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