git.sesse.net Git - ffmpeg/blob - 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
  24  * H.264 / AVC / MPEG4 part10 prediction functions.
  25  * @author Michael Niedermayer <michaelni@gmx.at>
  26  */
  27
  28 #include "libavutil/avassert.h"
  29 #include "h264pred.h"
  30 #include "dsputil.h" // for ff_cropTbl
  31 #include "avcodec.h" // for AV_CODEC_ID_*
  32
  33 #define BIT_DEPTH 8
  34 #include "h264pred_template.c"
  35 #undef BIT_DEPTH
  36
  37 #define BIT_DEPTH 9
  38 #include "h264pred_template.c"
  39 #undef BIT_DEPTH
  40
  41 #define BIT_DEPTH 10
  42 #include "h264pred_template.c"
  43 #undef BIT_DEPTH
  44
  45 #define BIT_DEPTH 12
  46 #include "h264pred_template.c"
  47 #undef BIT_DEPTH
  48
  49 #define BIT_DEPTH 14
  50 #include "h264pred_template.c"
  51 #undef BIT_DEPTH
  52
  53 static void pred4x4_vertical_vp8_c(uint8_t *src, const uint8_t *topright,
  54                                    ptrdiff_t stride)
  55 {
  56     const unsigned lt = src[-1-1*stride];
  57     LOAD_TOP_EDGE
  58     LOAD_TOP_RIGHT_EDGE
  59     uint32_t v = PACK_4U8((lt + 2*t0 + t1 + 2) >> 2,
  60                           (t0 + 2*t1 + t2 + 2) >> 2,
  61                           (t1 + 2*t2 + t3 + 2) >> 2,
  62                           (t2 + 2*t3 + t4 + 2) >> 2);
  63
  64     AV_WN32A(src+0*stride, v);
  65     AV_WN32A(src+1*stride, v);
  66     AV_WN32A(src+2*stride, v);
  67     AV_WN32A(src+3*stride, v);
  68 }
  69
  70 static void pred4x4_horizontal_vp8_c(uint8_t *src, const uint8_t *topright,
  71                                      ptrdiff_t stride)
  72 {
  73     const unsigned lt = src[-1-1*stride];
  74     LOAD_LEFT_EDGE
  75
  76     AV_WN32A(src+0*stride, ((lt + 2*l0 + l1 + 2) >> 2)*0x01010101);
  77     AV_WN32A(src+1*stride, ((l0 + 2*l1 + l2 + 2) >> 2)*0x01010101);
  78     AV_WN32A(src+2*stride, ((l1 + 2*l2 + l3 + 2) >> 2)*0x01010101);
  79     AV_WN32A(src+3*stride, ((l2 + 2*l3 + l3 + 2) >> 2)*0x01010101);
  80 }
  81
  82 static void pred4x4_down_left_svq3_c(uint8_t *src, const uint8_t *topright,
  83                                      ptrdiff_t stride)
  84 {
  85     LOAD_TOP_EDGE
  86     LOAD_LEFT_EDGE
  87
  88     src[0+0*stride]=(l1 + t1)>>1;
  89     src[1+0*stride]=
  90     src[0+1*stride]=(l2 + t2)>>1;
  91     src[2+0*stride]=
  92     src[1+1*stride]=
  93     src[0+2*stride]=
  94     src[3+0*stride]=
  95     src[2+1*stride]=
  96     src[1+2*stride]=
  97     src[0+3*stride]=
  98     src[3+1*stride]=
  99     src[2+2*stride]=
 100     src[1+3*stride]=
 101     src[3+2*stride]=
 102     src[2+3*stride]=
 103     src[3+3*stride]=(l3 + t3)>>1;
 104 }
 105
 106 static void pred4x4_down_left_rv40_c(uint8_t *src, const uint8_t *topright,
 107                                      ptrdiff_t stride)
 108 {
 109     LOAD_TOP_EDGE
 110     LOAD_TOP_RIGHT_EDGE
 111     LOAD_LEFT_EDGE
 112     LOAD_DOWN_LEFT_EDGE
 113
 114     src[0+0*stride]=(t0 + t2 + 2*t1 + 2 + l0 + l2 + 2*l1 + 2)>>3;
 115     src[1+0*stride]=
 116     src[0+1*stride]=(t1 + t3 + 2*t2 + 2 + l1 + l3 + 2*l2 + 2)>>3;
 117     src[2+0*stride]=
 118     src[1+1*stride]=
 119     src[0+2*stride]=(t2 + t4 + 2*t3 + 2 + l2 + l4 + 2*l3 + 2)>>3;
 120     src[3+0*stride]=
 121     src[2+1*stride]=
 122     src[1+2*stride]=
 123     src[0+3*stride]=(t3 + t5 + 2*t4 + 2 + l3 + l5 + 2*l4 + 2)>>3;
 124     src[3+1*stride]=
 125     src[2+2*stride]=
 126     src[1+3*stride]=(t4 + t6 + 2*t5 + 2 + l4 + l6 + 2*l5 + 2)>>3;
 127     src[3+2*stride]=
 128     src[2+3*stride]=(t5 + t7 + 2*t6 + 2 + l5 + l7 + 2*l6 + 2)>>3;
 129     src[3+3*stride]=(t6 + t7 + 1 + l6 + l7 + 1)>>2;
 130 }
 131
 132 static void pred4x4_down_left_rv40_nodown_c(uint8_t *src,
 133                                             const uint8_t *topright,
 134                                             ptrdiff_t stride)
 135 {
 136     LOAD_TOP_EDGE
 137     LOAD_TOP_RIGHT_EDGE
 138     LOAD_LEFT_EDGE
 139
 140     src[0+0*stride]=(t0 + t2 + 2*t1 + 2 + l0 + l2 + 2*l1 + 2)>>3;
 141     src[1+0*stride]=
 142     src[0+1*stride]=(t1 + t3 + 2*t2 + 2 + l1 + l3 + 2*l2 + 2)>>3;
 143     src[2+0*stride]=
 144     src[1+1*stride]=
 145     src[0+2*stride]=(t2 + t4 + 2*t3 + 2 + l2 + 3*l3 + 2)>>3;
 146     src[3+0*stride]=
 147     src[2+1*stride]=
 148     src[1+2*stride]=
 149     src[0+3*stride]=(t3 + t5 + 2*t4 + 2 + l3*4 + 2)>>3;
 150     src[3+1*stride]=
 151     src[2+2*stride]=
 152     src[1+3*stride]=(t4 + t6 + 2*t5 + 2 + l3*4 + 2)>>3;
 153     src[3+2*stride]=
 154     src[2+3*stride]=(t5 + t7 + 2*t6 + 2 + l3*4 + 2)>>3;
 155     src[3+3*stride]=(t6 + t7 + 1 + 2*l3 + 1)>>2;
 156 }
 157
 158 static void pred4x4_vertical_left_rv40(uint8_t *src, const uint8_t *topright,
 159                                        ptrdiff_t stride,
 160                                        const int l0, const int l1, const int l2,
 161                                        const int l3, const int l4)
 162 {
 163     LOAD_TOP_EDGE
 164     LOAD_TOP_RIGHT_EDGE
 165
 166     src[0+0*stride]=(2*t0 + 2*t1 + l1 + 2*l2 + l3 + 4)>>3;
 167     src[1+0*stride]=
 168     src[0+2*stride]=(t1 + t2 + 1)>>1;
 169     src[2+0*stride]=
 170     src[1+2*stride]=(t2 + t3 + 1)>>1;
 171     src[3+0*stride]=
 172     src[2+2*stride]=(t3 + t4+ 1)>>1;
 173     src[3+2*stride]=(t4 + t5+ 1)>>1;
 174     src[0+1*stride]=(t0 + 2*t1 + t2 + l2 + 2*l3 + l4 + 4)>>3;
 175     src[1+1*stride]=
 176     src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
 177     src[2+1*stride]=
 178     src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
 179     src[3+1*stride]=
 180     src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
 181     src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
 182 }
 183
 184 static void pred4x4_vertical_left_rv40_c(uint8_t *src, const uint8_t *topright,
 185                                          ptrdiff_t stride)
 186 {
 187     LOAD_LEFT_EDGE
 188     LOAD_DOWN_LEFT_EDGE
 189
 190     pred4x4_vertical_left_rv40(src, topright, stride, l0, l1, l2, l3, l4);
 191 }
 192
 193 static void pred4x4_vertical_left_rv40_nodown_c(uint8_t *src,
 194                                                 const uint8_t *topright,
 195                                                 ptrdiff_t stride)
 196 {
 197     LOAD_LEFT_EDGE
 198
 199     pred4x4_vertical_left_rv40(src, topright, stride, l0, l1, l2, l3, l3);
 200 }
 201
 202 static void pred4x4_vertical_left_vp8_c(uint8_t *src, const uint8_t *topright,
 203                                         ptrdiff_t stride)
 204 {
 205     LOAD_TOP_EDGE
 206     LOAD_TOP_RIGHT_EDGE
 207
 208     src[0+0*stride]=(t0 + t1 + 1)>>1;
 209     src[1+0*stride]=
 210     src[0+2*stride]=(t1 + t2 + 1)>>1;
 211     src[2+0*stride]=
 212     src[1+2*stride]=(t2 + t3 + 1)>>1;
 213     src[3+0*stride]=
 214     src[2+2*stride]=(t3 + t4 + 1)>>1;
 215     src[0+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
 216     src[1+1*stride]=
 217     src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
 218     src[2+1*stride]=
 219     src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
 220     src[3+1*stride]=
 221     src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
 222     src[3+2*stride]=(t4 + 2*t5 + t6 + 2)>>2;
 223     src[3+3*stride]=(t5 + 2*t6 + t7 + 2)>>2;
 224 }
 225
 226 static void pred4x4_horizontal_up_rv40_c(uint8_t *src, const uint8_t *topright,
 227                                          ptrdiff_t stride)
 228 {
 229     LOAD_LEFT_EDGE
 230     LOAD_DOWN_LEFT_EDGE
 231     LOAD_TOP_EDGE
 232     LOAD_TOP_RIGHT_EDGE
 233
 234     src[0+0*stride]=(t1 + 2*t2 + t3 + 2*l0 + 2*l1 + 4)>>3;
 235     src[1+0*stride]=(t2 + 2*t3 + t4 + l0 + 2*l1 + l2 + 4)>>3;
 236     src[2+0*stride]=
 237     src[0+1*stride]=(t3 + 2*t4 + t5 + 2*l1 + 2*l2 + 4)>>3;
 238     src[3+0*stride]=
 239     src[1+1*stride]=(t4 + 2*t5 + t6 + l1 + 2*l2 + l3 + 4)>>3;
 240     src[2+1*stride]=
 241     src[0+2*stride]=(t5 + 2*t6 + t7 + 2*l2 + 2*l3 + 4)>>3;
 242     src[3+1*stride]=
 243     src[1+2*stride]=(t6 + 3*t7 + l2 + 3*l3 + 4)>>3;
 244     src[3+2*stride]=
 245     src[1+3*stride]=(l3 + 2*l4 + l5 + 2)>>2;
 246     src[0+3*stride]=
 247     src[2+2*stride]=(t6 + t7 + l3 + l4 + 2)>>2;
 248     src[2+3*stride]=(l4 + l5 + 1)>>1;
 249     src[3+3*stride]=(l4 + 2*l5 + l6 + 2)>>2;
 250 }
 251
 252 static void pred4x4_horizontal_up_rv40_nodown_c(uint8_t *src,
 253                                                 const uint8_t *topright,
 254                                                 ptrdiff_t stride)
 255 {
 256     LOAD_LEFT_EDGE
 257     LOAD_TOP_EDGE
 258     LOAD_TOP_RIGHT_EDGE
 259
 260     src[0+0*stride]=(t1 + 2*t2 + t3 + 2*l0 + 2*l1 + 4)>>3;
 261     src[1+0*stride]=(t2 + 2*t3 + t4 + l0 + 2*l1 + l2 + 4)>>3;
 262     src[2+0*stride]=
 263     src[0+1*stride]=(t3 + 2*t4 + t5 + 2*l1 + 2*l2 + 4)>>3;
 264     src[3+0*stride]=
 265     src[1+1*stride]=(t4 + 2*t5 + t6 + l1 + 2*l2 + l3 + 4)>>3;
 266     src[2+1*stride]=
 267     src[0+2*stride]=(t5 + 2*t6 + t7 + 2*l2 + 2*l3 + 4)>>3;
 268     src[3+1*stride]=
 269     src[1+2*stride]=(t6 + 3*t7 + l2 + 3*l3 + 4)>>3;
 270     src[3+2*stride]=
 271     src[1+3*stride]=l3;
 272     src[0+3*stride]=
 273     src[2+2*stride]=(t6 + t7 + 2*l3 + 2)>>2;
 274     src[2+3*stride]=
 275     src[3+3*stride]=l3;
 276 }
 277
 278 static void pred4x4_tm_vp8_c(uint8_t *src, const uint8_t *topright,
 279                              ptrdiff_t stride)
 280 {
 281     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
 282     uint8_t *top = src-stride;
 283     int y;
 284
 285     for (y = 0; y < 4; y++) {
 286         uint8_t *cm_in = cm + src[-1];
 287         src[0] = cm_in[top[0]];
 288         src[1] = cm_in[top[1]];
 289         src[2] = cm_in[top[2]];
 290         src[3] = cm_in[top[3]];
 291         src += stride;
 292     }
 293 }
 294
 295 static void pred16x16_plane_svq3_c(uint8_t *src, ptrdiff_t stride)
 296 {
 297     pred16x16_plane_compat_8_c(src, stride, 1, 0);
 298 }
 299
 300 static void pred16x16_plane_rv40_c(uint8_t *src, ptrdiff_t stride)
 301 {
 302     pred16x16_plane_compat_8_c(src, stride, 0, 1);
 303 }
 304
 305 static void pred16x16_tm_vp8_c(uint8_t *src, ptrdiff_t stride)
 306 {
 307     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
 308     uint8_t *top = src-stride;
 309     int y;
 310
 311     for (y = 0; y < 16; y++) {
 312         uint8_t *cm_in = cm + src[-1];
 313         src[0]  = cm_in[top[0]];
 314         src[1]  = cm_in[top[1]];
 315         src[2]  = cm_in[top[2]];
 316         src[3]  = cm_in[top[3]];
 317         src[4]  = cm_in[top[4]];
 318         src[5]  = cm_in[top[5]];
 319         src[6]  = cm_in[top[6]];
 320         src[7]  = cm_in[top[7]];
 321         src[8]  = cm_in[top[8]];
 322         src[9]  = cm_in[top[9]];
 323         src[10] = cm_in[top[10]];
 324         src[11] = cm_in[top[11]];
 325         src[12] = cm_in[top[12]];
 326         src[13] = cm_in[top[13]];
 327         src[14] = cm_in[top[14]];
 328         src[15] = cm_in[top[15]];
 329         src += stride;
 330     }
 331 }
 332
 333 static void pred8x8_left_dc_rv40_c(uint8_t *src, ptrdiff_t stride)
 334 {
 335     int i;
 336     unsigned dc0;
 337
 338     dc0=0;
 339     for(i=0;i<8; i++)
 340         dc0+= src[-1+i*stride];
 341     dc0= 0x01010101*((dc0 + 4)>>3);
 342
 343     for(i=0; i<8; i++){
 344         ((uint32_t*)(src+i*stride))[0]=
 345         ((uint32_t*)(src+i*stride))[1]= dc0;
 346     }
 347 }
 348
 349 static void pred8x8_top_dc_rv40_c(uint8_t *src, ptrdiff_t stride)
 350 {
 351     int i;
 352     unsigned dc0;
 353
 354     dc0=0;
 355     for(i=0;i<8; i++)
 356         dc0+= src[i-stride];
 357     dc0= 0x01010101*((dc0 + 4)>>3);
 358
 359     for(i=0; i<8; i++){
 360         ((uint32_t*)(src+i*stride))[0]=
 361         ((uint32_t*)(src+i*stride))[1]= dc0;
 362     }
 363 }
 364
 365 static void pred8x8_dc_rv40_c(uint8_t *src, ptrdiff_t stride)
 366 {
 367     int i;
 368     unsigned dc0 = 0;
 369
 370     for(i=0;i<4; i++){
 371         dc0+= src[-1+i*stride] + src[i-stride];
 372         dc0+= src[4+i-stride];
 373         dc0+= src[-1+(i+4)*stride];
 374     }
 375     dc0= 0x01010101*((dc0 + 8)>>4);
 376
 377     for(i=0; i<4; i++){
 378         ((uint32_t*)(src+i*stride))[0]= dc0;
 379         ((uint32_t*)(src+i*stride))[1]= dc0;
 380     }
 381     for(i=4; i<8; i++){
 382         ((uint32_t*)(src+i*stride))[0]= dc0;
 383         ((uint32_t*)(src+i*stride))[1]= dc0;
 384     }
 385 }
 386
 387 static void pred8x8_tm_vp8_c(uint8_t *src, ptrdiff_t stride)
 388 {
 389     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
 390     uint8_t *top = src-stride;
 391     int y;
 392
 393     for (y = 0; y < 8; y++) {
 394         uint8_t *cm_in = cm + src[-1];
 395         src[0] = cm_in[top[0]];
 396         src[1] = cm_in[top[1]];
 397         src[2] = cm_in[top[2]];
 398         src[3] = cm_in[top[3]];
 399         src[4] = cm_in[top[4]];
 400         src[5] = cm_in[top[5]];
 401         src[6] = cm_in[top[6]];
 402         src[7] = cm_in[top[7]];
 403         src += stride;
 404     }
 405 }
 406
 407 /**
 408  * Set the intra prediction function pointers.
 409  */
 410 void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth,
 411                        int chroma_format_idc)
 412 {
 413 #undef FUNC
 414 #undef FUNCC
 415 #define FUNC(a, depth) a ## _ ## depth
 416 #define FUNCC(a, depth) a ## _ ## depth ## _c
 417 #define FUNCD(a) a ## _c
 418
 419 #define H264_PRED(depth) \
 420     if(codec_id != AV_CODEC_ID_RV40){\
 421         if(codec_id == AV_CODEC_ID_VP8) {\
 422             h->pred4x4[VERT_PRED       ]= FUNCD(pred4x4_vertical_vp8);\
 423             h->pred4x4[HOR_PRED        ]= FUNCD(pred4x4_horizontal_vp8);\
 424         } else {\
 425             h->pred4x4[VERT_PRED       ]= FUNCC(pred4x4_vertical          , depth);\
 426             h->pred4x4[HOR_PRED        ]= FUNCC(pred4x4_horizontal        , depth);\
 427         }\
 428         h->pred4x4[DC_PRED             ]= FUNCC(pred4x4_dc                , depth);\
 429         if(codec_id == AV_CODEC_ID_SVQ3)\
 430             h->pred4x4[DIAG_DOWN_LEFT_PRED ]= FUNCD(pred4x4_down_left_svq3);\
 431         else\
 432             h->pred4x4[DIAG_DOWN_LEFT_PRED ]= FUNCC(pred4x4_down_left     , depth);\
 433         h->pred4x4[DIAG_DOWN_RIGHT_PRED]= FUNCC(pred4x4_down_right        , depth);\
 434         h->pred4x4[VERT_RIGHT_PRED     ]= FUNCC(pred4x4_vertical_right    , depth);\
 435         h->pred4x4[HOR_DOWN_PRED       ]= FUNCC(pred4x4_horizontal_down   , depth);\
 436         if (codec_id == AV_CODEC_ID_VP8) {\
 437             h->pred4x4[VERT_LEFT_PRED  ]= FUNCD(pred4x4_vertical_left_vp8);\
 438         } else\
 439             h->pred4x4[VERT_LEFT_PRED  ]= FUNCC(pred4x4_vertical_left     , depth);\
 440         h->pred4x4[HOR_UP_PRED         ]= FUNCC(pred4x4_horizontal_up     , depth);\
 441         if(codec_id != AV_CODEC_ID_VP8) {\
 442             h->pred4x4[LEFT_DC_PRED    ]= FUNCC(pred4x4_left_dc           , depth);\
 443             h->pred4x4[TOP_DC_PRED     ]= FUNCC(pred4x4_top_dc            , depth);\
 444             h->pred4x4[DC_128_PRED     ]= FUNCC(pred4x4_128_dc            , depth);\
 445         } else {\
 446             h->pred4x4[TM_VP8_PRED     ]= FUNCD(pred4x4_tm_vp8);\
 447             h->pred4x4[DC_127_PRED     ]= FUNCC(pred4x4_127_dc            , depth);\
 448             h->pred4x4[DC_129_PRED     ]= FUNCC(pred4x4_129_dc            , depth);\
 449             h->pred4x4[VERT_VP8_PRED   ]= FUNCC(pred4x4_vertical          , depth);\
 450             h->pred4x4[HOR_VP8_PRED    ]= FUNCC(pred4x4_horizontal        , depth);\
 451         }\
 452     }else{\
 453         h->pred4x4[VERT_PRED           ]= FUNCC(pred4x4_vertical          , depth);\
 454         h->pred4x4[HOR_PRED            ]= FUNCC(pred4x4_horizontal        , depth);\
 455         h->pred4x4[DC_PRED             ]= FUNCC(pred4x4_dc                , depth);\
 456         h->pred4x4[DIAG_DOWN_LEFT_PRED ]= FUNCD(pred4x4_down_left_rv40);\
 457         h->pred4x4[DIAG_DOWN_RIGHT_PRED]= FUNCC(pred4x4_down_right        , depth);\
 458         h->pred4x4[VERT_RIGHT_PRED     ]= FUNCC(pred4x4_vertical_right    , depth);\
 459         h->pred4x4[HOR_DOWN_PRED       ]= FUNCC(pred4x4_horizontal_down   , depth);\
 460         h->pred4x4[VERT_LEFT_PRED      ]= FUNCD(pred4x4_vertical_left_rv40);\
 461         h->pred4x4[HOR_UP_PRED         ]= FUNCD(pred4x4_horizontal_up_rv40);\
 462         h->pred4x4[LEFT_DC_PRED        ]= FUNCC(pred4x4_left_dc           , depth);\
 463         h->pred4x4[TOP_DC_PRED         ]= FUNCC(pred4x4_top_dc            , depth);\
 464         h->pred4x4[DC_128_PRED         ]= FUNCC(pred4x4_128_dc            , depth);\
 465         h->pred4x4[DIAG_DOWN_LEFT_PRED_RV40_NODOWN]= FUNCD(pred4x4_down_left_rv40_nodown);\
 466         h->pred4x4[HOR_UP_PRED_RV40_NODOWN]= FUNCD(pred4x4_horizontal_up_rv40_nodown);\
 467         h->pred4x4[VERT_LEFT_PRED_RV40_NODOWN]= FUNCD(pred4x4_vertical_left_rv40_nodown);\
 468     }\
 469 \
 470     h->pred8x8l[VERT_PRED           ]= FUNCC(pred8x8l_vertical            , depth);\
 471     h->pred8x8l[HOR_PRED            ]= FUNCC(pred8x8l_horizontal          , depth);\
 472     h->pred8x8l[DC_PRED             ]= FUNCC(pred8x8l_dc                  , depth);\
 473     h->pred8x8l[DIAG_DOWN_LEFT_PRED ]= FUNCC(pred8x8l_down_left           , depth);\
 474     h->pred8x8l[DIAG_DOWN_RIGHT_PRED]= FUNCC(pred8x8l_down_right          , depth);\
 475     h->pred8x8l[VERT_RIGHT_PRED     ]= FUNCC(pred8x8l_vertical_right      , depth);\
 476     h->pred8x8l[HOR_DOWN_PRED       ]= FUNCC(pred8x8l_horizontal_down     , depth);\
 477     h->pred8x8l[VERT_LEFT_PRED      ]= FUNCC(pred8x8l_vertical_left       , depth);\
 478     h->pred8x8l[HOR_UP_PRED         ]= FUNCC(pred8x8l_horizontal_up       , depth);\
 479     h->pred8x8l[LEFT_DC_PRED        ]= FUNCC(pred8x8l_left_dc             , depth);\
 480     h->pred8x8l[TOP_DC_PRED         ]= FUNCC(pred8x8l_top_dc              , depth);\
 481     h->pred8x8l[DC_128_PRED         ]= FUNCC(pred8x8l_128_dc              , depth);\
 482 \
 483     if (chroma_format_idc == 1) {\
 484         h->pred8x8[VERT_PRED8x8   ]= FUNCC(pred8x8_vertical               , depth);\
 485         h->pred8x8[HOR_PRED8x8    ]= FUNCC(pred8x8_horizontal             , depth);\
 486     } else {\
 487         h->pred8x8[VERT_PRED8x8   ]= FUNCC(pred8x16_vertical              , depth);\
 488         h->pred8x8[HOR_PRED8x8    ]= FUNCC(pred8x16_horizontal            , depth);\
 489     }\
 490     if (codec_id != AV_CODEC_ID_VP8) {\
 491         if (chroma_format_idc == 1) {\
 492             h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x8_plane                , depth);\
 493         } else {\
 494             h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x16_plane               , depth);\
 495         }\
 496     } else\
 497         h->pred8x8[PLANE_PRED8x8]= FUNCD(pred8x8_tm_vp8);\
 498     if(codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP8){\
 499         if (chroma_format_idc == 1) {\
 500             h->pred8x8[DC_PRED8x8     ]= FUNCC(pred8x8_dc                     , depth);\
 501             h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x8_left_dc                , depth);\
 502             h->pred8x8[TOP_DC_PRED8x8 ]= FUNCC(pred8x8_top_dc                 , depth);\
 503             h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l0t, depth);\
 504             h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0lt, depth);\
 505             h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l00, depth);\
 506             h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0l0, depth);\
 507         } else {\
 508             h->pred8x8[DC_PRED8x8     ]= FUNCC(pred8x16_dc                    , depth);\
 509             h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x16_left_dc               , depth);\
 510             h->pred8x8[TOP_DC_PRED8x8 ]= FUNCC(pred8x16_top_dc                , depth);\
 511             h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= FUNC(pred8x16_mad_cow_dc_l0t, depth);\
 512             h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= FUNC(pred8x16_mad_cow_dc_0lt, depth);\
 513             h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= FUNC(pred8x16_mad_cow_dc_l00, depth);\
 514             h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= FUNC(pred8x16_mad_cow_dc_0l0, depth);\
 515         }\
 516     }else{\
 517         h->pred8x8[DC_PRED8x8     ]= FUNCD(pred8x8_dc_rv40);\
 518         h->pred8x8[LEFT_DC_PRED8x8]= FUNCD(pred8x8_left_dc_rv40);\
 519         h->pred8x8[TOP_DC_PRED8x8 ]= FUNCD(pred8x8_top_dc_rv40);\
 520         if (codec_id == AV_CODEC_ID_VP8) {\
 521             h->pred8x8[DC_127_PRED8x8]= FUNCC(pred8x8_127_dc              , depth);\
 522             h->pred8x8[DC_129_PRED8x8]= FUNCC(pred8x8_129_dc              , depth);\
 523         }\
 524     }\
 525     if (chroma_format_idc == 1) {\
 526         h->pred8x8[DC_128_PRED8x8 ]= FUNCC(pred8x8_128_dc                 , depth);\
 527     } else {\
 528         h->pred8x8[DC_128_PRED8x8 ]= FUNCC(pred8x16_128_dc                , depth);\
 529     }\
 530 \
 531     h->pred16x16[DC_PRED8x8     ]= FUNCC(pred16x16_dc                     , depth);\
 532     h->pred16x16[VERT_PRED8x8   ]= FUNCC(pred16x16_vertical               , depth);\
 533     h->pred16x16[HOR_PRED8x8    ]= FUNCC(pred16x16_horizontal             , depth);\
 534     switch(codec_id){\
 535     case AV_CODEC_ID_SVQ3:\
 536        h->pred16x16[PLANE_PRED8x8  ]= FUNCD(pred16x16_plane_svq3);\
 537        break;\
 538     case AV_CODEC_ID_RV40:\
 539        h->pred16x16[PLANE_PRED8x8  ]= FUNCD(pred16x16_plane_rv40);\
 540        break;\
 541     case AV_CODEC_ID_VP8:\
 542        h->pred16x16[PLANE_PRED8x8  ]= FUNCD(pred16x16_tm_vp8);\
 543        h->pred16x16[DC_127_PRED8x8]= FUNCC(pred16x16_127_dc               , depth);\
 544        h->pred16x16[DC_129_PRED8x8]= FUNCC(pred16x16_129_dc               , depth);\
 545        break;\
 546     default:\
 547        h->pred16x16[PLANE_PRED8x8  ]= FUNCC(pred16x16_plane               , depth);\
 548        break;\
 549     }\
 550     h->pred16x16[LEFT_DC_PRED8x8]= FUNCC(pred16x16_left_dc                , depth);\
 551     h->pred16x16[TOP_DC_PRED8x8 ]= FUNCC(pred16x16_top_dc                 , depth);\
 552     h->pred16x16[DC_128_PRED8x8 ]= FUNCC(pred16x16_128_dc                 , depth);\
 553 \
 554     /* special lossless h/v prediction for h264 */ \
 555     h->pred4x4_add  [VERT_PRED   ]= FUNCC(pred4x4_vertical_add            , depth);\
 556     h->pred4x4_add  [ HOR_PRED   ]= FUNCC(pred4x4_horizontal_add          , depth);\
 557     h->pred8x8l_add [VERT_PRED   ]= FUNCC(pred8x8l_vertical_add           , depth);\
 558     h->pred8x8l_add [ HOR_PRED   ]= FUNCC(pred8x8l_horizontal_add         , depth);\
 559     if (chroma_format_idc == 1) {\
 560     h->pred8x8_add  [VERT_PRED8x8]= FUNCC(pred8x8_vertical_add            , depth);\
 561     h->pred8x8_add  [ HOR_PRED8x8]= FUNCC(pred8x8_horizontal_add          , depth);\
 562     } else {\
 563         h->pred8x8_add  [VERT_PRED8x8]= FUNCC(pred8x16_vertical_add            , depth);\
 564         h->pred8x8_add  [ HOR_PRED8x8]= FUNCC(pred8x16_horizontal_add          , depth);\
 565     }\
 566     h->pred16x16_add[VERT_PRED8x8]= FUNCC(pred16x16_vertical_add          , depth);\
 567     h->pred16x16_add[ HOR_PRED8x8]= FUNCC(pred16x16_horizontal_add        , depth);\
 568
 569     if(!chroma_format_idc)
 570         chroma_format_idc = 1;
 571
 572     switch (bit_depth) {
 573         case 9:
 574             H264_PRED(9)
 575             break;
 576         case 10:
 577             H264_PRED(10)
 578             break;
 579         case 12:
 580             H264_PRED(12)
 581             break;
 582         case 14:
 583             H264_PRED(14)
 584             break;
 585         default:
 586             av_assert0(bit_depth<=8);
 587             H264_PRED(8)
 588             break;
 589     }
 590
 591     if (ARCH_ARM) ff_h264_pred_init_arm(h, codec_id, bit_depth, chroma_format_idc);
 592     if (ARCH_X86) ff_h264_pred_init_x86(h, codec_id, bit_depth, chroma_format_idc);
 593 }