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