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