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