc0 + 4)>>3);
 660
 661     for(i=0; i<8; i++){
 662         ((uint32_t*)(src+i*stride))[0]=
 663         ((uint32_t*)(src+i*stride))[1]= dc0;
 664     }
 665 }
 666
 667
 668 static void pred8x8_dc_c(uint8_t *src, int stride){
 669     int i;
 670     int dc0, dc1, dc2, dc3;
 671
 672     dc0=dc1=dc2=0;
 673     for(i=0;i<4; i++){
 674         dc0+= src[-1+i*stride] + src[i-stride];
 675         dc1+= src[4+i-stride];
 676         dc2+= src[-1+(i+4)*stride];
 677     }
 678     dc3= 0x01010101*((dc1 + dc2 + 4)>>3);
 679     dc0= 0x01010101*((dc0 + 4)>>3);
 680     dc1= 0x01010101*((dc1 + 2)>>2);
 681     dc2= 0x01010101*((dc2 + 2)>>2);
 682
 683     for(i=0; i<4; i++){
 684         ((uint32_t*)(src+i*stride))[0]= dc0;
 685         ((uint32_t*)(src+i*stride))[1]= dc1;
 686     }
 687     for(i=4; i<8; i++){
 688         ((uint32_t*)(src+i*stride))[0]= dc2;
 689         ((uint32_t*)(src+i*stride))[1]= dc3;
 690     }
 691 }
 692
 693 //the following 4 function should not be optimized!
 694 static void pred8x8_mad_cow_dc_l0t(uint8_t *src, int stride){
 695     pred8x8_top_dc_c(src, stride);
 696     pred4x4_dc_c(src, NULL, stride);
 697 }
 698
 699 static void pred8x8_mad_cow_dc_0lt(uint8_t *src, int stride){
 700     pred8x8_dc_c(src, stride);
 701     pred4x4_top_dc_c(src, NULL, stride);
 702 }
 703
 704 static void pred8x8_mad_cow_dc_l00(uint8_t *src, int stride){
 705     pred8x8_left_dc_c(src, stride);
 706     pred4x4_128_dc_c(src + 4*stride    , NULL, stride);
 707     pred4x4_128_dc_c(src + 4*stride + 4, NULL, stride);
 708 }
 709
 710 static void pred8x8_mad_cow_dc_0l0(uint8_t *src, int stride){
 711     pred8x8_left_dc_c(src, stride);
 712     pred4x4_128_dc_c(src    , NULL, stride);
 713     pred4x4_128_dc_c(src + 4, NULL, stride);
 714 }
 715
 716 static void pred8x8_dc_rv40_c(uint8_t *src, int stride){
 717     int i;
 718     int dc0=0;
 719
 720     for(i=0;i<4; i++){
 721         dc0+= src[-1+i*stride] + src[i-stride];
 722         dc0+= src[4+i-stride];
 723         dc0+= src[-1+(i+4)*stride];
 724     }
 725     dc0= 0x01010101*((dc0 + 8)>>4);
 726
 727     for(i=0; i<4; i++){
 728         ((uint32_t*)(src+i*stride))[0]= dc0;
 729         ((uint32_t*)(src+i*stride))[1]= dc0;
 730     }
 731     for(i=4; i<8; i++){
 732         ((uint32_t*)(src+i*stride))[0]= dc0;
 733         ((uint32_t*)(src+i*stride))[1]= dc0;
 734     }
 735 }
 736
 737 static void pred8x8_plane_c(uint8_t *src, int stride){
 738   int j, k;
 739   int a;
 740   uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 741   const uint8_t * const src0 = src+3-stride;
 742   const uint8_t *src1 = src+4*stride-1;
 743   const uint8_t *src2 = src1-2*stride;      // == src+2*stride-1;
 744   int H = src0[1] - src0[-1];
 745   int V = src1[0] - src2[ 0];
 746   for(k=2; k<=4; ++k) {
 747     src1 += stride; src2 -= stride;
 748     H += k*(src0[k] - src0[-k]);
 749     V += k*(src1[0] - src2[ 0]);
 750   }
 751   H = ( 17*H+16 ) >> 5;
 752   V = ( 17*V+16 ) >> 5;
 753
 754   a = 16*(src1[0] + src2[8]+1) - 3*(V+H);
 755   for(j=8; j>0; --j) {
 756     int b = a;
 757     a += V;
 758     src[0] = cm[ (b    ) >> 5 ];
 759     src[1] = cm[ (b+  H) >> 5 ];
 760     src[2] = cm[ (b+2*H) >> 5 ];
 761     src[3] = cm[ (b+3*H) >> 5 ];
 762     src[4] = cm[ (b+4*H) >> 5 ];
 763     src[5] = cm[ (b+5*H) >> 5 ];
 764     src[6] = cm[ (b+6*H) >> 5 ];
 765     src[7] = cm[ (b+7*H) >> 5 ];
 766     src += stride;
 767   }
 768 }
 769
 770 #define SRC(x,y) src[(x)+(y)*stride]
 771 #define PL(y) \
 772     const int l##y = (SRC(-1,y-1) + 2*SRC(-1,y) + SRC(-1,y+1) + 2) >> 2;
 773 #define PREDICT_8x8_LOAD_LEFT \
 774     const int l0 = ((has_topleft ? SRC(-1,-1) : SRC(-1,0)) \
 775                      + 2*SRC(-1,0) + SRC(-1,1) + 2) >> 2; \
 776     PL(1) PL(2) PL(3) PL(4) PL(5) PL(6) \
 777     const int l7 av_unused = (SRC(-1,6) + 3*SRC(-1,7) + 2) >> 2
 778
 779 #define PT(x) \
 780     const int t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
 781 #define PREDICT_8x8_LOAD_TOP \
 782     const int t0 = ((has_topleft ? SRC(-1,-1) : SRC(0,-1)) \
 783                      + 2*SRC(0,-1) + SRC(1,-1) + 2) >> 2; \
 784     PT(1) PT(2) PT(3) PT(4) PT(5) PT(6) \
 785     const int t7 av_unused = ((has_topright ? SRC(8,-1) : SRC(7,-1)) \
 786                      + 2*SRC(7,-1) + SRC(6,-1) + 2) >> 2
 787
 788 #define PTR(x) \
 789     t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
 790 #define PREDICT_8x8_LOAD_TOPRIGHT \
 791     int t8, t9, t10, t11, t12, t13, t14, t15; \
 792     if(has_topright) { \
 793         PTR(8) PTR(9) PTR(10) PTR(11) PTR(12) PTR(13) PTR(14) \
 794         t15 = (SRC(14,-1) + 3*SRC(15,-1) + 2) >> 2; \
 795     } else t8=t9=t10=t11=t12=t13=t14=t15= SRC(7,-1);
 796
 797 #define PREDICT_8x8_LOAD_TOPLEFT \
 798     const int lt = (SRC(-1,0) + 2*SRC(-1,-1) + SRC(0,-1) + 2) >> 2
 799
 800 #define PREDICT_8x8_DC(v) \
 801     int y; \
 802     for( y = 0; y < 8; y++ ) { \
 803         ((uint32_t*)src)[0] = \
 804         ((uint32_t*)src)[1] = v; \
 805         src += stride; \
 806     }
 807
 808 static void pred8x8l_128_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 809 {
 810     PREDICT_8x8_DC(0x80808080);
 811 }
 812 static void pred8x8l_left_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 813 {
 814     PREDICT_8x8_LOAD_LEFT;
 815     const uint32_t dc = ((l0+l1+l2+l3+l4+l5+l6+l7+4) >> 3) * 0x01010101;
 816     PREDICT_8x8_DC(dc);
 817 }
 818 static void pred8x8l_top_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 819 {
 820     PREDICT_8x8_LOAD_TOP;
 821     const uint32_t dc = ((t0+t1+t2+t3+t4+t5+t6+t7+4) >> 3) * 0x01010101;
 822     PREDICT_8x8_DC(dc);
 823 }
 824 static void pred8x8l_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 825 {
 826     PREDICT_8x8_LOAD_LEFT;
 827     PREDICT_8x8_LOAD_TOP;
 828     const uint32_t dc = ((l0+l1+l2+l3+l4+l5+l6+l7
 829                          +t0+t1+t2+t3+t4+t5+t6+t7+8) >> 4) * 0x01010101;
 830     PREDICT_8x8_DC(dc);
 831 }
 832 static void pred8x8l_horizontal_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 833 {
 834     PREDICT_8x8_LOAD_LEFT;
 835 #define ROW(y) ((uint32_t*)(src+y*stride))[0] =\
 836                ((uint32_t*)(src+y*stride))[1] = 0x01010101 * l##y
 837     ROW(0); ROW(1); ROW(2); ROW(3); ROW(4); ROW(5); ROW(6); ROW(7);
 838 #undef ROW
 839 }
 840 static void pred8x8l_vertical_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 841 {
 842     int y;
 843     PREDICT_8x8_LOAD_TOP;
 844     src[0] = t0;
 845     src[1] = t1;
 846     src[2] = t2;
 847     src[3] = t3;
 848     src[4] = t4;
 849     src[5] = t5;
 850     src[6] = t6;
 851     src[7] = t7;
 852     for( y = 1; y < 8; y++ )
 853         *(uint64_t*)(src+y*stride) = *(uint64_t*)src;
 854 }
 855 static void pred8x8l_down_left_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 856 {
 857     PREDICT_8x8_LOAD_TOP;
 858     PREDICT_8x8_LOAD_TOPRIGHT;
 859     SRC(0,0)= (t0 + 2*t1 + t2 + 2) >> 2;
 860     SRC(0,1)=SRC(1,0)= (t1 + 2*t2 + t3 + 2) >> 2;
 861     SRC(0,2)=SRC(1,1)=SRC(2,0)= (t2 + 2*t3 + t4 + 2) >> 2;
 862     SRC(0,3)=SRC(1,2)=SRC(2,1)=SRC(3,0)= (t3 + 2*t4 + t5 + 2) >> 2;
 863     SRC(0,4)=SRC(1,3)=SRC(2,2)=SRC(3,1)=SRC(4,0)= (t4 + 2*t5 + t6 + 2) >> 2;
 864     SRC(0,5)=SRC(1,4)=SRC(2,3)=SRC(3,2)=SRC(4,1)=SRC(5,0)= (t5 + 2*t6 + t7 + 2) >> 2;
 865     SRC(0,6)=SRC(1,5)=SRC(2,4)=SRC(3,3)=SRC(4,2)=SRC(5,1)=SRC(6,0)= (t6 + 2*t7 + t8 + 2) >> 2;
 866     SRC(0,7)=SRC(1,6)=SRC(2,5)=SRC(3,4)=SRC(4,3)=SRC(5,2)=SRC(6,1)=SRC(7,0)= (t7 + 2*t8 + t9 + 2) >> 2;
 867     SRC(1,7)=SRC(2,6)=SRC(3,5)=SRC(4,4)=SRC(5,3)=SRC(6,2)=SRC(7,1)= (t8 + 2*t9 + t10 + 2) >> 2;
 868     SRC(2,7)=SRC(3,6)=SRC(4,5)=SRC(5,4)=SRC(6,3)=SRC(7,2)= (t9 + 2*t10 + t11 + 2) >> 2;
 869     SRC(3,7)=SRC(4,6)=SRC(5,5)=SRC(6,4)=SRC(7,3)= (t10 + 2*t11 + t12 + 2) >> 2;
 870     SRC(4,7)=SRC(5,6)=SRC(6,5)=SRC(7,4)= (t11 + 2*t12 + t13 + 2) >> 2;
 871     SRC(5,7)=SRC(6,6)=SRC(7,5)= (t12 + 2*t13 + t14 + 2) >> 2;
 872     SRC(6,7)=SRC(7,6)= (t13 + 2*t14 + t15 + 2) >> 2;
 873     SRC(7,7)= (t14 + 3*t15 + 2) >> 2;
 874 }
 875 static void pred8x8l_down_right_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 876 {
 877     PREDICT_8x8_LOAD_TOP;
 878     PREDICT_8x8_LOAD_LEFT;
 879     PREDICT_8x8_LOAD_TOPLEFT;
 880     SRC(0,7)= (l7 + 2*l6 + l5 + 2) >> 2;
 881     SRC(0,6)=SRC(1,7)= (l6 + 2*l5 + l4 + 2) >> 2;
 882     SRC(0,5)=SRC(1,6)=SRC(2,7)= (l5 + 2*l4 + l3 + 2) >> 2;
 883     SRC(0,4)=SRC(1,5)=SRC(2,6)=SRC(3,7)= (l4 + 2*l3 + l2 + 2) >> 2;
 884     SRC(0,3)=SRC(1,4)=SRC(2,5)=SRC(3,6)=SRC(4,7)= (l3 + 2*l2 + l1 + 2) >> 2;
 885     SRC(0,2)=SRC(1,3)=SRC(2,4)=SRC(3,5)=SRC(4,6)=SRC(5,7)= (l2 + 2*l1 + l0 + 2) >> 2;
 886     SRC(0,1)=SRC(1,2)=SRC(2,3)=SRC(3,4)=SRC(4,5)=SRC(5,6)=SRC(6,7)= (l1 + 2*l0 + lt + 2) >> 2;
 887     SRC(0,0)=SRC(1,1)=SRC(2,2)=SRC(3,3)=SRC(4,4)=SRC(5,5)=SRC(6,6)=SRC(7,7)= (l0 + 2*lt + t0 + 2) >> 2;
 888     SRC(1,0)=SRC(2,1)=SRC(3,2)=SRC(4,3)=SRC(5,4)=SRC(6,5)=SRC(7,6)= (lt + 2*t0 + t1 + 2) >> 2;
 889     SRC(2,0)=SRC(3,1)=SRC(4,2)=SRC(5,3)=SRC(6,4)=SRC(7,5)= (t0 + 2*t1 + t2 + 2) >> 2;
 890     SRC(3,0)=SRC(4,1)=SRC(5,2)=SRC(6,3)=SRC(7,4)= (t1 + 2*t2 + t3 + 2) >> 2;
 891     SRC(4,0)=SRC(5,1)=SRC(6,2)=SRC(7,3)= (t2 + 2*t3 + t4 + 2) >> 2;
 892     SRC(5,0)=SRC(6,1)=SRC(7,2)= (t3 + 2*t4 + t5 + 2) >> 2;
 893     SRC(6,0)=SRC(7,1)= (t4 + 2*t5 + t6 + 2) >> 2;
 894     SRC(7,0)= (t5 + 2*t6 + t7 + 2) >> 2;
 895
 896 }
 897 static void pred8x8l_vertical_right_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 898 {
 899     PREDICT_8x8_LOAD_TOP;
 900     PREDICT_8x8_LOAD_LEFT;
 901     PREDICT_8x8_LOAD_TOPLEFT;
 902     SRC(0,6)= (l5 + 2*l4 + l3 + 2) >> 2;
 903     SRC(0,7)= (l6 + 2*l5 + l4 + 2) >> 2;
 904     SRC(0,4)=SRC(1,6)= (l3 + 2*l2 + l1 + 2) >> 2;
 905     SRC(0,5)=SRC(1,7)= (l4 + 2*l3 + l2 + 2) >> 2;
 906     SRC(0,2)=SRC(1,4)=SRC(2,6)= (l1 + 2*l0 + lt + 2) >> 2;
 907     SRC(0,3)=SRC(1,5)=SRC(2,7)= (l2 + 2*l1 + l0 + 2) >> 2;
 908     SRC(0,1)=SRC(1,3)=SRC(2,5)=SRC(3,7)= (l0 + 2*lt + t0 + 2) >> 2;
 909     SRC(0,0)=SRC(1,2)=SRC(2,4)=SRC(3,6)= (lt + t0 + 1) >> 1;
 910     SRC(1,1)=SRC(2,3)=SRC(3,5)=SRC(4,7)= (lt + 2*t0 + t1 + 2) >> 2;
 911     SRC(1,0)=SRC(2,2)=SRC(3,4)=SRC(4,6)= (t0 + t1 + 1) >> 1;
 912     SRC(2,1)=SRC(3,3)=SRC(4,5)=SRC(5,7)= (t0 + 2*t1 + t2 + 2) >> 2;
 913     SRC(2,0)=SRC(3,2)=SRC(4,4)=SRC(5,6)= (t1 + t2 + 1) >> 1;
 914     SRC(3,1)=SRC(4,3)=SRC(5,5)=SRC(6,7)= (t1 + 2*t2 + t3 + 2) >> 2;
 915     SRC(3,0)=SRC(4,2)=SRC(5,4)=SRC(6,6)= (t2 + t3 + 1) >> 1;
 916     SRC(4,1)=SRC(5,3)=SRC(6,5)=SRC(7,7)= (t2 + 2*t3 + t4 + 2) >> 2;
 917     SRC(4,0)=SRC(5,2)=SRC(6,4)=SRC(7,6)= (t3 + t4 + 1) >> 1;
 918     SRC(5,1)=SRC(6,3)=SRC(7,5)= (t3 + 2*t4 + t5 + 2) >> 2;
 919     SRC(5,0)=SRC(6,2)=SRC(7,4)= (t4 + t5 + 1) >> 1;
 920     SRC(6,1)=SRC(7,3)= (t4 + 2*t5 + t6 + 2) >> 2;
 921     SRC(6,0)=SRC(7,2)= (t5 + t6 + 1) >> 1;
 922     SRC(7,1)= (t5 + 2*t6 + t7 + 2) >> 2;
 923     SRC(7,0)= (t6 + t7 + 1) >> 1;
 924 }
 925 static void pred8x8l_horizontal_down_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 926 {
 927     PREDICT_8x8_LOAD_TOP;
 928     PREDICT_8x8_LOAD_LEFT;
 929     PREDICT_8x8_LOAD_TOPLEFT;
 930     SRC(0,7)= (l6 + l7 + 1) >> 1;
 931     SRC(1,7)= (l5 + 2*l6 + l7 + 2) >> 2;
 932     SRC(0,6)=SRC(2,7)= (l5 + l6 + 1) >> 1;
 933     SRC(1,6)=SRC(3,7)= (l4 + 2*l5 + l6 + 2) >> 2;
 934     SRC(0,5)=SRC(2,6)=SRC(4,7)= (l4 + l5 + 1) >> 1;
 935     SRC(1,5)=SRC(3,6)=SRC(5,7)= (l3 + 2*l4 + l5 + 2) >> 2;
 936     SRC(0,4)=SRC(2,5)=SRC(4,6)=SRC(6,7)= (l3 + l4 + 1) >> 1;
 937     SRC(1,4)=SRC(3,5)=SRC(5,6)=SRC(7,7)= (l2 + 2*l3 + l4 + 2) >> 2;
 938     SRC(0,3)=SRC(2,4)=SRC(4,5)=SRC(6,6)= (l2 + l3 + 1) >> 1;
 939     SRC(1,3)=SRC(3,4)=SRC(5,5)=SRC(7,6)= (l1 + 2*l2 + l3 + 2) >> 2;
 940     SRC(0,2)=SRC(2,3)=SRC(4,4)=SRC(6,5)= (l1 + l2 + 1) >> 1;
 941     SRC(1,2)=SRC(3,3)=SRC(5,4)=SRC(7,5)= (l0 + 2*l1 + l2 + 2) >> 2;
 942     SRC(0,1)=SRC(2,2)=SRC(4,3)=SRC(6,4)= (l0 + l1 + 1) >> 1;
 943     SRC(1,1)=SRC(3,2)=SRC(5,3)=SRC(7,4)= (lt + 2*l0 + l1 + 2) >> 2;
 944     SRC(0,0)=SRC(2,1)=SRC(4,2)=SRC(6,3)= (lt + l0 + 1) >> 1;
 945     SRC(1,0)=SRC(3,1)=SRC(5,2)=SRC(7,3)= (l0 + 2*lt + t0 + 2) >> 2;
 946     SRC(2,0)=SRC(4,1)=SRC(6,2)= (t1 + 2*t0 + lt + 2) >> 2;
 947     SRC(3,0)=SRC(5,1)=SRC(7,2)= (t2 + 2*t1 + t0 + 2) >> 2;
 948     SRC(4,0)=SRC(6,1)= (t3 + 2*t2 + t1 + 2) >> 2;
 949     SRC(5,0)=SRC(7,1)= (t4 + 2*t3 + t2 + 2) >> 2;
 950     SRC(6,0)= (t5 + 2*t4 + t3 + 2) >> 2;
 951     SRC(7,0)= (t6 + 2*t5 + t4 + 2) >> 2;
 952 }
 953 static void pred8x8l_vertical_left_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 954 {
 955     PREDICT_8x8_LOAD_TOP;
 956     PREDICT_8x8_LOAD_TOPRIGHT;
 957     SRC(0,0)= (t0 + t1 + 1) >> 1;
 958     SRC(0,1)= (t0 + 2*t1 + t2 + 2) >> 2;
 959     SRC(0,2)=SRC(1,0)= (t1 + t2 + 1) >> 1;
 960     SRC(0,3)=SRC(1,1)= (t1 + 2*t2 + t3 + 2) >> 2;
 961     SRC(0,4)=SRC(1,2)=SRC(2,0)= (t2 + t3 + 1) >> 1;
 962     SRC(0,5)=SRC(1,3)=SRC(2,1)= (t2 + 2*t3 + t4 + 2) >> 2;
 963     SRC(0,6)=SRC(1,4)=SRC(2,2)=SRC(3,0)= (t3 + t4 + 1) >> 1;
 964     SRC(0,7)=SRC(1,5)=SRC(2,3)=SRC(3,1)= (t3 + 2*t4 + t5 + 2) >> 2;
 965     SRC(1,6)=SRC(2,4)=SRC(3,2)=SRC(4,0)= (t4 + t5 + 1) >> 1;
 966     SRC(1,7)=SRC(2,5)=SRC(3,3)=SRC(4,1)= (t4 + 2*t5 + t6 + 2) >> 2;
 967     SRC(2,6)=SRC(3,4)=SRC(4,2)=SRC(5,0)= (t5 + t6 + 1) >> 1;
 968     SRC(2,7)=SRC(3,5)=SRC(4,3)=SRC(5,1)= (t5 + 2*t6 + t7 + 2) >> 2;
 969     SRC(3,6)=SRC(4,4)=SRC(5,2)=SRC(6,0)= (t6 + t7 + 1) >> 1;
 970     SRC(3,7)=SRC(4,5)=SRC(5,3)=SRC(6,1)= (t6 + 2*t7 + t8 + 2) >> 2;
 971     SRC(4,6)=SRC(5,4)=SRC(6,2)=SRC(7,0)= (t7 + t8 + 1) >> 1;
 972     SRC(4,7)=SRC(5,5)=SRC(6,3)=SRC(7,1)= (t7 + 2*t8 + t9 + 2) >> 2;
 973     SRC(5,6)=SRC(6,4)=SRC(7,2)= (t8 + t9 + 1) >> 1;
 974     SRC(5,7)=SRC(6,5)=SRC(7,3)= (t8 + 2*t9 + t10 + 2) >> 2;
 975     SRC(6,6)=SRC(7,4)= (t9 + t10 + 1) >> 1;
 976     SRC(6,7)=SRC(7,5)= (t9 + 2*t10 + t11 + 2) >> 2;
 977     SRC(7,6)= (t10 + t11 + 1) >> 1;
 978     SRC(7,7)= (t10 + 2*t11 + t12 + 2) >> 2;
 979 }
 980 static void pred8x8l_horizontal_up_c(uint8_t *src, int has_topleft, int has_topright, int stride)
 981 {
 982     PREDICT_8x8_LOAD_LEFT;
 983     SRC(0,0)= (l0 + l1 + 1) >> 1;
 984     SRC(1,0)= (l0 + 2*l1 + l2 + 2) >> 2;
 985     SRC(0,1)=SRC(2,0)= (l1 + l2 + 1) >> 1;
 986     SRC(1,1)=SRC(3,0)= (l1 + 2*l2 + l3 + 2) >> 2;
 987     SRC(0,2)=SRC(2,1)=SRC(4,0)= (l2 + l3 + 1) >> 1;
 988     SRC(1,2)=SRC(3,1)=SRC(5,0)= (l2 + 2*l3 + l4 + 2) >> 2;
 989     SRC(0,3)=SRC(2,2)=SRC(4,1)=SRC(6,0)= (l3 + l4 + 1) >> 1;
 990     SRC(1,3)=SRC(3,2)=SRC(5,1)=SRC(7,0)= (l3 + 2*l4 + l5 + 2) >> 2;
 991     SRC(0,4)=SRC(2,3)=SRC(4,2)=SRC(6,1)= (l4 + l5 + 1) >> 1;
 992     SRC(1,4)=SRC(3,3)=SRC(5,2)=SRC(7,1)= (l4 + 2*l5 + l6 + 2) >> 2;
 993     SRC(0,5)=SRC(2,4)=SRC(4,3)=SRC(6,2)= (l5 + l6 + 1) >> 1;
 994     SRC(1,5)=SRC(3,4)=SRC(5,3)=SRC(7,2)= (l5 + 2*l6 + l7 + 2) >> 2;
 995     SRC(0,6)=SRC(2,5)=SRC(4,4)=SRC(6,3)= (l6 + l7 + 1) >> 1;
 996     SRC(1,6)=SRC(3,5)=SRC(5,4)=SRC(7,3)= (l6 + 3*l7 + 2) >> 2;
 997     SRC(0,7)=SRC(1,7)=SRC(2,6)=SRC(2,7)=SRC(3,6)=
 998     SRC(3,7)=SRC(4,5)=SRC(4,6)=SRC(4,7)=SRC(5,5)=
 999     SRC(5,6)=SRC(5,7)=SRC(6,4)=SRC(6,5)=SRC(6,6)=
1000     SRC(6,7)=SRC(7,4)=SRC(7,5)=SRC(7,6)=SRC(7,7)= l7;
1001 }
1002 #undef PREDICT_8x8_LOAD_LEFT
1003 #undef PREDICT_8x8_LOAD_TOP
1004 #undef PREDICT_8x8_LOAD_TOPLEFT
1005 #undef PREDICT_8x8_LOAD_TOPRIGHT
1006 #undef PREDICT_8x8_DC
1007 #undef PTR
1008 #undef PT
1009 #undef PL
1010 #undef SRC
1011
1012 /**
1013  * Sets the intra prediction function pointers.
1014  */
1015 void ff_h264_pred_init(H264PredContext *h, int codec_id){
1016 //    MpegEncContext * const s = &h->s;
1017
1018     if(codec_id != CODEC_ID_RV40){
1019         h->pred4x4[VERT_PRED           ]= pred4x4_vertical_c;
1020         h->pred4x4[HOR_PRED            ]= pred4x4_horizontal_c;
1021         h->pred4x4[DC_PRED             ]= pred4x4_dc_c;
1022         if(codec_id == CODEC_ID_SVQ3)
1023             h->pred4x4[DIAG_DOWN_LEFT_PRED ]= pred4x4_down_left_svq3_c;
1024         else
1025             h->pred4x4[DIAG_DOWN_LEFT_PRED ]= pred4x4_down_left_c;
1026         h->pred4x4[DIAG_DOWN_RIGHT_PRED]= pred4x4_down_right_c;
1027         h->pred4x4[VERT_RIGHT_PRED     ]= pred4x4_vertical_right_c;
1028         h->pred4x4[HOR_DOWN_PRED       ]= pred4x4_horizontal_down_c;
1029         h->pred4x4[VERT_LEFT_PRED      ]= pred4x4_vertical_left_c;
1030         h->pred4x4[HOR_UP_PRED         ]= pred4x4_horizontal_up_c;
1031         h->pred4x4[LEFT_DC_PRED        ]= pred4x4_left_dc_c;
1032         h->pred4x4[TOP_DC_PRED         ]= pred4x4_top_dc_c;
1033         h->pred4x4[DC_128_PRED         ]= pred4x4_128_dc_c;
1034     }else{
1035         h->pred4x4[VERT_PRED           ]= pred4x4_vertical_c;
1036         h->pred4x4[HOR_PRED            ]= pred4x4_horizontal_c;
1037         h->pred4x4[DC_PRED             ]= pred4x4_dc_c;
1038         h->pred4x4[DIAG_DOWN_LEFT_PRED ]= pred4x4_down_left_rv40_c;
1039         h->pred4x4[DIAG_DOWN_RIGHT_PRED]= pred4x4_down_right_c;
1040         h->pred4x4[VERT_RIGHT_PRED     ]= pred4x4_vertical_right_c;
1041         h->pred4x4[HOR_DOWN_PRED       ]= pred4x4_horizontal_down_c;
1042         h->pred4x4[VERT_LEFT_PRED      ]= pred4x4_vertical_left_rv40_c;
1043         h->pred4x4[HOR_UP_PRED         ]= pred4x4_horizontal_up_rv40_c;
1044         h->pred4x4[LEFT_DC_PRED        ]= pred4x4_left_dc_c;
1045         h->pred4x4[TOP_DC_PRED         ]= pred4x4_top_dc_c;
1046         h->pred4x4[DC_128_PRED         ]= pred4x4_128_dc_c;
1047         h->pred4x4[DIAG_DOWN_LEFT_PRED_RV40_NODOWN]= pred4x4_down_left_rv40_nodown_c;
1048         h->pred4x4[HOR_UP_PRED_RV40_NODOWN]= pred4x4_horizontal_up_rv40_nodown_c;
1049         h->pred4x4[VERT_LEFT_PRED_RV40_NODOWN]= pred4x4_vertical_left_rv40_nodown_c;
1050     }
1051
1052     h->pred8x8l[VERT_PRED           ]= pred8x8l_vertical_c;
1053     h->pred8x8l[HOR_PRED            ]= pred8x8l_horizontal_c;
1054     h->pred8x8l[DC_PRED             ]= pred8x8l_dc_c;
1055     h->pred8x8l[DIAG_DOWN_LEFT_PRED ]= pred8x8l_down_left_c;
1056     h->pred8x8l[DIAG_DOWN_RIGHT_PRED]= pred8x8l_down_right_c;
1057     h->pred8x8l[VERT_RIGHT_PRED     ]= pred8x8l_vertical_right_c;
1058     h->pred8x8l[HOR_DOWN_PRED       ]= pred8x8l_horizontal_down_c;
1059     h->pred8x8l[VERT_LEFT_PRED      ]= pred8x8l_vertical_left_c;
1060     h->pred8x8l[HOR_UP_PRED         ]= pred8x8l_horizontal_up_c;
1061     h->pred8x8l[LEFT_DC_PRED        ]= pred8x8l_left_dc_c;
1062     h->pred8x8l[TOP_DC_PRED         ]= pred8x8l_top_dc_c;
1063     h->pred8x8l[DC_128_PRED         ]= pred8x8l_128_dc_c;
1064
1065     h->pred8x8[VERT_PRED8x8   ]= pred8x8_vertical_c;
1066     h->pred8x8[HOR_PRED8x8    ]= pred8x8_horizontal_c;
1067     h->pred8x8[PLANE_PRED8x8  ]= pred8x8_plane_c;
1068     if(codec_id != CODEC_ID_RV40){
1069         h->pred8x8[DC_PRED8x8     ]= pred8x8_dc_c;
1070         h->pred8x8[LEFT_DC_PRED8x8]= pred8x8_left_dc_c;
1071         h->pred8x8[TOP_DC_PRED8x8 ]= pred8x8_top_dc_c;
1072         h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= pred8x8_mad_cow_dc_l0t;
1073         h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= pred8x8_mad_cow_dc_0lt;
1074         h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= pred8x8_mad_cow_dc_l00;
1075         h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= pred8x8_mad_cow_dc_0l0;
1076     }else{
1077         h->pred8x8[DC_PRED8x8     ]= pred8x8_dc_rv40_c;
1078         h->pred8x8[LEFT_DC_PRED8x8]= pred8x8_left_dc_rv40_c;
1079         h->pred8x8[TOP_DC_PRED8x8 ]= pred8x8_top_dc_rv40_c;
1080     }
1081     h->pred8x8[DC_128_PRED8x8 ]= pred8x8_128_dc_c;
1082
1083     h->pred16x16[DC_PRED8x8     ]= pred16x16_dc_c;
1084     h->pred16x16[VERT_PRED8x8   ]= pred16x16_vertical_c;
1085     h->pred16x16[HOR_PRED8x8    ]= pred16x16_horizontal_c;
1086     h->pred16x16[PLANE_PRED8x8  ]= pred16x16_plane_c;
1087     switch(codec_id){
1088     case CODEC_ID_SVQ3:
1089        h->pred16x16[PLANE_PRED8x8  ]= pred16x16_plane_svq3_c;
1090        break;
1091     case CODEC_ID_RV40:
1092        h->pred16x16[PLANE_PRED8x8  ]= pred16x16_plane_rv40_c;
1093        break;
1094     default:
1095        h->pred16x16[PLANE_PRED8x8  ]= pred16x16_plane_c;
1096     }
1097     h->pred16x16[LEFT_DC_PRED8x8]= pred16x16_left_dc_c;
1098     h->pred16x16[TOP_DC_PRED8x8 ]= pred16x16_top_dc_c;
1099     h->pred16x16[DC_128_PRED8x8 ]= pred16x16_128_dc_c;
1100 }