h264.c 209 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
/*
 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 */
 
/**
 * @file h264.c
 * H.264 / AVC / MPEG4 part10 codec.
 * @author Michael Niedermayer <michaelni@gmx.at>
 */

#include "common.h"
#include "dsputil.h"
#include "avcodec.h"
#include "mpegvideo.h"
#include "h264data.h"
#include "golomb.h"

34 35
#include "cabac.h"

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71
#undef NDEBUG
#include <assert.h>

#define interlaced_dct interlaced_dct_is_a_bad_name
#define mb_intra mb_intra_isnt_initalized_see_mb_type

#define LUMA_DC_BLOCK_INDEX   25
#define CHROMA_DC_BLOCK_INDEX 26

#define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
#define COEFF_TOKEN_VLC_BITS           8
#define TOTAL_ZEROS_VLC_BITS           9
#define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
#define RUN_VLC_BITS                   3
#define RUN7_VLC_BITS                  6

#define MAX_SPS_COUNT 32
#define MAX_PPS_COUNT 256

#define MAX_MMCO_COUNT 66

/**
 * Sequence parameter set
 */
typedef struct SPS{
    
    int profile_idc;
    int level_idc;
    int log2_max_frame_num;            ///< log2_max_frame_num_minus4 + 4
    int poc_type;                      ///< pic_order_cnt_type
    int log2_max_poc_lsb;              ///< log2_max_pic_order_cnt_lsb_minus4
    int delta_pic_order_always_zero_flag;
    int offset_for_non_ref_pic;
    int offset_for_top_to_bottom_field;
    int poc_cycle_length;              ///< num_ref_frames_in_pic_order_cnt_cycle
    int ref_frame_count;               ///< num_ref_frames
72
    int gaps_in_frame_num_allowed_flag;
73 74 75 76 77
    int mb_width;                      ///< frame_width_in_mbs_minus1 + 1
    int mb_height;                     ///< frame_height_in_mbs_minus1 + 1
    int frame_mbs_only_flag;
    int mb_aff;                        ///<mb_adaptive_frame_field_flag
    int direct_8x8_inference_flag;
78 79 80 81 82
    int crop;                   ///< frame_cropping_flag
    int crop_left;              ///< frame_cropping_rect_left_offset
    int crop_right;             ///< frame_cropping_rect_right_offset
    int crop_top;               ///< frame_cropping_rect_top_offset
    int crop_bottom;            ///< frame_cropping_rect_bottom_offset
83
    int vui_parameters_present_flag;
Michael Niedermayer's avatar
Michael Niedermayer committed
84
    AVRational sar;
85 86 87 88
    int timing_info_present_flag;
    uint32_t num_units_in_tick;
    uint32_t time_scale;
    int fixed_frame_rate_flag;
89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
    short offset_for_ref_frame[256]; //FIXME dyn aloc?
}SPS;

/**
 * Picture parameter set
 */
typedef struct PPS{
    int sps_id;
    int cabac;                  ///< entropy_coding_mode_flag
    int pic_order_present;      ///< pic_order_present_flag
    int slice_group_count;      ///< num_slice_groups_minus1 + 1
    int mb_slice_group_map_type;
    int ref_count[2];           ///< num_ref_idx_l0/1_active_minus1 + 1
    int weighted_pred;          ///< weighted_pred_flag
    int weighted_bipred_idc;
    int init_qp;                ///< pic_init_qp_minus26 + 26
    int init_qs;                ///< pic_init_qs_minus26 + 26
    int chroma_qp_index_offset;
    int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
    int constrained_intra_pred; ///< constrained_intra_pred_flag
    int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
}PPS;

/**
 * Memory management control operation opcode.
 */
typedef enum MMCOOpcode{
    MMCO_END=0,
    MMCO_SHORT2UNUSED,
    MMCO_LONG2UNUSED,
    MMCO_SHORT2LONG,
    MMCO_SET_MAX_LONG,
    MMCO_RESET, 
    MMCO_LONG,
} MMCOOpcode;

/**
 * Memory management control operation.
 */
typedef struct MMCO{
    MMCOOpcode opcode;
    int short_frame_num;
    int long_index;
} MMCO;

/**
 * H264Context
 */
typedef struct H264Context{
    MpegEncContext s;
    int nal_ref_idc;	
    int nal_unit_type;
#define NAL_SLICE		1
#define NAL_DPA			2
#define NAL_DPB			3
#define NAL_DPC			4
#define NAL_IDR_SLICE		5
#define NAL_SEI			6
#define NAL_SPS			7
#define NAL_PPS			8
#define NAL_PICTURE_DELIMITER	9
#define NAL_FILTER_DATA		10
    uint8_t *rbsp_buffer;
    int rbsp_buffer_size;

154 155 156 157 158 159 160
    /**
      * Used to parse AVC variant of h264
      */
    int is_avc; ///< this flag is != 0 if codec is avc1
    int got_avcC; ///< flag used to parse avcC data only once
    int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)

161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
    int chroma_qp; //QPc

    int prev_mb_skiped; //FIXME remove (IMHO not used)

    //prediction stuff
    int chroma_pred_mode;
    int intra16x16_pred_mode;
    
    int8_t intra4x4_pred_mode_cache[5*8];
    int8_t (*intra4x4_pred_mode)[8];
    void (*pred4x4  [9+3])(uint8_t *src, uint8_t *topright, int stride);//FIXME move to dsp?
    void (*pred8x8  [4+3])(uint8_t *src, int stride);
    void (*pred16x16[4+3])(uint8_t *src, int stride);
    unsigned int topleft_samples_available;
    unsigned int top_samples_available;
    unsigned int topright_samples_available;
    unsigned int left_samples_available;
178 179
    uint8_t (*top_border)[16+2*8];
    uint8_t left_border[17+2*9];
180 181 182 183 184 185

    /**
     * non zero coeff count cache.
     * is 64 if not available.
     */
    uint8_t non_zero_count_cache[6*8];
186
    uint8_t (*non_zero_count)[16];
187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208

    /**
     * Motion vector cache.
     */
    int16_t mv_cache[2][5*8][2];
    int8_t ref_cache[2][5*8];
#define LIST_NOT_USED -1 //FIXME rename?
#define PART_NOT_AVAILABLE -2
    
    /**
     * is 1 if the specific list MV&references are set to 0,0,-2.
     */
    int mv_cache_clean[2];

    int block_offset[16+8];
    int chroma_subblock_offset[16]; //FIXME remove
    
    uint16_t *mb2b_xy; //FIXME are these 4 a good idea?
    uint16_t *mb2b8_xy;
    int b_stride;
    int b8_stride;

209 210 211
    int halfpel_flag;
    int thirdpel_flag;

212 213 214
    int unknown_svq3_flag;
    int next_slice_index;

215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265
    SPS sps_buffer[MAX_SPS_COUNT];
    SPS sps; ///< current sps
    
    PPS pps_buffer[MAX_PPS_COUNT];
    /**
     * current pps
     */
    PPS pps; //FIXME move tp Picture perhaps? (->no) do we need that?

    int slice_num;
    uint8_t *slice_table_base;
    uint8_t *slice_table;      ///< slice_table_base + mb_stride + 1
    int slice_type;
    int slice_type_fixed;
    
    //interlacing specific flags
    int mb_field_decoding_flag;
    
    int sub_mb_type[4];
    
    //POC stuff
    int poc_lsb;
    int poc_msb;
    int delta_poc_bottom;
    int delta_poc[2];
    int frame_num;
    int prev_poc_msb;             ///< poc_msb of the last reference pic for POC type 0
    int prev_poc_lsb;             ///< poc_lsb of the last reference pic for POC type 0
    int frame_num_offset;         ///< for POC type 2
    int prev_frame_num_offset;    ///< for POC type 2
    int prev_frame_num;           ///< frame_num of the last pic for POC type 1/2

    /**
     * frame_num for frames or 2*frame_num for field pics.
     */
    int curr_pic_num;
    
    /**
     * max_frame_num or 2*max_frame_num for field pics.
     */
    int max_pic_num;

    //Weighted pred stuff
    int luma_log2_weight_denom;
    int chroma_log2_weight_denom;
    int luma_weight[2][16];
    int luma_offset[2][16];
    int chroma_weight[2][16][2];
    int chroma_offset[2][16][2];
   
    //deblock
266
    int deblocking_filter;         ///< disable_deblocking_filter_idc with 1<->0 
267 268
    int slice_alpha_c0_offset;
    int slice_beta_offset;
269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299
     
    int redundant_pic_count;
    
    int direct_spatial_mv_pred;

    /**
     * num_ref_idx_l0/1_active_minus1 + 1
     */
    int ref_count[2];// FIXME split for AFF
    Picture *short_ref[16];
    Picture *long_ref[16];
    Picture default_ref_list[2][32];
    Picture ref_list[2][32]; //FIXME size?
    Picture field_ref_list[2][32]; //FIXME size?
    
    /**
     * memory management control operations buffer.
     */
    MMCO mmco[MAX_MMCO_COUNT];
    int mmco_index;
    
    int long_ref_count;  ///< number of actual long term references
    int short_ref_count; ///< number of actual short term references
    
    //data partitioning
    GetBitContext intra_gb;
    GetBitContext inter_gb;
    GetBitContext *intra_gb_ptr;
    GetBitContext *inter_gb_ptr;
    
    DCTELEM mb[16*24] __align8;
300 301 302 303 304 305 306 307 308 309

    /**
     * Cabac
     */
    CABACContext cabac;
    uint8_t      cabac_state[399];
    int          cabac_init_idc;

    /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
    uint16_t     *cbp_table;
Michael Niedermayer's avatar
Michael Niedermayer committed
310 311
    int top_cbp;
    int left_cbp;
312
    /* chroma_pred_mode for i4x4 or i16x16, else 0 */
313 314
    uint8_t     *chroma_pred_mode_table;
    int         last_qscale_diff;
315 316
    int16_t     (*mvd_table[2])[2];
    int16_t     mvd_cache[2][5*8][2];
317

318 319 320 321 322 323 324 325 326 327 328
}H264Context;

static VLC coeff_token_vlc[4];
static VLC chroma_dc_coeff_token_vlc;

static VLC total_zeros_vlc[15];
static VLC chroma_dc_total_zeros_vlc[3];

static VLC run_vlc[6];
static VLC run7_vlc;

329 330
static void svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);
static void svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
331
static void filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr);
332

Michael Niedermayer's avatar
Michael Niedermayer committed
333 334 335 336 337 338 339 340
static inline uint32_t pack16to32(int a, int b){
#ifdef WORDS_BIGENDIAN
   return (b&0xFFFF) + (a<<16);
#else
   return (a&0xFFFF) + (b<<16);
#endif
}

341 342 343 344 345 346
/**
 * fill a rectangle.
 * @param h height of the recatangle, should be a constant
 * @param w width of the recatangle, should be a constant
 * @param size the size of val (1 or 4), should be a constant
 */
347 348
static inline void fill_rectangle(void *vp, int w, int h, int stride, uint32_t val, int size){ //FIXME ensure this IS inlined
    uint8_t *p= (uint8_t*)vp;
349 350 351 352 353 354 355 356 357 358 359 360 361 362
    assert(size==1 || size==4);
    
    w      *= size;
    stride *= size;
    
//FIXME check what gcc generates for 64 bit on x86 and possible write a 32 bit ver of it
    if(w==2 && h==2){
        *(uint16_t*)(p + 0)=
        *(uint16_t*)(p + stride)= size==4 ? val : val*0x0101;
    }else if(w==2 && h==4){
        *(uint16_t*)(p + 0*stride)=
        *(uint16_t*)(p + 1*stride)=
        *(uint16_t*)(p + 2*stride)=
        *(uint16_t*)(p + 3*stride)= size==4 ? val : val*0x0101;
Michael Niedermayer's avatar
Michael Niedermayer committed
363 364
    }else if(w==4 && h==1){
        *(uint32_t*)(p + 0*stride)= size==4 ? val : val*0x01010101;
365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
    }else if(w==4 && h==2){
        *(uint32_t*)(p + 0*stride)=
        *(uint32_t*)(p + 1*stride)= size==4 ? val : val*0x01010101;
    }else if(w==4 && h==4){
        *(uint32_t*)(p + 0*stride)=
        *(uint32_t*)(p + 1*stride)=
        *(uint32_t*)(p + 2*stride)=
        *(uint32_t*)(p + 3*stride)= size==4 ? val : val*0x01010101;
    }else if(w==8 && h==1){
        *(uint32_t*)(p + 0)=
        *(uint32_t*)(p + 4)= size==4 ? val : val*0x01010101;
    }else if(w==8 && h==2){
        *(uint32_t*)(p + 0 + 0*stride)=
        *(uint32_t*)(p + 4 + 0*stride)=
        *(uint32_t*)(p + 0 + 1*stride)=
        *(uint32_t*)(p + 4 + 1*stride)=  size==4 ? val : val*0x01010101;
    }else if(w==8 && h==4){
        *(uint64_t*)(p + 0*stride)=
        *(uint64_t*)(p + 1*stride)=
        *(uint64_t*)(p + 2*stride)=
        *(uint64_t*)(p + 3*stride)= size==4 ? val*0x0100000001ULL : val*0x0101010101010101ULL;
    }else if(w==16 && h==2){
        *(uint64_t*)(p + 0+0*stride)=
        *(uint64_t*)(p + 8+0*stride)=
        *(uint64_t*)(p + 0+1*stride)=
        *(uint64_t*)(p + 8+1*stride)= size==4 ? val*0x0100000001ULL : val*0x0101010101010101ULL;
    }else if(w==16 && h==4){
        *(uint64_t*)(p + 0+0*stride)=
        *(uint64_t*)(p + 8+0*stride)=
        *(uint64_t*)(p + 0+1*stride)=
        *(uint64_t*)(p + 8+1*stride)=
        *(uint64_t*)(p + 0+2*stride)=
        *(uint64_t*)(p + 8+2*stride)=
        *(uint64_t*)(p + 0+3*stride)=
        *(uint64_t*)(p + 8+3*stride)= size==4 ? val*0x0100000001ULL : val*0x0101010101010101ULL;
    }else
        assert(0);
}

static inline void fill_caches(H264Context *h, int mb_type){
    MpegEncContext * const s = &h->s;
406
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
407 408 409 410 411 412 413 414 415
    int topleft_xy, top_xy, topright_xy, left_xy[2];
    int topleft_type, top_type, topright_type, left_type[2];
    int left_block[4];
    int i;

    //wow what a mess, why didnt they simplify the interlacing&intra stuff, i cant imagine that these complex rules are worth it 
    
    if(h->sps.mb_aff){
    //FIXME
Fabrice Bellard's avatar
Fabrice Bellard committed
416 417 418
        topleft_xy = 0; /* avoid warning */
        top_xy = 0; /* avoid warning */
        topright_xy = 0; /* avoid warning */
419
    }else{
420 421 422
        topleft_xy = mb_xy-1 - s->mb_stride;
        top_xy     = mb_xy   - s->mb_stride;
        topright_xy= mb_xy+1 - s->mb_stride;
423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507
        left_xy[0]   = mb_xy-1;
        left_xy[1]   = mb_xy-1;
        left_block[0]= 0;
        left_block[1]= 1;
        left_block[2]= 2;
        left_block[3]= 3;
    }

    topleft_type = h->slice_table[topleft_xy ] == h->slice_num ? s->current_picture.mb_type[topleft_xy] : 0;
    top_type     = h->slice_table[top_xy     ] == h->slice_num ? s->current_picture.mb_type[top_xy]     : 0;
    topright_type= h->slice_table[topright_xy] == h->slice_num ? s->current_picture.mb_type[topright_xy]: 0;
    left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
    left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;

    if(IS_INTRA(mb_type)){
        h->topleft_samples_available= 
        h->top_samples_available= 
        h->left_samples_available= 0xFFFF;
        h->topright_samples_available= 0xEEEA;

        if(!IS_INTRA(top_type) && (top_type==0 || h->pps.constrained_intra_pred)){
            h->topleft_samples_available= 0xB3FF;
            h->top_samples_available= 0x33FF;
            h->topright_samples_available= 0x26EA;
        }
        for(i=0; i<2; i++){
            if(!IS_INTRA(left_type[i]) && (left_type[i]==0 || h->pps.constrained_intra_pred)){
                h->topleft_samples_available&= 0xDF5F;
                h->left_samples_available&= 0x5F5F;
            }
        }
        
        if(!IS_INTRA(topleft_type) && (topleft_type==0 || h->pps.constrained_intra_pred))
            h->topleft_samples_available&= 0x7FFF;
        
        if(!IS_INTRA(topright_type) && (topright_type==0 || h->pps.constrained_intra_pred))
            h->topright_samples_available&= 0xFBFF;
    
        if(IS_INTRA4x4(mb_type)){
            if(IS_INTRA4x4(top_type)){
                h->intra4x4_pred_mode_cache[4+8*0]= h->intra4x4_pred_mode[top_xy][4];
                h->intra4x4_pred_mode_cache[5+8*0]= h->intra4x4_pred_mode[top_xy][5];
                h->intra4x4_pred_mode_cache[6+8*0]= h->intra4x4_pred_mode[top_xy][6];
                h->intra4x4_pred_mode_cache[7+8*0]= h->intra4x4_pred_mode[top_xy][3];
            }else{
                int pred;
                if(IS_INTRA16x16(top_type) || (IS_INTER(top_type) && !h->pps.constrained_intra_pred))
                    pred= 2;
                else{
                    pred= -1;
                }
                h->intra4x4_pred_mode_cache[4+8*0]=
                h->intra4x4_pred_mode_cache[5+8*0]=
                h->intra4x4_pred_mode_cache[6+8*0]=
                h->intra4x4_pred_mode_cache[7+8*0]= pred;
            }
            for(i=0; i<2; i++){
                if(IS_INTRA4x4(left_type[i])){
                    h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[0+2*i]];
                    h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[1+2*i]];
                }else{
                    int pred;
                    if(IS_INTRA16x16(left_type[i]) || (IS_INTER(left_type[i]) && !h->pps.constrained_intra_pred))
                        pred= 2;
                    else{
                        pred= -1;
                    }
                    h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]=
                    h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= pred;
                }
            }
        }
    }
    
    
/*
0 . T T. T T T T 
1 L . .L . . . . 
2 L . .L . . . . 
3 . T TL . . . . 
4 L . .L . . . . 
5 L . .. . . . . 
*/
//FIXME constraint_intra_pred & partitioning & nnz (lets hope this is just a typo in the spec)
    if(top_type){
508 509 510 511
        h->non_zero_count_cache[4+8*0]= h->non_zero_count[top_xy][0];
        h->non_zero_count_cache[5+8*0]= h->non_zero_count[top_xy][1];
        h->non_zero_count_cache[6+8*0]= h->non_zero_count[top_xy][2];
        h->non_zero_count_cache[7+8*0]= h->non_zero_count[top_xy][3];
512
    
513 514
        h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][7];
        h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][8];
515
    
516 517
        h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][10];
        h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][11];
Michael Niedermayer's avatar
Michael Niedermayer committed
518 519
        
        h->top_cbp= h->cbp_table[top_xy];
520 521 522 523 524 525 526 527 528 529
    }else{
        h->non_zero_count_cache[4+8*0]=      
        h->non_zero_count_cache[5+8*0]=
        h->non_zero_count_cache[6+8*0]=
        h->non_zero_count_cache[7+8*0]=
    
        h->non_zero_count_cache[1+8*0]=
        h->non_zero_count_cache[2+8*0]=
    
        h->non_zero_count_cache[1+8*3]=
Michael Niedermayer's avatar
Michael Niedermayer committed
530 531 532 533
        h->non_zero_count_cache[2+8*3]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
        
        if(IS_INTRA(mb_type)) h->top_cbp= 0x1C0;
        else                  h->top_cbp= 0;
534 535 536
    }
    
    if(left_type[0]){
537 538 539 540
        h->non_zero_count_cache[3+8*1]= h->non_zero_count[left_xy[0]][6];
        h->non_zero_count_cache[3+8*2]= h->non_zero_count[left_xy[0]][5];
        h->non_zero_count_cache[0+8*1]= h->non_zero_count[left_xy[0]][9]; //FIXME left_block
        h->non_zero_count_cache[0+8*4]= h->non_zero_count[left_xy[0]][12];
Michael Niedermayer's avatar
Michael Niedermayer committed
541
        h->left_cbp= h->cbp_table[left_xy[0]]; //FIXME interlacing
542 543 544 545
    }else{
        h->non_zero_count_cache[3+8*1]= 
        h->non_zero_count_cache[3+8*2]= 
        h->non_zero_count_cache[0+8*1]= 
Michael Niedermayer's avatar
Michael Niedermayer committed
546 547 548 549
        h->non_zero_count_cache[0+8*4]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
        
        if(IS_INTRA(mb_type)) h->left_cbp= 0x1C0;//FIXME interlacing
        else                  h->left_cbp= 0;
550 551 552
    }
    
    if(left_type[1]){
553 554 555 556
        h->non_zero_count_cache[3+8*3]= h->non_zero_count[left_xy[1]][4];
        h->non_zero_count_cache[3+8*4]= h->non_zero_count[left_xy[1]][3];
        h->non_zero_count_cache[0+8*2]= h->non_zero_count[left_xy[1]][8];
        h->non_zero_count_cache[0+8*5]= h->non_zero_count[left_xy[1]][11];
557 558 559 560
    }else{
        h->non_zero_count_cache[3+8*3]= 
        h->non_zero_count_cache[3+8*4]= 
        h->non_zero_count_cache[0+8*2]= 
Michael Niedermayer's avatar
Michael Niedermayer committed
561
        h->non_zero_count_cache[0+8*5]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
    }
    
#if 1
    if(IS_INTER(mb_type)){
        int list;
        for(list=0; list<2; list++){
            if((!IS_8X8(mb_type)) && !USES_LIST(mb_type, list)){
                /*if(!h->mv_cache_clean[list]){
                    memset(h->mv_cache [list],  0, 8*5*2*sizeof(int16_t)); //FIXME clean only input? clean at all?
                    memset(h->ref_cache[list], PART_NOT_AVAILABLE, 8*5*sizeof(int8_t));
                    h->mv_cache_clean[list]= 1;
                }*/
                continue; //FIXME direct mode ...
            }
            h->mv_cache_clean[list]= 0;
            
            if(IS_INTER(topleft_type)){
                const int b_xy = h->mb2b_xy[topleft_xy] + 3 + 3*h->b_stride;
                const int b8_xy= h->mb2b8_xy[topleft_xy] + 1 + h->b8_stride;
                *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
                h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
            }else{
                *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= 0;
                h->ref_cache[list][scan8[0] - 1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
            }
            
            if(IS_INTER(top_type)){
                const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
                const int b8_xy= h->mb2b8_xy[top_xy] + h->b8_stride;
                *(uint32_t*)h->mv_cache[list][scan8[0] + 0 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 0];
                *(uint32_t*)h->mv_cache[list][scan8[0] + 1 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 1];
                *(uint32_t*)h->mv_cache[list][scan8[0] + 2 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 2];
                *(uint32_t*)h->mv_cache[list][scan8[0] + 3 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 3];
                h->ref_cache[list][scan8[0] + 0 - 1*8]=
                h->ref_cache[list][scan8[0] + 1 - 1*8]= s->current_picture.ref_index[list][b8_xy + 0];
                h->ref_cache[list][scan8[0] + 2 - 1*8]=
                h->ref_cache[list][scan8[0] + 3 - 1*8]= s->current_picture.ref_index[list][b8_xy + 1];
            }else{
                *(uint32_t*)h->mv_cache [list][scan8[0] + 0 - 1*8]= 
                *(uint32_t*)h->mv_cache [list][scan8[0] + 1 - 1*8]= 
                *(uint32_t*)h->mv_cache [list][scan8[0] + 2 - 1*8]= 
                *(uint32_t*)h->mv_cache [list][scan8[0] + 3 - 1*8]= 0;
                *(uint32_t*)&h->ref_cache[list][scan8[0] + 0 - 1*8]= ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101;
            }

            if(IS_INTER(topright_type)){
                const int b_xy= h->mb2b_xy[topright_xy] + 3*h->b_stride;
                const int b8_xy= h->mb2b8_xy[topright_xy] + h->b8_stride;
                *(uint32_t*)h->mv_cache[list][scan8[0] + 4 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
                h->ref_cache[list][scan8[0] + 4 - 1*8]= s->current_picture.ref_index[list][b8_xy];
            }else{
                *(uint32_t*)h->mv_cache [list][scan8[0] + 4 - 1*8]= 0;
                h->ref_cache[list][scan8[0] + 4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
            }
            
            //FIXME unify cleanup or sth
            if(IS_INTER(left_type[0])){
                const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
                const int b8_xy= h->mb2b8_xy[left_xy[0]] + 1;
                *(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 0*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0]];
                *(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[1]];
                h->ref_cache[list][scan8[0] - 1 + 0*8]= 
                h->ref_cache[list][scan8[0] - 1 + 1*8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[0]>>1)];
            }else{
                *(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 0*8]=
                *(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 1*8]= 0;
                h->ref_cache[list][scan8[0] - 1 + 0*8]=
                h->ref_cache[list][scan8[0] - 1 + 1*8]= left_type[0] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
            }
            
            if(IS_INTER(left_type[1])){
                const int b_xy= h->mb2b_xy[left_xy[1]] + 3;
                const int b8_xy= h->mb2b8_xy[left_xy[1]] + 1;
                *(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 2*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[2]];
                *(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 3*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[3]];
                h->ref_cache[list][scan8[0] - 1 + 2*8]= 
                h->ref_cache[list][scan8[0] - 1 + 3*8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[2]>>1)];
            }else{
                *(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 2*8]=
                *(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 3*8]= 0;
                h->ref_cache[list][scan8[0] - 1 + 2*8]=
                h->ref_cache[list][scan8[0] - 1 + 3*8]= left_type[0] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
            }

            h->ref_cache[list][scan8[5 ]+1] = 
            h->ref_cache[list][scan8[7 ]+1] = 
            h->ref_cache[list][scan8[13]+1] =  //FIXME remove past 3 (init somewher else)
            h->ref_cache[list][scan8[4 ]] = 
            h->ref_cache[list][scan8[12]] = PART_NOT_AVAILABLE;
            *(uint32_t*)h->mv_cache [list][scan8[5 ]+1]=
            *(uint32_t*)h->mv_cache [list][scan8[7 ]+1]=
            *(uint32_t*)h->mv_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewher else)
            *(uint32_t*)h->mv_cache [list][scan8[4 ]]=
            *(uint32_t*)h->mv_cache [list][scan8[12]]= 0;
656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699

            if( h->pps.cabac ) {
                /* XXX beurk, Load mvd */
                if(IS_INTER(topleft_type)){
                    const int b_xy = h->mb2b_xy[topleft_xy] + 3 + 3*h->b_stride;
                    *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy];
                }else{
                    *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 - 1*8]= 0;
                }

                if(IS_INTER(top_type)){
                    const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
                    *(uint32_t*)h->mvd_cache[list][scan8[0] + 0 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 0];
                    *(uint32_t*)h->mvd_cache[list][scan8[0] + 1 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 1];
                    *(uint32_t*)h->mvd_cache[list][scan8[0] + 2 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 2];
                    *(uint32_t*)h->mvd_cache[list][scan8[0] + 3 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 3];
                }else{
                    *(uint32_t*)h->mvd_cache [list][scan8[0] + 0 - 1*8]= 
                    *(uint32_t*)h->mvd_cache [list][scan8[0] + 1 - 1*8]= 
                    *(uint32_t*)h->mvd_cache [list][scan8[0] + 2 - 1*8]= 
                    *(uint32_t*)h->mvd_cache [list][scan8[0] + 3 - 1*8]= 0;
                }
                if(IS_INTER(left_type[0])){
                    const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
                    *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 0*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[0]];
                    *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[1]];
                }else{
                    *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 0*8]=
                    *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 1*8]= 0;
                }
                if(IS_INTER(left_type[1])){
                    const int b_xy= h->mb2b_xy[left_xy[1]] + 3;
                    *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 2*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[2]];
                    *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 3*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[3]];
                }else{
                    *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 2*8]=
                    *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 3*8]= 0;
                }
                *(uint32_t*)h->mvd_cache [list][scan8[5 ]+1]=
                *(uint32_t*)h->mvd_cache [list][scan8[7 ]+1]=
                *(uint32_t*)h->mvd_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewher else)
                *(uint32_t*)h->mvd_cache [list][scan8[4 ]]=
                *(uint32_t*)h->mvd_cache [list][scan8[12]]= 0;
            }
700 701 702 703 704 705 706 707
        }
//FIXME
    }
#endif
}

static inline void write_back_intra_pred_mode(H264Context *h){
    MpegEncContext * const s = &h->s;
708
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731

    h->intra4x4_pred_mode[mb_xy][0]= h->intra4x4_pred_mode_cache[7+8*1];
    h->intra4x4_pred_mode[mb_xy][1]= h->intra4x4_pred_mode_cache[7+8*2];
    h->intra4x4_pred_mode[mb_xy][2]= h->intra4x4_pred_mode_cache[7+8*3];
    h->intra4x4_pred_mode[mb_xy][3]= h->intra4x4_pred_mode_cache[7+8*4];
    h->intra4x4_pred_mode[mb_xy][4]= h->intra4x4_pred_mode_cache[4+8*4];
    h->intra4x4_pred_mode[mb_xy][5]= h->intra4x4_pred_mode_cache[5+8*4];
    h->intra4x4_pred_mode[mb_xy][6]= h->intra4x4_pred_mode_cache[6+8*4];
}

/**
 * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
 */
static inline int check_intra4x4_pred_mode(H264Context *h){
    MpegEncContext * const s = &h->s;
    static const int8_t top [12]= {-1, 0,LEFT_DC_PRED,-1,-1,-1,-1,-1, 0};
    static const int8_t left[12]= { 0,-1, TOP_DC_PRED, 0,-1,-1,-1, 0,-1,DC_128_PRED};
    int i;
    
    if(!(h->top_samples_available&0x8000)){
        for(i=0; i<4; i++){
            int status= top[ h->intra4x4_pred_mode_cache[scan8[0] + i] ];
            if(status<0){
732
                av_log(h->s.avctx, AV_LOG_ERROR, "top block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
733 734 735 736 737 738 739 740 741 742 743
                return -1;
            } else if(status){
                h->intra4x4_pred_mode_cache[scan8[0] + i]= status;
            }
        }
    }
    
    if(!(h->left_samples_available&0x8000)){
        for(i=0; i<4; i++){
            int status= left[ h->intra4x4_pred_mode_cache[scan8[0] + 8*i] ];
            if(status<0){
744
                av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762
                return -1;
            } else if(status){
                h->intra4x4_pred_mode_cache[scan8[0] + 8*i]= status;
            }
        }
    }

    return 0;
} //FIXME cleanup like next

/**
 * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
 */
static inline int check_intra_pred_mode(H264Context *h, int mode){
    MpegEncContext * const s = &h->s;
    static const int8_t top [7]= {LEFT_DC_PRED8x8, 1,-1,-1};
    static const int8_t left[7]= { TOP_DC_PRED8x8,-1, 2,-1,DC_128_PRED8x8};
    
Michael Niedermayer's avatar
Michael Niedermayer committed
763 764 765
    if(mode < 0 || mode > 6)
        return -1;
    
766 767 768
    if(!(h->top_samples_available&0x8000)){
        mode= top[ mode ];
        if(mode<0){
769
            av_log(h->s.avctx, AV_LOG_ERROR, "top block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
770 771 772 773 774 775 776
            return -1;
        }
    }
    
    if(!(h->left_samples_available&0x8000)){
        mode= left[ mode ];
        if(mode<0){
777
            av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793
            return -1;
        } 
    }

    return mode;
}

/**
 * gets the predicted intra4x4 prediction mode.
 */
static inline int pred_intra_mode(H264Context *h, int n){
    const int index8= scan8[n];
    const int left= h->intra4x4_pred_mode_cache[index8 - 1];
    const int top = h->intra4x4_pred_mode_cache[index8 - 8];
    const int min= FFMIN(left, top);

794
    tprintf("mode:%d %d min:%d\n", left ,top, min);
795 796 797 798 799 800 801

    if(min<0) return DC_PRED;
    else      return min;
}

static inline void write_back_non_zero_count(H264Context *h){
    MpegEncContext * const s = &h->s;
802
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
803

804 805 806 807 808 809 810 811 812 813 814 815 816 817 818
    h->non_zero_count[mb_xy][0]= h->non_zero_count_cache[4+8*4];
    h->non_zero_count[mb_xy][1]= h->non_zero_count_cache[5+8*4];
    h->non_zero_count[mb_xy][2]= h->non_zero_count_cache[6+8*4];
    h->non_zero_count[mb_xy][3]= h->non_zero_count_cache[7+8*4];
    h->non_zero_count[mb_xy][4]= h->non_zero_count_cache[7+8*3];
    h->non_zero_count[mb_xy][5]= h->non_zero_count_cache[7+8*2];
    h->non_zero_count[mb_xy][6]= h->non_zero_count_cache[7+8*1];
    
    h->non_zero_count[mb_xy][7]= h->non_zero_count_cache[1+8*2];
    h->non_zero_count[mb_xy][8]= h->non_zero_count_cache[2+8*2];
    h->non_zero_count[mb_xy][9]= h->non_zero_count_cache[2+8*1];

    h->non_zero_count[mb_xy][10]=h->non_zero_count_cache[1+8*5];
    h->non_zero_count[mb_xy][11]=h->non_zero_count_cache[2+8*5];
    h->non_zero_count[mb_xy][12]=h->non_zero_count_cache[2+8*4];
819 820 821 822 823 824 825 826 827 828 829 830 831 832
}

/**
 * gets the predicted number of non zero coefficients.
 * @param n block index
 */
static inline int pred_non_zero_count(H264Context *h, int n){
    const int index8= scan8[n];
    const int left= h->non_zero_count_cache[index8 - 1];
    const int top = h->non_zero_count_cache[index8 - 8];
    int i= left + top;
    
    if(i<64) i= (i+1)>>1;

833
    tprintf("pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);
834 835 836 837

    return i&31;
}

838 839 840 841 842 843 844
static inline int fetch_diagonal_mv(H264Context *h, const int16_t **C, int i, int list, int part_width){
    const int topright_ref= h->ref_cache[list][ i - 8 + part_width ];

    if(topright_ref != PART_NOT_AVAILABLE){
        *C= h->mv_cache[list][ i - 8 + part_width ];
        return topright_ref;
    }else{
845 846
        tprintf("topright MV not available\n");

847 848 849 850 851
        *C= h->mv_cache[list][ i - 8 - 1 ];
        return h->ref_cache[list][ i - 8 - 1 ];
    }
}

852 853 854 855 856 857 858 859 860 861 862 863 864
/**
 * gets the predicted MV.
 * @param n the block index
 * @param part_width the width of the partition (4, 8,16) -> (1, 2, 4)
 * @param mx the x component of the predicted motion vector
 * @param my the y component of the predicted motion vector
 */
static inline void pred_motion(H264Context * const h, int n, int part_width, int list, int ref, int * const mx, int * const my){
    const int index8= scan8[n];
    const int top_ref=      h->ref_cache[list][ index8 - 8 ];
    const int left_ref=     h->ref_cache[list][ index8 - 1 ];
    const int16_t * const A= h->mv_cache[list][ index8 - 1 ];
    const int16_t * const B= h->mv_cache[list][ index8 - 8 ];
865 866 867
    const int16_t * C;
    int diagonal_ref, match_count;

868
    assert(part_width==1 || part_width==2 || part_width==4);
869

870 871 872 873 874 875 876
/* mv_cache
  B . . A T T T T 
  U . . L . . , .
  U . . L . . . .
  U . . L . . , .
  . . . L . . . .
*/
877 878 879 880 881 882 883 884 885 886 887 888 889

    diagonal_ref= fetch_diagonal_mv(h, &C, index8, list, part_width);
    match_count= (diagonal_ref==ref) + (top_ref==ref) + (left_ref==ref);
    if(match_count > 1){ //most common
        *mx= mid_pred(A[0], B[0], C[0]);
        *my= mid_pred(A[1], B[1], C[1]);
    }else if(match_count==1){
        if(left_ref==ref){
            *mx= A[0];
            *my= A[1];        
        }else if(top_ref==ref){
            *mx= B[0];
            *my= B[1];        
890
        }else{
891 892
            *mx= C[0];
            *my= C[1];        
893 894
        }
    }else{
895
        if(top_ref == PART_NOT_AVAILABLE && diagonal_ref == PART_NOT_AVAILABLE && left_ref != PART_NOT_AVAILABLE){
896
            *mx= A[0];
897
            *my= A[1];        
898
        }else{
899 900
            *mx= mid_pred(A[0], B[0], C[0]);
            *my= mid_pred(A[1], B[1], C[1]);
901 902
        }
    }
903
        
904
    tprintf("pred_motion (%2d %2d %2d) (%2d %2d %2d) (%2d %2d %2d) -> (%2d %2d %2d) at %2d %2d %d list %d\n", top_ref, B[0], B[1],                    diagonal_ref, C[0], C[1], left_ref, A[0], A[1], ref, *mx, *my, h->s.mb_x, h->s.mb_y, n, list);
905 906 907 908 909 910 911 912 913 914 915 916 917
}

/**
 * gets the directionally predicted 16x8 MV.
 * @param n the block index
 * @param mx the x component of the predicted motion vector
 * @param my the y component of the predicted motion vector
 */
static inline void pred_16x8_motion(H264Context * const h, int n, int list, int ref, int * const mx, int * const my){
    if(n==0){
        const int top_ref=      h->ref_cache[list][ scan8[0] - 8 ];
        const int16_t * const B= h->mv_cache[list][ scan8[0] - 8 ];

918
        tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d", top_ref, B[0], B[1], h->s.mb_x, h->s.mb_y, n, list);
919 920 921 922 923 924 925 926 927 928
        
        if(top_ref == ref){
            *mx= B[0];
            *my= B[1];
            return;
        }
    }else{
        const int left_ref=     h->ref_cache[list][ scan8[8] - 1 ];
        const int16_t * const A= h->mv_cache[list][ scan8[8] - 1 ];
        
929
        tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952

        if(left_ref == ref){
            *mx= A[0];
            *my= A[1];
            return;
        }
    }

    //RARE
    pred_motion(h, n, 4, list, ref, mx, my);
}

/**
 * gets the directionally predicted 8x16 MV.
 * @param n the block index
 * @param mx the x component of the predicted motion vector
 * @param my the y component of the predicted motion vector
 */
static inline void pred_8x16_motion(H264Context * const h, int n, int list, int ref, int * const mx, int * const my){
    if(n==0){
        const int left_ref=      h->ref_cache[list][ scan8[0] - 1 ];
        const int16_t * const A=  h->mv_cache[list][ scan8[0] - 1 ];
        
953
        tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
954 955 956 957 958 959 960

        if(left_ref == ref){
            *mx= A[0];
            *my= A[1];
            return;
        }
    }else{
961 962 963 964
        const int16_t * C;
        int diagonal_ref;

        diagonal_ref= fetch_diagonal_mv(h, &C, scan8[4], list, 2);
965
        
966
        tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d", diagonal_ref, C[0], C[1], h->s.mb_x, h->s.mb_y, n, list);
967

968
        if(diagonal_ref == ref){ 
969 970 971 972 973 974 975 976 977 978 979 980 981 982
            *mx= C[0];
            *my= C[1];
            return;
        }
    }

    //RARE
    pred_motion(h, n, 2, list, ref, mx, my);
}

static inline void pred_pskip_motion(H264Context * const h, int * const mx, int * const my){
    const int top_ref = h->ref_cache[0][ scan8[0] - 8 ];
    const int left_ref= h->ref_cache[0][ scan8[0] - 1 ];

983
    tprintf("pred_pskip: (%d) (%d) at %2d %2d", top_ref, left_ref, h->s.mb_x, h->s.mb_y);
984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011

    if(top_ref == PART_NOT_AVAILABLE || left_ref == PART_NOT_AVAILABLE
       || (top_ref == 0  && *(uint32_t*)h->mv_cache[0][ scan8[0] - 8 ] == 0)
       || (left_ref == 0 && *(uint32_t*)h->mv_cache[0][ scan8[0] - 1 ] == 0)){
       
        *mx = *my = 0;
        return;
    }
        
    pred_motion(h, 0, 4, 0, 0, mx, my);

    return;
}

static inline void write_back_motion(H264Context *h, int mb_type){
    MpegEncContext * const s = &h->s;
    const int b_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
    const int b8_xy= 2*s->mb_x + 2*s->mb_y*h->b8_stride;
    int list;

    for(list=0; list<2; list++){
        int y;
        if((!IS_8X8(mb_type)) && !USES_LIST(mb_type, list)){
            if(1){ //FIXME skip or never read if mb_type doesnt use it
                for(y=0; y<4; y++){
                    *(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]=
                    *(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= 0;
                }
1012 1013 1014 1015 1016 1017 1018
                if( h->pps.cabac ) {
                    /* FIXME needed ? */
                    for(y=0; y<4; y++){
                        *(uint64_t*)h->mvd_table[list][b_xy + 0 + y*h->b_stride]=
                        *(uint64_t*)h->mvd_table[list][b_xy + 2 + y*h->b_stride]= 0;
                    }
                }
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
                for(y=0; y<2; y++){
                    *(uint16_t*)s->current_picture.motion_val[list][b8_xy + y*h->b8_stride]= (LIST_NOT_USED&0xFF)*0x0101;
                }
            }
            continue; //FIXME direct mode ...
        }
        
        for(y=0; y<4; y++){
            *(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+0 + 8*y];
            *(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+2 + 8*y];
        }
1030 1031 1032 1033 1034 1035
        if( h->pps.cabac ) {
            for(y=0; y<4; y++){
                *(uint64_t*)h->mvd_table[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mvd_cache[list][scan8[0]+0 + 8*y];
                *(uint64_t*)h->mvd_table[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mvd_cache[list][scan8[0]+2 + 8*y];
            }
        }
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
        for(y=0; y<2; y++){
            s->current_picture.ref_index[list][b8_xy + 0 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+0 + 16*y];
            s->current_picture.ref_index[list][b8_xy + 1 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+2 + 16*y];
        }
    }
}

/**
 * Decodes a network abstraction layer unit.
 * @param consumed is the number of bytes used as input
 * @param length is the length of the array
 * @param dst_length is the number of decoded bytes FIXME here or a decode rbsp ttailing?
 * @returns decoded bytes, might be src+1 if no escapes 
 */
static uint8_t *decode_nal(H264Context *h, uint8_t *src, int *dst_length, int *consumed, int length){
    int i, si, di;
    uint8_t *dst;

//    src[0]&0x80;		//forbidden bit
    h->nal_ref_idc= src[0]>>5;
    h->nal_unit_type= src[0]&0x1F;

    src++; length--;
#if 0    
    for(i=0; i<length; i++)
        printf("%2X ", src[i]);
#endif
    for(i=0; i+1<length; i+=2){
        if(src[i]) continue;
        if(i>0 && src[i-1]==0) i--;
        if(i+2<length && src[i+1]==0 && src[i+2]<=3){
            if(src[i+2]!=3){
                /* startcode, so we must be past the end */
                length=i;
            }
            break;
        }
    }

    if(i>=length-1){ //no escaped 0
        *dst_length= length;
        *consumed= length+1; //+1 for the header
        return src; 
    }

    h->rbsp_buffer= av_fast_realloc(h->rbsp_buffer, &h->rbsp_buffer_size, length);
    dst= h->rbsp_buffer;

//printf("deoding esc\n");
    si=di=0;
    while(si<length){ 
        //remove escapes (very rare 1:2^22)
        if(si+2<length && src[si]==0 && src[si+1]==0 && src[si+2]<=3){
            if(src[si+2]==3){ //escape
                dst[di++]= 0;
                dst[di++]= 0;
                si+=3;
1093
                continue;
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
            }else //next start code
                break;
        }

        dst[di++]= src[si++];
    }

    *dst_length= di;
    *consumed= si + 1;//+1 for the header
//FIXME store exact number of bits in the getbitcontext (its needed for decoding)
    return dst;
}

1107
#if 0
1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
/**
 * @param src the data which should be escaped
 * @param dst the target buffer, dst+1 == src is allowed as a special case
 * @param length the length of the src data
 * @param dst_length the length of the dst array
 * @returns length of escaped data in bytes or -1 if an error occured
 */
static int encode_nal(H264Context *h, uint8_t *dst, uint8_t *src, int length, int dst_length){
    int i, escape_count, si, di;
    uint8_t *temp;
    
    assert(length>=0);
    assert(dst_length>0);
    
    dst[0]= (h->nal_ref_idc<<5) + h->nal_unit_type;

    if(length==0) return 1;

    escape_count= 0;
    for(i=0; i<length; i+=2){
        if(src[i]) continue;
        if(i>0 && src[i-1]==0) 
            i--;
        if(i+2<length && src[i+1]==0 && src[i+2]<=3){
            escape_count++;
            i+=2;
        }
    }
    
    if(escape_count==0){ 
        if(dst+1 != src)
            memcpy(dst+1, src, length);
        return length + 1;
    }
    
    if(length + escape_count + 1> dst_length)
        return -1;

    //this should be damn rare (hopefully)

    h->rbsp_buffer= av_fast_realloc(h->rbsp_buffer, &h->rbsp_buffer_size, length + escape_count);
    temp= h->rbsp_buffer;
//printf("encoding esc\n");
    
    si= 0;
    di= 0;
    while(si < length){
        if(si+2<length && src[si]==0 && src[si+1]==0 && src[si+2]<=3){
            temp[di++]= 0; si++;
            temp[di++]= 0; si++;
            temp[di++]= 3; 
            temp[di++]= src[si++];
        }
        else
            temp[di++]= src[si++];
    }
    memcpy(dst+1, temp, length+escape_count);
    
    assert(di == length+escape_count);
    
    return di + 1;
}

/**
 * write 1,10,100,1000,... for alignment, yes its exactly inverse to mpeg4
 */
static void encode_rbsp_trailing(PutBitContext *pb){
    int length;
    put_bits(pb, 1, 1);
1177
    length= (-put_bits_count(pb))&7;
1178 1179
    if(length) put_bits(pb, length, 0);
}
1180
#endif
1181 1182 1183 1184 1185 1186 1187 1188 1189

/**
 * identifies the exact end of the bitstream
 * @return the length of the trailing, or 0 if damaged
 */
static int decode_rbsp_trailing(uint8_t *src){
    int v= *src;
    int r;

1190
    tprintf("rbsp trailing %X\n", v);
1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239

    for(r=1; r<9; r++){
        if(v&1) return r;
        v>>=1;
    }
    return 0;
}

/**
 * idct tranforms the 16 dc values and dequantize them.
 * @param qp quantization parameter
 */
static void h264_luma_dc_dequant_idct_c(DCTELEM *block, int qp){
    const int qmul= dequant_coeff[qp][0];
#define stride 16
    int i;
    int temp[16]; //FIXME check if this is a good idea
    static const int x_offset[4]={0, 1*stride, 4* stride,  5*stride};
    static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride};

//memset(block, 64, 2*256);
//return;
    for(i=0; i<4; i++){
        const int offset= y_offset[i];
        const int z0= block[offset+stride*0] + block[offset+stride*4];
        const int z1= block[offset+stride*0] - block[offset+stride*4];
        const int z2= block[offset+stride*1] - block[offset+stride*5];
        const int z3= block[offset+stride*1] + block[offset+stride*5];

        temp[4*i+0]= z0+z3;
        temp[4*i+1]= z1+z2;
        temp[4*i+2]= z1-z2;
        temp[4*i+3]= z0-z3;
    }

    for(i=0; i<4; i++){
        const int offset= x_offset[i];
        const int z0= temp[4*0+i] + temp[4*2+i];
        const int z1= temp[4*0+i] - temp[4*2+i];
        const int z2= temp[4*1+i] - temp[4*3+i];
        const int z3= temp[4*1+i] + temp[4*3+i];

        block[stride*0 +offset]= ((z0 + z3)*qmul + 2)>>2; //FIXME think about merging this into decode_resdual
        block[stride*2 +offset]= ((z1 + z2)*qmul + 2)>>2;
        block[stride*8 +offset]= ((z1 - z2)*qmul + 2)>>2;
        block[stride*10+offset]= ((z0 - z3)*qmul + 2)>>2;
    }
}

1240
#if 0
1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277
/**
 * dct tranforms the 16 dc values.
 * @param qp quantization parameter ??? FIXME
 */
static void h264_luma_dc_dct_c(DCTELEM *block/*, int qp*/){
//    const int qmul= dequant_coeff[qp][0];
    int i;
    int temp[16]; //FIXME check if this is a good idea
    static const int x_offset[4]={0, 1*stride, 4* stride,  5*stride};
    static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride};

    for(i=0; i<4; i++){
        const int offset= y_offset[i];
        const int z0= block[offset+stride*0] + block[offset+stride*4];
        const int z1= block[offset+stride*0] - block[offset+stride*4];
        const int z2= block[offset+stride*1] - block[offset+stride*5];
        const int z3= block[offset+stride*1] + block[offset+stride*5];

        temp[4*i+0]= z0+z3;
        temp[4*i+1]= z1+z2;
        temp[4*i+2]= z1-z2;
        temp[4*i+3]= z0-z3;
    }

    for(i=0; i<4; i++){
        const int offset= x_offset[i];
        const int z0= temp[4*0+i] + temp[4*2+i];
        const int z1= temp[4*0+i] - temp[4*2+i];
        const int z2= temp[4*1+i] - temp[4*3+i];
        const int z3= temp[4*1+i] + temp[4*3+i];

        block[stride*0 +offset]= (z0 + z3)>>1;
        block[stride*2 +offset]= (z1 + z2)>>1;
        block[stride*8 +offset]= (z1 - z2)>>1;
        block[stride*10+offset]= (z0 - z3)>>1;
    }
}
1278 1279
#endif

1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304
#undef xStride
#undef stride

static void chroma_dc_dequant_idct_c(DCTELEM *block, int qp){
    const int qmul= dequant_coeff[qp][0];
    const int stride= 16*2;
    const int xStride= 16;
    int a,b,c,d,e;

    a= block[stride*0 + xStride*0];
    b= block[stride*0 + xStride*1];
    c= block[stride*1 + xStride*0];
    d= block[stride*1 + xStride*1];

    e= a-b;
    a= a+b;
    b= c-d;
    c= c+d;

    block[stride*0 + xStride*0]= ((a+c)*qmul + 0)>>1;
    block[stride*0 + xStride*1]= ((e+b)*qmul + 0)>>1;
    block[stride*1 + xStride*0]= ((a-c)*qmul + 0)>>1;
    block[stride*1 + xStride*1]= ((e-b)*qmul + 0)>>1;
}

1305
#if 0
1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325
static void chroma_dc_dct_c(DCTELEM *block){
    const int stride= 16*2;
    const int xStride= 16;
    int a,b,c,d,e;

    a= block[stride*0 + xStride*0];
    b= block[stride*0 + xStride*1];
    c= block[stride*1 + xStride*0];
    d= block[stride*1 + xStride*1];

    e= a-b;
    a= a+b;
    b= c-d;
    c= c+d;

    block[stride*0 + xStride*0]= (a+c);
    block[stride*0 + xStride*1]= (e+b);
    block[stride*1 + xStride*0]= (a-c);
    block[stride*1 + xStride*1]= (e-b);
}
1326
#endif
1327 1328 1329 1330 1331 1332 1333 1334 1335 1336

/**
 * gets the chroma qp.
 */
static inline int get_chroma_qp(H264Context *h, int qscale){
    
    return chroma_qp[clip(qscale + h->pps.chroma_qp_index_offset, 0, 51)];
}


1337
#if 0
1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369
static void h264_diff_dct_c(DCTELEM *block, uint8_t *src1, uint8_t *src2, int stride){
    int i;
    //FIXME try int temp instead of block
    
    for(i=0; i<4; i++){
        const int d0= src1[0 + i*stride] - src2[0 + i*stride];
        const int d1= src1[1 + i*stride] - src2[1 + i*stride];
        const int d2= src1[2 + i*stride] - src2[2 + i*stride];
        const int d3= src1[3 + i*stride] - src2[3 + i*stride];
        const int z0= d0 + d3;
        const int z3= d0 - d3;
        const int z1= d1 + d2;
        const int z2= d1 - d2;
        
        block[0 + 4*i]=   z0 +   z1;
        block[1 + 4*i]= 2*z3 +   z2;
        block[2 + 4*i]=   z0 -   z1;
        block[3 + 4*i]=   z3 - 2*z2;
    }    

    for(i=0; i<4; i++){
        const int z0= block[0*4 + i] + block[3*4 + i];
        const int z3= block[0*4 + i] - block[3*4 + i];
        const int z1= block[1*4 + i] + block[2*4 + i];
        const int z2= block[1*4 + i] - block[2*4 + i];
        
        block[0*4 + i]=   z0 +   z1;
        block[1*4 + i]= 2*z3 +   z2;
        block[2*4 + i]=   z0 -   z1;
        block[3*4 + i]=   z3 - 2*z2;
    }
}
1370
#endif
1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540

//FIXME need to check that this doesnt overflow signed 32 bit for low qp, iam not sure, its very close
//FIXME check that gcc inlines this (and optimizes intra & seperate_dc stuff away)
static inline int quantize_c(DCTELEM *block, uint8_t *scantable, int qscale, int intra, int seperate_dc){
    int i;
    const int * const quant_table= quant_coeff[qscale];
    const int bias= intra ? (1<<QUANT_SHIFT)/3 : (1<<QUANT_SHIFT)/6;
    const unsigned int threshold1= (1<<QUANT_SHIFT) - bias - 1;
    const unsigned int threshold2= (threshold1<<1);
    int last_non_zero;

    if(seperate_dc){
        if(qscale<=18){
            //avoid overflows
            const int dc_bias= intra ? (1<<(QUANT_SHIFT-2))/3 : (1<<(QUANT_SHIFT-2))/6;
            const unsigned int dc_threshold1= (1<<(QUANT_SHIFT-2)) - dc_bias - 1;
            const unsigned int dc_threshold2= (dc_threshold1<<1);

            int level= block[0]*quant_coeff[qscale+18][0];
            if(((unsigned)(level+dc_threshold1))>dc_threshold2){
                if(level>0){
                    level= (dc_bias + level)>>(QUANT_SHIFT-2);
                    block[0]= level;
                }else{
                    level= (dc_bias - level)>>(QUANT_SHIFT-2);
                    block[0]= -level;
                }
//                last_non_zero = i;
            }else{
                block[0]=0;
            }
        }else{
            const int dc_bias= intra ? (1<<(QUANT_SHIFT+1))/3 : (1<<(QUANT_SHIFT+1))/6;
            const unsigned int dc_threshold1= (1<<(QUANT_SHIFT+1)) - dc_bias - 1;
            const unsigned int dc_threshold2= (dc_threshold1<<1);

            int level= block[0]*quant_table[0];
            if(((unsigned)(level+dc_threshold1))>dc_threshold2){
                if(level>0){
                    level= (dc_bias + level)>>(QUANT_SHIFT+1);
                    block[0]= level;
                }else{
                    level= (dc_bias - level)>>(QUANT_SHIFT+1);
                    block[0]= -level;
                }
//                last_non_zero = i;
            }else{
                block[0]=0;
            }
        }
        last_non_zero= 0;
        i=1;
    }else{
        last_non_zero= -1;
        i=0;
    }

    for(; i<16; i++){
        const int j= scantable[i];
        int level= block[j]*quant_table[j];

//        if(   bias+level >= (1<<(QMAT_SHIFT - 3))
//           || bias-level >= (1<<(QMAT_SHIFT - 3))){
        if(((unsigned)(level+threshold1))>threshold2){
            if(level>0){
                level= (bias + level)>>QUANT_SHIFT;
                block[j]= level;
            }else{
                level= (bias - level)>>QUANT_SHIFT;
                block[j]= -level;
            }
            last_non_zero = i;
        }else{
            block[j]=0;
        }
    }

    return last_non_zero;
}

static void pred4x4_vertical_c(uint8_t *src, uint8_t *topright, int stride){
    const uint32_t a= ((uint32_t*)(src-stride))[0];
    ((uint32_t*)(src+0*stride))[0]= a;
    ((uint32_t*)(src+1*stride))[0]= a;
    ((uint32_t*)(src+2*stride))[0]= a;
    ((uint32_t*)(src+3*stride))[0]= a;
}

static void pred4x4_horizontal_c(uint8_t *src, uint8_t *topright, int stride){
    ((uint32_t*)(src+0*stride))[0]= src[-1+0*stride]*0x01010101;
    ((uint32_t*)(src+1*stride))[0]= src[-1+1*stride]*0x01010101;
    ((uint32_t*)(src+2*stride))[0]= src[-1+2*stride]*0x01010101;
    ((uint32_t*)(src+3*stride))[0]= src[-1+3*stride]*0x01010101;
}

static void pred4x4_dc_c(uint8_t *src, uint8_t *topright, int stride){
    const int dc= (  src[-stride] + src[1-stride] + src[2-stride] + src[3-stride]
                   + src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
    
    ((uint32_t*)(src+0*stride))[0]= 
    ((uint32_t*)(src+1*stride))[0]= 
    ((uint32_t*)(src+2*stride))[0]= 
    ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101; 
}

static void pred4x4_left_dc_c(uint8_t *src, uint8_t *topright, int stride){
    const int dc= (  src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 2) >>2;
    
    ((uint32_t*)(src+0*stride))[0]= 
    ((uint32_t*)(src+1*stride))[0]= 
    ((uint32_t*)(src+2*stride))[0]= 
    ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101; 
}

static void pred4x4_top_dc_c(uint8_t *src, uint8_t *topright, int stride){
    const int dc= (  src[-stride] + src[1-stride] + src[2-stride] + src[3-stride] + 2) >>2;
    
    ((uint32_t*)(src+0*stride))[0]= 
    ((uint32_t*)(src+1*stride))[0]= 
    ((uint32_t*)(src+2*stride))[0]= 
    ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101; 
}

static void pred4x4_128_dc_c(uint8_t *src, uint8_t *topright, int stride){
    ((uint32_t*)(src+0*stride))[0]= 
    ((uint32_t*)(src+1*stride))[0]= 
    ((uint32_t*)(src+2*stride))[0]= 
    ((uint32_t*)(src+3*stride))[0]= 128U*0x01010101U;
}


#define LOAD_TOP_RIGHT_EDGE\
    const int t4= topright[0];\
    const int t5= topright[1];\
    const int t6= topright[2];\
    const int t7= topright[3];\

#define LOAD_LEFT_EDGE\
    const int l0= src[-1+0*stride];\
    const int l1= src[-1+1*stride];\
    const int l2= src[-1+2*stride];\
    const int l3= src[-1+3*stride];\

#define LOAD_TOP_EDGE\
    const int t0= src[ 0-1*stride];\
    const int t1= src[ 1-1*stride];\
    const int t2= src[ 2-1*stride];\
    const int t3= src[ 3-1*stride];\

static void pred4x4_down_right_c(uint8_t *src, uint8_t *topright, int stride){
    const int lt= src[-1-1*stride];
    LOAD_TOP_EDGE
    LOAD_LEFT_EDGE

    src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2; 
    src[0+2*stride]=
    src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2; 
    src[0+1*stride]=
    src[1+2*stride]=
    src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2; 
    src[0+0*stride]=
    src[1+1*stride]=
    src[2+2*stride]=
    src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2; 
    src[1+0*stride]=
    src[2+1*stride]=
    src[3+2*stride]=(lt + 2*t0 + t1 + 2)>>2;
    src[2+0*stride]=
    src[3+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
    src[3+0*stride]=(t1 + 2*t2 + t3 + 2)>>2;
1541
}
1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563

static void pred4x4_down_left_c(uint8_t *src, uint8_t *topright, int stride){
    LOAD_TOP_EDGE    
    LOAD_TOP_RIGHT_EDGE    
//    LOAD_LEFT_EDGE    

    src[0+0*stride]=(t0 + t2 + 2*t1 + 2)>>2;
    src[1+0*stride]=
    src[0+1*stride]=(t1 + t3 + 2*t2 + 2)>>2;
    src[2+0*stride]=
    src[1+1*stride]=
    src[0+2*stride]=(t2 + t4 + 2*t3 + 2)>>2;
    src[3+0*stride]=
    src[2+1*stride]=
    src[1+2*stride]=
    src[0+3*stride]=(t3 + t5 + 2*t4 + 2)>>2;
    src[3+1*stride]=
    src[2+2*stride]=
    src[1+3*stride]=(t4 + t6 + 2*t5 + 2)>>2;
    src[3+2*stride]=
    src[2+3*stride]=(t5 + t7 + 2*t6 + 2)>>2;
    src[3+3*stride]=(t6 + 3*t7 + 2)>>2;
1564
}
1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587

static void pred4x4_vertical_right_c(uint8_t *src, uint8_t *topright, int stride){
    const int lt= src[-1-1*stride];
    LOAD_TOP_EDGE    
    LOAD_LEFT_EDGE    
    const __attribute__((unused)) int unu= l3;

    src[0+0*stride]=
    src[1+2*stride]=(lt + t0 + 1)>>1;
    src[1+0*stride]=
    src[2+2*stride]=(t0 + t1 + 1)>>1;
    src[2+0*stride]=
    src[3+2*stride]=(t1 + t2 + 1)>>1;
    src[3+0*stride]=(t2 + t3 + 1)>>1;
    src[0+1*stride]=
    src[1+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
    src[1+1*stride]=
    src[2+3*stride]=(lt + 2*t0 + t1 + 2)>>2;
    src[2+1*stride]=
    src[3+3*stride]=(t0 + 2*t1 + t2 + 2)>>2;
    src[3+1*stride]=(t1 + 2*t2 + t3 + 2)>>2;
    src[0+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
    src[0+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
1588
}
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610

static void pred4x4_vertical_left_c(uint8_t *src, uint8_t *topright, int stride){
    LOAD_TOP_EDGE    
    LOAD_TOP_RIGHT_EDGE    
    const __attribute__((unused)) int unu= t7;

    src[0+0*stride]=(t0 + t1 + 1)>>1;
    src[1+0*stride]=
    src[0+2*stride]=(t1 + t2 + 1)>>1;
    src[2+0*stride]=
    src[1+2*stride]=(t2 + t3 + 1)>>1;
    src[3+0*stride]=
    src[2+2*stride]=(t3 + t4+ 1)>>1;
    src[3+2*stride]=(t4 + t5+ 1)>>1;
    src[0+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
    src[1+1*stride]=
    src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
    src[2+1*stride]=
    src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
    src[3+1*stride]=
    src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
    src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
1611
}
1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631

static void pred4x4_horizontal_up_c(uint8_t *src, uint8_t *topright, int stride){
    LOAD_LEFT_EDGE    

    src[0+0*stride]=(l0 + l1 + 1)>>1;
    src[1+0*stride]=(l0 + 2*l1 + l2 + 2)>>2;
    src[2+0*stride]=
    src[0+1*stride]=(l1 + l2 + 1)>>1;
    src[3+0*stride]=
    src[1+1*stride]=(l1 + 2*l2 + l3 + 2)>>2;
    src[2+1*stride]=
    src[0+2*stride]=(l2 + l3 + 1)>>1;
    src[3+1*stride]=
    src[1+2*stride]=(l2 + 2*l3 + l3 + 2)>>2;
    src[3+2*stride]=
    src[1+3*stride]=
    src[0+3*stride]=
    src[2+2*stride]=
    src[2+3*stride]=
    src[3+3*stride]=l3;
1632
}
1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655
    
static void pred4x4_horizontal_down_c(uint8_t *src, uint8_t *topright, int stride){
    const int lt= src[-1-1*stride];
    LOAD_TOP_EDGE    
    LOAD_LEFT_EDGE    
    const __attribute__((unused)) int unu= t3;

    src[0+0*stride]=
    src[2+1*stride]=(lt + l0 + 1)>>1;
    src[1+0*stride]=
    src[3+1*stride]=(l0 + 2*lt + t0 + 2)>>2;
    src[2+0*stride]=(lt + 2*t0 + t1 + 2)>>2;
    src[3+0*stride]=(t0 + 2*t1 + t2 + 2)>>2;
    src[0+1*stride]=
    src[2+2*stride]=(l0 + l1 + 1)>>1;
    src[1+1*stride]=
    src[3+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
    src[0+2*stride]=
    src[2+3*stride]=(l1 + l2+ 1)>>1;
    src[1+2*stride]=
    src[3+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
    src[0+3*stride]=(l2 + l3 + 1)>>1;
    src[1+3*stride]=(l1 + 2*l2 + l3 + 2)>>2;
1656
}
1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748

static void pred16x16_vertical_c(uint8_t *src, int stride){
    int i;
    const uint32_t a= ((uint32_t*)(src-stride))[0];
    const uint32_t b= ((uint32_t*)(src-stride))[1];
    const uint32_t c= ((uint32_t*)(src-stride))[2];
    const uint32_t d= ((uint32_t*)(src-stride))[3];
    
    for(i=0; i<16; i++){
        ((uint32_t*)(src+i*stride))[0]= a;
        ((uint32_t*)(src+i*stride))[1]= b;
        ((uint32_t*)(src+i*stride))[2]= c;
        ((uint32_t*)(src+i*stride))[3]= d;
    }
}

static void pred16x16_horizontal_c(uint8_t *src, int stride){
    int i;

    for(i=0; i<16; i++){
        ((uint32_t*)(src+i*stride))[0]=
        ((uint32_t*)(src+i*stride))[1]=
        ((uint32_t*)(src+i*stride))[2]=
        ((uint32_t*)(src+i*stride))[3]= src[-1+i*stride]*0x01010101;
    }
}

static void pred16x16_dc_c(uint8_t *src, int stride){
    int i, dc=0;

    for(i=0;i<16; i++){
        dc+= src[-1+i*stride];
    }
    
    for(i=0;i<16; i++){
        dc+= src[i-stride];
    }

    dc= 0x01010101*((dc + 16)>>5);

    for(i=0; i<16; i++){
        ((uint32_t*)(src+i*stride))[0]=
        ((uint32_t*)(src+i*stride))[1]=
        ((uint32_t*)(src+i*stride))[2]=
        ((uint32_t*)(src+i*stride))[3]= dc;
    }
}

static void pred16x16_left_dc_c(uint8_t *src, int stride){
    int i, dc=0;

    for(i=0;i<16; i++){
        dc+= src[-1+i*stride];
    }
    
    dc= 0x01010101*((dc + 8)>>4);

    for(i=0; i<16; i++){
        ((uint32_t*)(src+i*stride))[0]=
        ((uint32_t*)(src+i*stride))[1]=
        ((uint32_t*)(src+i*stride))[2]=
        ((uint32_t*)(src+i*stride))[3]= dc;
    }
}

static void pred16x16_top_dc_c(uint8_t *src, int stride){
    int i, dc=0;

    for(i=0;i<16; i++){
        dc+= src[i-stride];
    }
    dc= 0x01010101*((dc + 8)>>4);

    for(i=0; i<16; i++){
        ((uint32_t*)(src+i*stride))[0]=
        ((uint32_t*)(src+i*stride))[1]=
        ((uint32_t*)(src+i*stride))[2]=
        ((uint32_t*)(src+i*stride))[3]= dc;
    }
}

static void pred16x16_128_dc_c(uint8_t *src, int stride){
    int i;

    for(i=0; i<16; i++){
        ((uint32_t*)(src+i*stride))[0]=
        ((uint32_t*)(src+i*stride))[1]=
        ((uint32_t*)(src+i*stride))[2]=
        ((uint32_t*)(src+i*stride))[3]= 0x01010101U*128U;
    }
}

1749
static inline void pred16x16_plane_compat_c(uint8_t *src, int stride, const int svq3){
1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762
  int i, j, k;
  int a;
  uint8_t *cm = cropTbl + MAX_NEG_CROP;
  const uint8_t * const src0 = src+7-stride;
  const uint8_t *src1 = src+8*stride-1;
  const uint8_t *src2 = src1-2*stride;      // == src+6*stride-1;
  int H = src0[1] - src0[-1];
  int V = src1[0] - src2[ 0];
  for(k=2; k<=8; ++k) {
    src1 += stride; src2 -= stride;
    H += k*(src0[k] - src0[-k]);
    V += k*(src1[0] - src2[ 0]);
  }
1763 1764 1765
  if(svq3){
    H = ( 5*(H/4) ) / 16;
    V = ( 5*(V/4) ) / 16;
1766 1767 1768

    /* required for 100% accuracy */
    i = H; H = V; V = i;
1769 1770 1771 1772
  }else{
    H = ( 5*H+32 ) >> 6;
    V = ( 5*V+32 ) >> 6;
  }
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783

  a = 16*(src1[0] + src2[16] + 1) - 7*(V+H);
  for(j=16; j>0; --j) {
    int b = a;
    a += V;
    for(i=-16; i<0; i+=4) {
      src[16+i] = cm[ (b    ) >> 5 ];
      src[17+i] = cm[ (b+  H) >> 5 ];
      src[18+i] = cm[ (b+2*H) >> 5 ];
      src[19+i] = cm[ (b+3*H) >> 5 ];
      b += 4*H;
1784
    }
1785 1786
    src += stride;
  }
1787 1788
}

1789 1790 1791 1792
static void pred16x16_plane_c(uint8_t *src, int stride){
    pred16x16_plane_compat_c(src, stride, 0);
}

1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896
static void pred8x8_vertical_c(uint8_t *src, int stride){
    int i;
    const uint32_t a= ((uint32_t*)(src-stride))[0];
    const uint32_t b= ((uint32_t*)(src-stride))[1];
    
    for(i=0; i<8; i++){
        ((uint32_t*)(src+i*stride))[0]= a;
        ((uint32_t*)(src+i*stride))[1]= b;
    }
}

static void pred8x8_horizontal_c(uint8_t *src, int stride){
    int i;

    for(i=0; i<8; i++){
        ((uint32_t*)(src+i*stride))[0]=
        ((uint32_t*)(src+i*stride))[1]= src[-1+i*stride]*0x01010101;
    }
}

static void pred8x8_128_dc_c(uint8_t *src, int stride){
    int i;

    for(i=0; i<4; i++){
        ((uint32_t*)(src+i*stride))[0]= 
        ((uint32_t*)(src+i*stride))[1]= 0x01010101U*128U;
    }
    for(i=4; i<8; i++){
        ((uint32_t*)(src+i*stride))[0]= 
        ((uint32_t*)(src+i*stride))[1]= 0x01010101U*128U;
    }
}

static void pred8x8_left_dc_c(uint8_t *src, int stride){
    int i;
    int dc0, dc2;

    dc0=dc2=0;
    for(i=0;i<4; i++){
        dc0+= src[-1+i*stride];
        dc2+= src[-1+(i+4)*stride];
    }
    dc0= 0x01010101*((dc0 + 2)>>2);
    dc2= 0x01010101*((dc2 + 2)>>2);

    for(i=0; i<4; i++){
        ((uint32_t*)(src+i*stride))[0]=
        ((uint32_t*)(src+i*stride))[1]= dc0;
    }
    for(i=4; i<8; i++){
        ((uint32_t*)(src+i*stride))[0]=
        ((uint32_t*)(src+i*stride))[1]= dc2;
    }
}

static void pred8x8_top_dc_c(uint8_t *src, int stride){
    int i;
    int dc0, dc1;

    dc0=dc1=0;
    for(i=0;i<4; i++){
        dc0+= src[i-stride];
        dc1+= src[4+i-stride];
    }
    dc0= 0x01010101*((dc0 + 2)>>2);
    dc1= 0x01010101*((dc1 + 2)>>2);

    for(i=0; i<4; i++){
        ((uint32_t*)(src+i*stride))[0]= dc0;
        ((uint32_t*)(src+i*stride))[1]= dc1;
    }
    for(i=4; i<8; i++){
        ((uint32_t*)(src+i*stride))[0]= dc0;
        ((uint32_t*)(src+i*stride))[1]= dc1;
    }
}


static void pred8x8_dc_c(uint8_t *src, int stride){
    int i;
    int dc0, dc1, dc2, dc3;

    dc0=dc1=dc2=0;
    for(i=0;i<4; i++){
        dc0+= src[-1+i*stride] + src[i-stride];
        dc1+= src[4+i-stride];
        dc2+= src[-1+(i+4)*stride];
    }
    dc3= 0x01010101*((dc1 + dc2 + 4)>>3);
    dc0= 0x01010101*((dc0 + 4)>>3);
    dc1= 0x01010101*((dc1 + 2)>>2);
    dc2= 0x01010101*((dc2 + 2)>>2);

    for(i=0; i<4; i++){
        ((uint32_t*)(src+i*stride))[0]= dc0;
        ((uint32_t*)(src+i*stride))[1]= dc1;
    }
    for(i=4; i<8; i++){
        ((uint32_t*)(src+i*stride))[0]= dc2;
        ((uint32_t*)(src+i*stride))[1]= dc3;
    }
}

static void pred8x8_plane_c(uint8_t *src, int stride){
1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
  int j, k;
  int a;
  uint8_t *cm = cropTbl + MAX_NEG_CROP;
  const uint8_t * const src0 = src+3-stride;
  const uint8_t *src1 = src+4*stride-1;
  const uint8_t *src2 = src1-2*stride;      // == src+2*stride-1;
  int H = src0[1] - src0[-1];
  int V = src1[0] - src2[ 0];
  for(k=2; k<=4; ++k) {
    src1 += stride; src2 -= stride;
    H += k*(src0[k] - src0[-k]);
    V += k*(src1[0] - src2[ 0]);
  }
  H = ( 17*H+16 ) >> 5;
  V = ( 17*V+16 ) >> 5;

  a = 16*(src1[0] + src2[8]+1) - 3*(V+H);
  for(j=8; j>0; --j) {
    int b = a;
    a += V;
    src[0] = cm[ (b    ) >> 5 ];
    src[1] = cm[ (b+  H) >> 5 ];
    src[2] = cm[ (b+2*H) >> 5 ];
    src[3] = cm[ (b+3*H) >> 5 ];
    src[4] = cm[ (b+4*H) >> 5 ];
    src[5] = cm[ (b+5*H) >> 5 ];
    src[6] = cm[ (b+6*H) >> 5 ];
    src[7] = cm[ (b+7*H) >> 5 ];
    src += stride;
  }
1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954
}

static inline void mc_dir_part(H264Context *h, Picture *pic, int n, int square, int chroma_height, int delta, int list,
                           uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                           int src_x_offset, int src_y_offset,
                           qpel_mc_func *qpix_op, h264_chroma_mc_func chroma_op){
    MpegEncContext * const s = &h->s;
    const int mx= h->mv_cache[list][ scan8[n] ][0] + src_x_offset*8;
    const int my= h->mv_cache[list][ scan8[n] ][1] + src_y_offset*8;
    const int luma_xy= (mx&3) + ((my&3)<<2);
    uint8_t * src_y = pic->data[0] + (mx>>2) + (my>>2)*s->linesize;
    uint8_t * src_cb= pic->data[1] + (mx>>3) + (my>>3)*s->uvlinesize;
    uint8_t * src_cr= pic->data[2] + (mx>>3) + (my>>3)*s->uvlinesize;
    int extra_width= (s->flags&CODEC_FLAG_EMU_EDGE) ? 0 : 16; //FIXME increase edge?, IMHO not worth it
    int extra_height= extra_width;
    int emu=0;
    const int full_mx= mx>>2;
    const int full_my= my>>2;
    
    assert(pic->data[0]);
    
    if(mx&7) extra_width -= 3;
    if(my&7) extra_height -= 3;
    
    if(   full_mx < 0-extra_width 
       || full_my < 0-extra_height 
       || full_mx + 16/*FIXME*/ > s->width + extra_width 
       || full_my + 16/*FIXME*/ > s->height + extra_height){
1955
        ff_emulated_edge_mc(s->edge_emu_buffer, src_y - 2 - 2*s->linesize, s->linesize, 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, s->width, s->height);
1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967
            src_y= s->edge_emu_buffer + 2 + 2*s->linesize;
        emu=1;
    }
    
    qpix_op[luma_xy](dest_y, src_y, s->linesize); //FIXME try variable height perhaps?
    if(!square){
        qpix_op[luma_xy](dest_y + delta, src_y + delta, s->linesize);
    }
    
    if(s->flags&CODEC_FLAG_GRAY) return;
    
    if(emu){
1968
        ff_emulated_edge_mc(s->edge_emu_buffer, src_cb, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), s->width>>1, s->height>>1);
1969 1970 1971 1972 1973
            src_cb= s->edge_emu_buffer;
    }
    chroma_op(dest_cb, src_cb, s->uvlinesize, chroma_height, mx&7, my&7);

    if(emu){
1974
        ff_emulated_edge_mc(s->edge_emu_buffer, src_cr, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), s->width>>1, s->height>>1);
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
            src_cr= s->edge_emu_buffer;
    }
    chroma_op(dest_cr, src_cr, s->uvlinesize, chroma_height, mx&7, my&7);
}

static inline void mc_part(H264Context *h, int n, int square, int chroma_height, int delta,
                           uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                           int x_offset, int y_offset,
                           qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
                           qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg,
                           int list0, int list1){
    MpegEncContext * const s = &h->s;
    qpel_mc_func *qpix_op=  qpix_put;
    h264_chroma_mc_func chroma_op= chroma_put;
    
    dest_y  += 2*x_offset + 2*y_offset*s->  linesize;
    dest_cb +=   x_offset +   y_offset*s->uvlinesize;
    dest_cr +=   x_offset +   y_offset*s->uvlinesize;
    x_offset += 8*s->mb_x;
    y_offset += 8*s->mb_y;
    
    if(list0){
1997
        Picture *ref= &h->ref_list[0][ h->ref_cache[0][ scan8[n] ] ];
1998 1999 2000 2001 2002 2003 2004 2005 2006
        mc_dir_part(h, ref, n, square, chroma_height, delta, 0,
                           dest_y, dest_cb, dest_cr, x_offset, y_offset,
                           qpix_op, chroma_op);

        qpix_op=  qpix_avg;
        chroma_op= chroma_avg;
    }

    if(list1){
2007
        Picture *ref= &h->ref_list[1][ h->ref_cache[1][ scan8[n] ] ];
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
        mc_dir_part(h, ref, n, square, chroma_height, delta, 1,
                           dest_y, dest_cb, dest_cr, x_offset, y_offset,
                           qpix_op, chroma_op);
    }
}

static void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                      qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
                      qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg)){
    MpegEncContext * const s = &h->s;
2018
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161
    const int mb_type= s->current_picture.mb_type[mb_xy];
    
    assert(IS_INTER(mb_type));
    
    if(IS_16X16(mb_type)){
        mc_part(h, 0, 1, 8, 0, dest_y, dest_cb, dest_cr, 0, 0,
                qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0],
                IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
    }else if(IS_16X8(mb_type)){
        mc_part(h, 0, 0, 4, 8, dest_y, dest_cb, dest_cr, 0, 0,
                qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0],
                IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
        mc_part(h, 8, 0, 4, 8, dest_y, dest_cb, dest_cr, 0, 4,
                qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0],
                IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1));
    }else if(IS_8X16(mb_type)){
        mc_part(h, 0, 0, 8, 8*s->linesize, dest_y, dest_cb, dest_cr, 0, 0,
                qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
                IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
        mc_part(h, 4, 0, 8, 8*s->linesize, dest_y, dest_cb, dest_cr, 4, 0,
                qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
                IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1));
    }else{
        int i;
        
        assert(IS_8X8(mb_type));

        for(i=0; i<4; i++){
            const int sub_mb_type= h->sub_mb_type[i];
            const int n= 4*i;
            int x_offset= (i&1)<<2;
            int y_offset= (i&2)<<1;

            if(IS_SUB_8X8(sub_mb_type)){
                mc_part(h, n, 1, 4, 0, dest_y, dest_cb, dest_cr, x_offset, y_offset,
                    qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
                    IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
            }else if(IS_SUB_8X4(sub_mb_type)){
                mc_part(h, n  , 0, 2, 4, dest_y, dest_cb, dest_cr, x_offset, y_offset,
                    qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1],
                    IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
                mc_part(h, n+2, 0, 2, 4, dest_y, dest_cb, dest_cr, x_offset, y_offset+2,
                    qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1],
                    IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
            }else if(IS_SUB_4X8(sub_mb_type)){
                mc_part(h, n  , 0, 4, 4*s->linesize, dest_y, dest_cb, dest_cr, x_offset, y_offset,
                    qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
                    IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
                mc_part(h, n+1, 0, 4, 4*s->linesize, dest_y, dest_cb, dest_cr, x_offset+2, y_offset,
                    qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
                    IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
            }else{
                int j;
                assert(IS_SUB_4X4(sub_mb_type));
                for(j=0; j<4; j++){
                    int sub_x_offset= x_offset + 2*(j&1);
                    int sub_y_offset= y_offset +   (j&2);
                    mc_part(h, n+j, 1, 2, 0, dest_y, dest_cb, dest_cr, sub_x_offset, sub_y_offset,
                        qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
                        IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
                }
            }
        }
    }
}

static void decode_init_vlc(H264Context *h){
    static int done = 0;

    if (!done) {
        int i;
        done = 1;

        init_vlc(&chroma_dc_coeff_token_vlc, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 4*5, 
                 &chroma_dc_coeff_token_len [0], 1, 1,
                 &chroma_dc_coeff_token_bits[0], 1, 1);

        for(i=0; i<4; i++){
            init_vlc(&coeff_token_vlc[i], COEFF_TOKEN_VLC_BITS, 4*17, 
                     &coeff_token_len [i][0], 1, 1,
                     &coeff_token_bits[i][0], 1, 1);
        }

        for(i=0; i<3; i++){
            init_vlc(&chroma_dc_total_zeros_vlc[i], CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 4,
                     &chroma_dc_total_zeros_len [i][0], 1, 1,
                     &chroma_dc_total_zeros_bits[i][0], 1, 1);
        }
        for(i=0; i<15; i++){
            init_vlc(&total_zeros_vlc[i], TOTAL_ZEROS_VLC_BITS, 16, 
                     &total_zeros_len [i][0], 1, 1,
                     &total_zeros_bits[i][0], 1, 1);
        }

        for(i=0; i<6; i++){
            init_vlc(&run_vlc[i], RUN_VLC_BITS, 7, 
                     &run_len [i][0], 1, 1,
                     &run_bits[i][0], 1, 1);
        }
        init_vlc(&run7_vlc, RUN7_VLC_BITS, 16, 
                 &run_len [6][0], 1, 1,
                 &run_bits[6][0], 1, 1);
    }
}

/**
 * Sets the intra prediction function pointers.
 */
static void init_pred_ptrs(H264Context *h){
//    MpegEncContext * const s = &h->s;

    h->pred4x4[VERT_PRED           ]= pred4x4_vertical_c;
    h->pred4x4[HOR_PRED            ]= pred4x4_horizontal_c;
    h->pred4x4[DC_PRED             ]= pred4x4_dc_c;
    h->pred4x4[DIAG_DOWN_LEFT_PRED ]= pred4x4_down_left_c;
    h->pred4x4[DIAG_DOWN_RIGHT_PRED]= pred4x4_down_right_c;
    h->pred4x4[VERT_RIGHT_PRED     ]= pred4x4_vertical_right_c;
    h->pred4x4[HOR_DOWN_PRED       ]= pred4x4_horizontal_down_c;
    h->pred4x4[VERT_LEFT_PRED      ]= pred4x4_vertical_left_c;
    h->pred4x4[HOR_UP_PRED         ]= pred4x4_horizontal_up_c;
    h->pred4x4[LEFT_DC_PRED        ]= pred4x4_left_dc_c;
    h->pred4x4[TOP_DC_PRED         ]= pred4x4_top_dc_c;
    h->pred4x4[DC_128_PRED         ]= pred4x4_128_dc_c;

    h->pred8x8[DC_PRED8x8     ]= pred8x8_dc_c;
    h->pred8x8[VERT_PRED8x8   ]= pred8x8_vertical_c;
    h->pred8x8[HOR_PRED8x8    ]= pred8x8_horizontal_c;
    h->pred8x8[PLANE_PRED8x8  ]= pred8x8_plane_c;
    h->pred8x8[LEFT_DC_PRED8x8]= pred8x8_left_dc_c;
    h->pred8x8[TOP_DC_PRED8x8 ]= pred8x8_top_dc_c;
    h->pred8x8[DC_128_PRED8x8 ]= pred8x8_128_dc_c;

    h->pred16x16[DC_PRED8x8     ]= pred16x16_dc_c;
    h->pred16x16[VERT_PRED8x8   ]= pred16x16_vertical_c;
    h->pred16x16[HOR_PRED8x8    ]= pred16x16_horizontal_c;
    h->pred16x16[PLANE_PRED8x8  ]= pred16x16_plane_c;
    h->pred16x16[LEFT_DC_PRED8x8]= pred16x16_left_dc_c;
    h->pred16x16[TOP_DC_PRED8x8 ]= pred16x16_top_dc_c;
    h->pred16x16[DC_128_PRED8x8 ]= pred16x16_128_dc_c;
}

static void free_tables(H264Context *h){
    av_freep(&h->intra4x4_pred_mode);
2162 2163
    av_freep(&h->chroma_pred_mode_table);
    av_freep(&h->cbp_table);
2164 2165
    av_freep(&h->mvd_table[0]);
    av_freep(&h->mvd_table[1]);
2166 2167
    av_freep(&h->non_zero_count);
    av_freep(&h->slice_table_base);
2168
    av_freep(&h->top_border);
2169
    h->slice_table= NULL;
2170

2171 2172 2173 2174 2175 2176 2177 2178 2179 2180
    av_freep(&h->mb2b_xy);
    av_freep(&h->mb2b8_xy);
}

/**
 * allocates tables.
 * needs widzh/height
 */
static int alloc_tables(H264Context *h){
    MpegEncContext * const s = &h->s;
2181
    const int big_mb_num= s->mb_stride * (s->mb_height+1);
2182 2183 2184
    int x,y;

    CHECKED_ALLOCZ(h->intra4x4_pred_mode, big_mb_num * 8  * sizeof(uint8_t))
2185

2186
    CHECKED_ALLOCZ(h->non_zero_count    , big_mb_num * 16 * sizeof(uint8_t))
2187
    CHECKED_ALLOCZ(h->slice_table_base  , big_mb_num * sizeof(uint8_t))
2188
    CHECKED_ALLOCZ(h->top_border       , s->mb_width * (16+8+8) * sizeof(uint8_t))
Michael Niedermayer's avatar
Michael Niedermayer committed
2189
    CHECKED_ALLOCZ(h->cbp_table, big_mb_num * sizeof(uint16_t))
2190

2191 2192
    if( h->pps.cabac ) {
        CHECKED_ALLOCZ(h->chroma_pred_mode_table, big_mb_num * sizeof(uint8_t))
2193 2194
        CHECKED_ALLOCZ(h->mvd_table[0], 32*big_mb_num * sizeof(uint16_t));
        CHECKED_ALLOCZ(h->mvd_table[1], 32*big_mb_num * sizeof(uint16_t));
2195 2196
    }

2197
    memset(h->slice_table_base, -1, big_mb_num  * sizeof(uint8_t));
2198
    h->slice_table= h->slice_table_base + s->mb_stride + 1;
2199 2200 2201 2202 2203

    CHECKED_ALLOCZ(h->mb2b_xy  , big_mb_num * sizeof(uint16_t));
    CHECKED_ALLOCZ(h->mb2b8_xy , big_mb_num * sizeof(uint16_t));
    for(y=0; y<s->mb_height; y++){
        for(x=0; x<s->mb_width; x++){
2204
            const int mb_xy= x + y*s->mb_stride;
2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227
            const int b_xy = 4*x + 4*y*h->b_stride;
            const int b8_xy= 2*x + 2*y*h->b8_stride;
        
            h->mb2b_xy [mb_xy]= b_xy;
            h->mb2b8_xy[mb_xy]= b8_xy;
        }
    }
    
    return 0;
fail:
    free_tables(h);
    return -1;
}

static void common_init(H264Context *h){
    MpegEncContext * const s = &h->s;

    s->width = s->avctx->width;
    s->height = s->avctx->height;
    s->codec_id= s->avctx->codec->id;
    
    init_pred_ptrs(h);

Michael Niedermayer's avatar
Michael Niedermayer committed
2228
    s->unrestricted_mv=1;
2229 2230 2231 2232 2233 2234 2235
    s->decode=1; //FIXME
}

static int decode_init(AVCodecContext *avctx){
    H264Context *h= avctx->priv_data;
    MpegEncContext * const s = &h->s;

Michael Niedermayer's avatar
Michael Niedermayer committed
2236 2237
    MPV_decode_defaults(s);
    
2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250
    s->avctx = avctx;
    common_init(h);

    s->out_format = FMT_H264;
    s->workaround_bugs= avctx->workaround_bugs;

    // set defaults
//    s->decode_mb= ff_h263_decode_mb;
    s->low_delay= 1;
    avctx->pix_fmt= PIX_FMT_YUV420P;

    decode_init_vlc(h);
    
2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261
    if(avctx->codec_tag != 0x31637661) // avc1
        h->is_avc = 0;
    else {
        if((avctx->extradata_size == 0) || (avctx->extradata == NULL)) {
            av_log(avctx, AV_LOG_ERROR, "AVC codec requires avcC data\n");
            return -1;
        }
        h->is_avc = 1;
        h->got_avcC = 0;
    }

2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
    return 0;
}

static void frame_start(H264Context *h){
    MpegEncContext * const s = &h->s;
    int i;

    MPV_frame_start(s, s->avctx);
    ff_er_frame_start(s);
    h->mmco_index=0;

    assert(s->linesize && s->uvlinesize);

    for(i=0; i<16; i++){
        h->block_offset[i]= 4*((scan8[i] - scan8[0])&7) + 4*s->linesize*((scan8[i] - scan8[0])>>3);
        h->chroma_subblock_offset[i]= 2*((scan8[i] - scan8[0])&7) + 2*s->uvlinesize*((scan8[i] - scan8[0])>>3);
    }
    for(i=0; i<4; i++){
        h->block_offset[16+i]=
        h->block_offset[20+i]= 4*((scan8[i] - scan8[0])&7) + 4*s->uvlinesize*((scan8[i] - scan8[0])>>3);
    }

//    s->decode= (s->flags&CODEC_FLAG_PSNR) || !s->encoding || s->current_picture.reference /*|| h->contains_intra*/ || 1;
}

2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318
static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize){
    MpegEncContext * const s = &h->s;
    int i;
    
    src_y  -=   linesize;
    src_cb -= uvlinesize;
    src_cr -= uvlinesize;

    h->left_border[0]= h->top_border[s->mb_x][15];
    for(i=1; i<17; i++){
        h->left_border[i]= src_y[15+i*  linesize];
    }
    
    *(uint64_t*)(h->top_border[s->mb_x]+0)= *(uint64_t*)(src_y +  16*linesize);
    *(uint64_t*)(h->top_border[s->mb_x]+8)= *(uint64_t*)(src_y +8+16*linesize);

    if(!(s->flags&CODEC_FLAG_GRAY)){
        h->left_border[17  ]= h->top_border[s->mb_x][16+7];
        h->left_border[17+9]= h->top_border[s->mb_x][24+7];
        for(i=1; i<9; i++){
            h->left_border[i+17  ]= src_cb[7+i*uvlinesize];
            h->left_border[i+17+9]= src_cr[7+i*uvlinesize];
        }
        *(uint64_t*)(h->top_border[s->mb_x]+16)= *(uint64_t*)(src_cb+8*uvlinesize);
        *(uint64_t*)(h->top_border[s->mb_x]+24)= *(uint64_t*)(src_cr+8*uvlinesize);
    }
}

static inline void xchg_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int xchg){
    MpegEncContext * const s = &h->s;
    int temp8, i;
    uint64_t temp64;
2319 2320
    int deblock_left = (s->mb_x > 0);
    int deblock_top  = (s->mb_y > 0);
2321 2322 2323 2324 2325 2326 2327 2328 2329 2330

    src_y  -=   linesize + 1;
    src_cb -= uvlinesize + 1;
    src_cr -= uvlinesize + 1;

#define XCHG(a,b,t,xchg)\
t= a;\
if(xchg)\
    a= b;\
b= t;
2331 2332 2333 2334 2335 2336 2337 2338 2339 2340

    if(deblock_left){
        for(i = !deblock_top; i<17; i++){
            XCHG(h->left_border[i     ], src_y [i*  linesize], temp8, xchg);
        }
    }

    if(deblock_top){
        XCHG(*(uint64_t*)(h->top_border[s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
        XCHG(*(uint64_t*)(h->top_border[s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
2341 2342 2343
    }

    if(!(s->flags&CODEC_FLAG_GRAY)){
2344 2345 2346 2347 2348 2349 2350 2351 2352
        if(deblock_left){
            for(i = !deblock_top; i<9; i++){
                XCHG(h->left_border[i+17  ], src_cb[i*uvlinesize], temp8, xchg);
                XCHG(h->left_border[i+17+9], src_cr[i*uvlinesize], temp8, xchg);
            }
        }
        if(deblock_top){
            XCHG(*(uint64_t*)(h->top_border[s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
            XCHG(*(uint64_t*)(h->top_border[s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
2353 2354 2355 2356
        }
    }
}

2357 2358 2359 2360
static void hl_decode_mb(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_x= s->mb_x;
    const int mb_y= s->mb_y;
2361
    const int mb_xy= mb_x + mb_y*s->mb_stride;
2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391
    const int mb_type= s->current_picture.mb_type[mb_xy];
    uint8_t  *dest_y, *dest_cb, *dest_cr;
    int linesize, uvlinesize /*dct_offset*/;
    int i;

    if(!s->decode)
        return;

    if(s->mb_skiped){
    }

    dest_y  = s->current_picture.data[0] + (mb_y * 16* s->linesize  ) + mb_x * 16;
    dest_cb = s->current_picture.data[1] + (mb_y * 8 * s->uvlinesize) + mb_x * 8;
    dest_cr = s->current_picture.data[2] + (mb_y * 8 * s->uvlinesize) + mb_x * 8;

    if (h->mb_field_decoding_flag) {
        linesize = s->linesize * 2;
        uvlinesize = s->uvlinesize * 2;
        if(mb_y&1){ //FIXME move out of this func?
            dest_y -= s->linesize*15;
            dest_cb-= s->linesize*7;
            dest_cr-= s->linesize*7;
        }
    } else {
        linesize = s->linesize;
        uvlinesize = s->uvlinesize;
//        dct_offset = s->linesize * 16;
    }

    if(IS_INTRA(mb_type)){
2392 2393 2394
        if(h->deblocking_filter)
            xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1);

2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411
        if(!(s->flags&CODEC_FLAG_GRAY)){
            h->pred8x8[ h->chroma_pred_mode ](dest_cb, uvlinesize);
            h->pred8x8[ h->chroma_pred_mode ](dest_cr, uvlinesize);
        }

        if(IS_INTRA4x4(mb_type)){
            if(!s->encoding){
                for(i=0; i<16; i++){
                    uint8_t * const ptr= dest_y + h->block_offset[i];
                    uint8_t *topright= ptr + 4 - linesize;
                    const int topright_avail= (h->topright_samples_available<<i)&0x8000;
                    const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ];
                    int tr;

                    if(!topright_avail){
                        tr= ptr[3 - linesize]*0x01010101;
                        topright= (uint8_t*) &tr;
2412 2413 2414
                    }else if(i==5 && h->deblocking_filter){
                        tr= *(uint32_t*)h->top_border[mb_x+1];
                        topright= (uint8_t*) &tr;
2415 2416 2417
                    }

                    h->pred4x4[ dir ](ptr, topright, linesize);
2418 2419
                    if(h->non_zero_count_cache[ scan8[i] ]){
                        if(s->codec_id == CODEC_ID_H264)
2420
                            s->dsp.h264_idct_add(ptr, h->mb + i*16, linesize);
2421 2422 2423
                        else
                            svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, 0);
                    }
2424 2425 2426 2427
                }
            }
        }else{
            h->pred16x16[ h->intra16x16_pred_mode ](dest_y , linesize);
2428 2429 2430 2431
            if(s->codec_id == CODEC_ID_H264)
                h264_luma_dc_dequant_idct_c(h->mb, s->qscale);
            else
                svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
2432
        }
2433 2434
        if(h->deblocking_filter)
            xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0);
2435
    }else if(s->codec_id == CODEC_ID_H264){
2436 2437 2438 2439 2440 2441 2442
        hl_motion(h, dest_y, dest_cb, dest_cr,
                  s->dsp.put_h264_qpel_pixels_tab, s->dsp.put_h264_chroma_pixels_tab, 
                  s->dsp.avg_h264_qpel_pixels_tab, s->dsp.avg_h264_chroma_pixels_tab);
    }


    if(!IS_INTRA4x4(mb_type)){
Michael Niedermayer's avatar
Michael Niedermayer committed
2443 2444 2445 2446
        if(s->codec_id == CODEC_ID_H264){
            for(i=0; i<16; i++){
                if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
                    uint8_t * const ptr= dest_y + h->block_offset[i];
2447
                    s->dsp.h264_idct_add(ptr, h->mb + i*16, linesize);
Michael Niedermayer's avatar
Michael Niedermayer committed
2448 2449 2450 2451 2452 2453
                }
            }
        }else{
            for(i=0; i<16; i++){
                if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
                    uint8_t * const ptr= dest_y + h->block_offset[i];
2454
                    svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0);
Michael Niedermayer's avatar
Michael Niedermayer committed
2455
                }
2456 2457 2458 2459 2460 2461 2462
            }
        }
    }

    if(!(s->flags&CODEC_FLAG_GRAY)){
        chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp);
        chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp);
Michael Niedermayer's avatar
Michael Niedermayer committed
2463 2464 2465 2466
        if(s->codec_id == CODEC_ID_H264){
            for(i=16; i<16+4; i++){
                if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
                    uint8_t * const ptr= dest_cb + h->block_offset[i];
2467
                    s->dsp.h264_idct_add(ptr, h->mb + i*16, uvlinesize);
Michael Niedermayer's avatar
Michael Niedermayer committed
2468
                }
2469
            }
Michael Niedermayer's avatar
Michael Niedermayer committed
2470 2471 2472
            for(i=20; i<20+4; i++){
                if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
                    uint8_t * const ptr= dest_cr + h->block_offset[i];
2473
                    s->dsp.h264_idct_add(ptr, h->mb + i*16, uvlinesize);
Michael Niedermayer's avatar
Michael Niedermayer committed
2474 2475 2476 2477 2478 2479
                }
            }
        }else{
            for(i=16; i<16+4; i++){
                if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
                    uint8_t * const ptr= dest_cb + h->block_offset[i];
2480
                    svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, chroma_qp[s->qscale + 12] - 12, 2);
Michael Niedermayer's avatar
Michael Niedermayer committed
2481 2482 2483 2484 2485 2486 2487
                }
            }
            for(i=20; i<20+4; i++){
                if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
                    uint8_t * const ptr= dest_cr + h->block_offset[i];
                    svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, chroma_qp[s->qscale + 12] - 12, 2);
                }
2488 2489 2490
            }
        }
    }
2491 2492 2493 2494
    if(h->deblocking_filter) {
        backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
        filter_mb(h, mb_x, mb_y, dest_y, dest_cb, dest_cr);
    }
2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510
}

/**
 * fills the default_ref_list.
 */
static int fill_default_ref_list(H264Context *h){
    MpegEncContext * const s = &h->s;
    int i;
    Picture sorted_short_ref[16];
    
    if(h->slice_type==B_TYPE){
        int out_i;
        int limit= -1;

        for(out_i=0; out_i<h->short_ref_count; out_i++){
            int best_i=-1;
2511
            int best_poc=INT_MAX;
2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605

            for(i=0; i<h->short_ref_count; i++){
                const int poc= h->short_ref[i]->poc;
                if(poc > limit && poc < best_poc){
                    best_poc= poc;
                    best_i= i;
                }
            }
            
            assert(best_i != -1);
            
            limit= best_poc;
            sorted_short_ref[out_i]= *h->short_ref[best_i];
        }
    }

    if(s->picture_structure == PICT_FRAME){
        if(h->slice_type==B_TYPE){
            const int current_poc= s->current_picture_ptr->poc;
            int list;

            for(list=0; list<2; list++){
                int index=0;

                for(i=0; i<h->short_ref_count && index < h->ref_count[list]; i++){
                    const int i2= list ? h->short_ref_count - i - 1 : i;
                    const int poc= sorted_short_ref[i2].poc;
                    
                    if(sorted_short_ref[i2].reference != 3) continue; //FIXME refernce field shit

                    if((list==1 && poc > current_poc) || (list==0 && poc < current_poc)){
                        h->default_ref_list[list][index  ]= sorted_short_ref[i2];
                        h->default_ref_list[list][index++].pic_id= sorted_short_ref[i2].frame_num;
                    }
                }

                for(i=0; i<h->long_ref_count && index < h->ref_count[ list ]; i++){
                    if(h->long_ref[i]->reference != 3) continue;

                    h->default_ref_list[ list ][index  ]= *h->long_ref[i];
                    h->default_ref_list[ list ][index++].pic_id= i;;
                }
                
                if(h->long_ref_count > 1 && h->short_ref_count==0){
                    Picture temp= h->default_ref_list[1][0];
                    h->default_ref_list[1][0] = h->default_ref_list[1][1];
                    h->default_ref_list[1][0] = temp;
                }

                if(index < h->ref_count[ list ])
                    memset(&h->default_ref_list[list][index], 0, sizeof(Picture)*(h->ref_count[ list ] - index));
            }
        }else{
            int index=0;
            for(i=0; i<h->short_ref_count && index < h->ref_count[0]; i++){
                if(h->short_ref[i]->reference != 3) continue; //FIXME refernce field shit
                h->default_ref_list[0][index  ]= *h->short_ref[i];
                h->default_ref_list[0][index++].pic_id= h->short_ref[i]->frame_num;
            }
            for(i=0; i<h->long_ref_count && index < h->ref_count[0]; i++){
                if(h->long_ref[i]->reference != 3) continue;
                h->default_ref_list[0][index  ]= *h->long_ref[i];
                h->default_ref_list[0][index++].pic_id= i;;
            }
            if(index < h->ref_count[0])
                memset(&h->default_ref_list[0][index], 0, sizeof(Picture)*(h->ref_count[0] - index));
        }
    }else{ //FIELD
        if(h->slice_type==B_TYPE){
        }else{
            //FIXME second field balh
        }
    }
    return 0;
}

static int decode_ref_pic_list_reordering(H264Context *h){
    MpegEncContext * const s = &h->s;
    int list;
    
    if(h->slice_type==I_TYPE || h->slice_type==SI_TYPE) return 0; //FIXME move beofre func
    
    for(list=0; list<2; list++){
        memcpy(h->ref_list[list], h->default_ref_list[list], sizeof(Picture)*h->ref_count[list]);

        if(get_bits1(&s->gb)){
            int pred= h->curr_pic_num;
            int index;

            for(index=0; ; index++){
                int reordering_of_pic_nums_idc= get_ue_golomb(&s->gb);
                int pic_id;
                int i;
                
2606 2607
                if(reordering_of_pic_nums_idc==3) 
                    break;
2608 2609
                
                if(index >= h->ref_count[list]){
2610
                    av_log(h->s.avctx, AV_LOG_ERROR, "reference count overflow\n");
2611 2612 2613 2614 2615 2616 2617 2618
                    return -1;
                }
                
                if(reordering_of_pic_nums_idc<3){
                    if(reordering_of_pic_nums_idc<2){
                        const int abs_diff_pic_num= get_ue_golomb(&s->gb) + 1;

                        if(abs_diff_pic_num >= h->max_pic_num){
2619
                            av_log(h->s.avctx, AV_LOG_ERROR, "abs_diff_pic_num overflow\n");
2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640
                            return -1;
                        }

                        if(reordering_of_pic_nums_idc == 0) pred-= abs_diff_pic_num;
                        else                                pred+= abs_diff_pic_num;
                        pred &= h->max_pic_num - 1;
                    
                        for(i= h->ref_count[list]-1; i>=index; i--){
                            if(h->ref_list[list][i].pic_id == pred && h->ref_list[list][i].long_ref==0)
                                break;
                        }
                    }else{
                        pic_id= get_ue_golomb(&s->gb); //long_term_pic_idx

                        for(i= h->ref_count[list]-1; i>=index; i--){
                            if(h->ref_list[list][i].pic_id == pic_id && h->ref_list[list][i].long_ref==1)
                                break;
                        }
                    }

                    if(i < index){
2641
                        av_log(h->s.avctx, AV_LOG_ERROR, "reference picture missing during reorder\n");
2642 2643 2644 2645 2646 2647 2648 2649
                        memset(&h->ref_list[list][index], 0, sizeof(Picture)); //FIXME
                    }else if(i > index){
                        Picture tmp= h->ref_list[list][i];
                        for(; i>index; i--){
                            h->ref_list[list][i]= h->ref_list[list][i-1];
                        }
                        h->ref_list[list][index]= tmp;
                    }
2650
                }else{
2651
                    av_log(h->s.avctx, AV_LOG_ERROR, "illegal reordering_of_pic_nums_idc\n");
2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716
                    return -1;
                }
            }
        }

        if(h->slice_type!=B_TYPE) break;
    }
    return 0;    
}

static int pred_weight_table(H264Context *h){
    MpegEncContext * const s = &h->s;
    int list, i;
    
    h->luma_log2_weight_denom= get_ue_golomb(&s->gb);
    h->chroma_log2_weight_denom= get_ue_golomb(&s->gb);

    for(list=0; list<2; list++){
        for(i=0; i<h->ref_count[list]; i++){
            int luma_weight_flag, chroma_weight_flag;
            
            luma_weight_flag= get_bits1(&s->gb);
            if(luma_weight_flag){
                h->luma_weight[list][i]= get_se_golomb(&s->gb);
                h->luma_offset[list][i]= get_se_golomb(&s->gb);
            }

            chroma_weight_flag= get_bits1(&s->gb);
            if(chroma_weight_flag){
                int j;
                for(j=0; j<2; j++){
                    h->chroma_weight[list][i][j]= get_se_golomb(&s->gb);
                    h->chroma_offset[list][i][j]= get_se_golomb(&s->gb);
                }
            }
        }
        if(h->slice_type != B_TYPE) break;
    }
    return 0;
}

/**
 * instantaneos decoder refresh.
 */
static void idr(H264Context *h){
    int i;

    for(i=0; i<h->long_ref_count; i++){
        h->long_ref[i]->reference=0;
        h->long_ref[i]= NULL;
    }
    h->long_ref_count=0;

    for(i=0; i<h->short_ref_count; i++){
        h->short_ref[i]->reference=0;
        h->short_ref[i]= NULL;
    }
    h->short_ref_count=0;
}

/**
 *
 * @return the removed picture or NULL if an error occures
 */
static Picture * remove_short(H264Context *h, int frame_num){
2717
    MpegEncContext * const s = &h->s;
2718 2719
    int i;
    
2720
    if(s->avctx->debug&FF_DEBUG_MMCO)
2721
        av_log(h->s.avctx, AV_LOG_DEBUG, "remove short %d count %d\n", frame_num, h->short_ref_count);
2722
    
2723 2724
    for(i=0; i<h->short_ref_count; i++){
        Picture *pic= h->short_ref[i];
2725
        if(s->avctx->debug&FF_DEBUG_MMCO)
2726
            av_log(h->s.avctx, AV_LOG_DEBUG, "%d %d %p\n", i, pic->frame_num, pic);
2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764
        if(pic->frame_num == frame_num){
            h->short_ref[i]= NULL;
            memmove(&h->short_ref[i], &h->short_ref[i+1], (h->short_ref_count - i - 1)*sizeof(Picture*));
            h->short_ref_count--;
            return pic;
        }
    }
    return NULL;
}

/**
 *
 * @return the removed picture or NULL if an error occures
 */
static Picture * remove_long(H264Context *h, int i){
    Picture *pic;

    if(i >= h->long_ref_count) return NULL;
    pic= h->long_ref[i];
    if(pic==NULL) return NULL;
    
    h->long_ref[i]= NULL;
    memmove(&h->long_ref[i], &h->long_ref[i+1], (h->long_ref_count - i - 1)*sizeof(Picture*));
    h->long_ref_count--;

    return pic;
}

/**
 * Executes the reference picture marking (memory management control operations).
 */
static int execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
    MpegEncContext * const s = &h->s;
    int i;
    int current_is_long=0;
    Picture *pic;
    
    if((s->avctx->debug&FF_DEBUG_MMCO) && mmco_count==0)
2765
        av_log(h->s.avctx, AV_LOG_DEBUG, "no mmco here\n");
2766 2767 2768
        
    for(i=0; i<mmco_count; i++){
        if(s->avctx->debug&FF_DEBUG_MMCO)
2769
            av_log(h->s.avctx, AV_LOG_DEBUG, "mmco:%d %d %d\n", h->mmco[i].opcode, h->mmco[i].short_frame_num, h->mmco[i].long_index);
2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826

        switch(mmco[i].opcode){
        case MMCO_SHORT2UNUSED:
            pic= remove_short(h, mmco[i].short_frame_num);
            if(pic==NULL) return -1;
            pic->reference= 0;
            break;
        case MMCO_SHORT2LONG:
            pic= remove_long(h, mmco[i].long_index);
            if(pic) pic->reference=0;
            
            h->long_ref[ mmco[i].long_index ]= remove_short(h, mmco[i].short_frame_num);
            h->long_ref[ mmco[i].long_index ]->long_ref=1;
            break;
        case MMCO_LONG2UNUSED:
            pic= remove_long(h, mmco[i].long_index);
            if(pic==NULL) return -1;
            pic->reference= 0;
            break;
        case MMCO_LONG:
            pic= remove_long(h, mmco[i].long_index);
            if(pic) pic->reference=0;
            
            h->long_ref[ mmco[i].long_index ]= s->current_picture_ptr;
            h->long_ref[ mmco[i].long_index ]->long_ref=1;
            h->long_ref_count++;
            
            current_is_long=1;
            break;
        case MMCO_SET_MAX_LONG:
            assert(mmco[i].long_index <= 16);
            while(mmco[i].long_index < h->long_ref_count){
                pic= remove_long(h, mmco[i].long_index);
                pic->reference=0;
            }
            while(mmco[i].long_index > h->long_ref_count){
                h->long_ref[ h->long_ref_count++ ]= NULL;
            }
            break;
        case MMCO_RESET:
            while(h->short_ref_count){
                pic= remove_short(h, h->short_ref[0]->frame_num);
                pic->reference=0;
            }
            while(h->long_ref_count){
                pic= remove_long(h, h->long_ref_count-1);
                pic->reference=0;
            }
            break;
        default: assert(0);
        }
    }
    
    if(!current_is_long){
        pic= remove_short(h, s->current_picture_ptr->frame_num);
        if(pic){
            pic->reference=0;
2827
            av_log(h->s.avctx, AV_LOG_ERROR, "illegal short term buffer state detected\n");
2828 2829 2830
        }
        
        if(h->short_ref_count)
2831 2832 2833
            memmove(&h->short_ref[1], &h->short_ref[0], h->short_ref_count*sizeof(Picture*));

        h->short_ref[0]= s->current_picture_ptr;
2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869
        h->short_ref[0]->long_ref=0;
        h->short_ref_count++;
    }
    
    return 0; 
}

static int decode_ref_pic_marking(H264Context *h){
    MpegEncContext * const s = &h->s;
    int i;
    
    if(h->nal_unit_type == NAL_IDR_SLICE){ //FIXME fields
        s->broken_link= get_bits1(&s->gb) -1;
        h->mmco[0].long_index= get_bits1(&s->gb) - 1; // current_long_term_idx
        if(h->mmco[0].long_index == -1)
            h->mmco_index= 0;
        else{
            h->mmco[0].opcode= MMCO_LONG;
            h->mmco_index= 1;
        } 
    }else{
        if(get_bits1(&s->gb)){ // adaptive_ref_pic_marking_mode_flag
            for(i= h->mmco_index; i<MAX_MMCO_COUNT; i++) { 
                MMCOOpcode opcode= get_ue_golomb(&s->gb);;

                h->mmco[i].opcode= opcode;
                if(opcode==MMCO_SHORT2UNUSED || opcode==MMCO_SHORT2LONG){
                    h->mmco[i].short_frame_num= (h->frame_num - get_ue_golomb(&s->gb) - 1) & ((1<<h->sps.log2_max_frame_num)-1); //FIXME fields
/*                    if(h->mmco[i].short_frame_num >= h->short_ref_count || h->short_ref[ h->mmco[i].short_frame_num ] == NULL){
                        fprintf(stderr, "illegal short ref in memory management control operation %d\n", mmco);
                        return -1;
                    }*/
                }
                if(opcode==MMCO_SHORT2LONG || opcode==MMCO_LONG2UNUSED || opcode==MMCO_LONG || opcode==MMCO_SET_MAX_LONG){
                    h->mmco[i].long_index= get_ue_golomb(&s->gb);
                    if(/*h->mmco[i].long_index >= h->long_ref_count || h->long_ref[ h->mmco[i].long_index ] == NULL*/ h->mmco[i].long_index >= 16){
2870
                        av_log(h->s.avctx, AV_LOG_ERROR, "illegal long ref in memory management control operation %d\n", opcode);
2871 2872 2873 2874 2875
                        return -1;
                    }
                }
                    
                if(opcode > MMCO_LONG){
2876
                    av_log(h->s.avctx, AV_LOG_ERROR, "illegal memory management control operation %d\n", opcode);
2877 2878
                    return -1;
                }
2879 2880
                if(opcode == MMCO_END)
                    break;
2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997
            }
            h->mmco_index= i;
        }else{
            assert(h->long_ref_count + h->short_ref_count <= h->sps.ref_frame_count);

            if(h->long_ref_count + h->short_ref_count == h->sps.ref_frame_count){ //FIXME fields
                h->mmco[0].opcode= MMCO_SHORT2UNUSED;
                h->mmco[0].short_frame_num= h->short_ref[ h->short_ref_count - 1 ]->frame_num;
                h->mmco_index= 1;
            }else
                h->mmco_index= 0;
        }
    }
    
    return 0; 
}

static int init_poc(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int max_frame_num= 1<<h->sps.log2_max_frame_num;
    int field_poc[2];

    if(h->nal_unit_type == NAL_IDR_SLICE){
        h->frame_num_offset= 0;
    }else{
        if(h->frame_num < h->prev_frame_num)
            h->frame_num_offset= h->prev_frame_num_offset + max_frame_num;
        else
            h->frame_num_offset= h->prev_frame_num_offset;
    }

    if(h->sps.poc_type==0){
        const int max_poc_lsb= 1<<h->sps.log2_max_poc_lsb;

        if     (h->poc_lsb < h->prev_poc_lsb && h->prev_poc_lsb - h->poc_lsb >= max_poc_lsb/2)
            h->poc_msb = h->prev_poc_msb + max_poc_lsb;
        else if(h->poc_lsb > h->prev_poc_lsb && h->prev_poc_lsb - h->poc_lsb < -max_poc_lsb/2)
            h->poc_msb = h->prev_poc_msb - max_poc_lsb;
        else
            h->poc_msb = h->prev_poc_msb;
//printf("poc: %d %d\n", h->poc_msb, h->poc_lsb);
        field_poc[0] = 
        field_poc[1] = h->poc_msb + h->poc_lsb;
        if(s->picture_structure == PICT_FRAME) 
            field_poc[1] += h->delta_poc_bottom;
    }else if(h->sps.poc_type==1){
        int abs_frame_num, expected_delta_per_poc_cycle, expectedpoc;
        int i;

        if(h->sps.poc_cycle_length != 0)
            abs_frame_num = h->frame_num_offset + h->frame_num;
        else
            abs_frame_num = 0;

        if(h->nal_ref_idc==0 && abs_frame_num > 0)
            abs_frame_num--;
            
        expected_delta_per_poc_cycle = 0;
        for(i=0; i < h->sps.poc_cycle_length; i++)
            expected_delta_per_poc_cycle += h->sps.offset_for_ref_frame[ i ]; //FIXME integrate during sps parse

        if(abs_frame_num > 0){
            int poc_cycle_cnt          = (abs_frame_num - 1) / h->sps.poc_cycle_length;
            int frame_num_in_poc_cycle = (abs_frame_num - 1) % h->sps.poc_cycle_length;

            expectedpoc = poc_cycle_cnt * expected_delta_per_poc_cycle;
            for(i = 0; i <= frame_num_in_poc_cycle; i++)
                expectedpoc = expectedpoc + h->sps.offset_for_ref_frame[ i ];
        } else
            expectedpoc = 0;

        if(h->nal_ref_idc == 0) 
            expectedpoc = expectedpoc + h->sps.offset_for_non_ref_pic;
        
        field_poc[0] = expectedpoc + h->delta_poc[0];
        field_poc[1] = field_poc[0] + h->sps.offset_for_top_to_bottom_field;

        if(s->picture_structure == PICT_FRAME)
            field_poc[1] += h->delta_poc[1];
    }else{
        int poc;
        if(h->nal_unit_type == NAL_IDR_SLICE){
            poc= 0;
        }else{
            if(h->nal_ref_idc) poc= 2*(h->frame_num_offset + h->frame_num);
            else               poc= 2*(h->frame_num_offset + h->frame_num) - 1;
        }
        field_poc[0]= poc;
        field_poc[1]= poc;
    }
    
    if(s->picture_structure != PICT_BOTTOM_FIELD)
        s->current_picture_ptr->field_poc[0]= field_poc[0];
    if(s->picture_structure != PICT_TOP_FIELD)
        s->current_picture_ptr->field_poc[1]= field_poc[1];
    if(s->picture_structure == PICT_FRAME) // FIXME field pix?
        s->current_picture_ptr->poc= FFMIN(field_poc[0], field_poc[1]);

    return 0;
}

/**
 * decodes a slice header.
 * this will allso call MPV_common_init() and frame_start() as needed
 */
static int decode_slice_header(H264Context *h){
    MpegEncContext * const s = &h->s;
    int first_mb_in_slice, pps_id;
    int num_ref_idx_active_override_flag;
    static const uint8_t slice_type_map[5]= {P_TYPE, B_TYPE, I_TYPE, SP_TYPE, SI_TYPE};

    s->current_picture.reference= h->nal_ref_idc != 0;

    first_mb_in_slice= get_ue_golomb(&s->gb);

    h->slice_type= get_ue_golomb(&s->gb);
    if(h->slice_type > 9){
2998
        av_log(h->s.avctx, AV_LOG_ERROR, "slice type too large (%d) at %d %d\n", h->slice_type, s->mb_x, s->mb_y);
2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011
    }
    if(h->slice_type > 4){
        h->slice_type -= 5;
        h->slice_type_fixed=1;
    }else
        h->slice_type_fixed=0;
    
    h->slice_type= slice_type_map[ h->slice_type ];
    
    s->pict_type= h->slice_type; // to make a few old func happy, its wrong though
        
    pps_id= get_ue_golomb(&s->gb);
    if(pps_id>255){
3012
        av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
3013 3014 3015
        return -1;
    }
    h->pps= h->pps_buffer[pps_id];
3016
    if(h->pps.slice_group_count == 0){
3017
        av_log(h->s.avctx, AV_LOG_ERROR, "non existing PPS referenced\n");
3018 3019 3020
        return -1;
    }

3021
    h->sps= h->sps_buffer[ h->pps.sps_id ];
3022
    if(h->sps.log2_max_frame_num == 0){
3023
        av_log(h->s.avctx, AV_LOG_ERROR, "non existing SPS referenced\n");
3024 3025
        return -1;
    }
3026 3027 3028 3029 3030 3031 3032 3033 3034 3035
    
    s->mb_width= h->sps.mb_width;
    s->mb_height= h->sps.mb_height;
    
    h->b_stride=  s->mb_width*4;
    h->b8_stride= s->mb_width*2;

    s->mb_x = first_mb_in_slice % s->mb_width;
    s->mb_y = first_mb_in_slice / s->mb_width; //FIXME AFFW
    
3036
    s->width = 16*s->mb_width - 2*(h->sps.crop_left + h->sps.crop_right );
3037
    if(h->sps.frame_mbs_only_flag)
3038
        s->height= 16*s->mb_height - 2*(h->sps.crop_top  + h->sps.crop_bottom);
3039
    else
3040
        s->height= 16*s->mb_height - 4*(h->sps.crop_top  + h->sps.crop_bottom); //FIXME recheck
3041 3042
    
    if (s->context_initialized 
Michael Niedermayer's avatar
Michael Niedermayer committed
3043
        && (   s->width != s->avctx->width || s->height != s->avctx->height)) {
3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054
        free_tables(h);
        MPV_common_end(s);
    }
    if (!s->context_initialized) {
        if (MPV_common_init(s) < 0)
            return -1;

        alloc_tables(h);

        s->avctx->width = s->width;
        s->avctx->height = s->height;
Michael Niedermayer's avatar
Michael Niedermayer committed
3055
        s->avctx->sample_aspect_ratio= h->sps.sar;
3056 3057 3058 3059 3060

        if(h->sps.timing_info_present_flag && h->sps.fixed_frame_rate_flag){
            s->avctx->frame_rate = h->sps.time_scale;
            s->avctx->frame_rate_base = h->sps.num_units_in_tick;
        }
3061 3062 3063 3064 3065 3066
    }

    if(first_mb_in_slice == 0){
        frame_start(h);
    }

3067
    s->current_picture_ptr->frame_num= //FIXME frame_num cleanup
3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087
    h->frame_num= get_bits(&s->gb, h->sps.log2_max_frame_num);

    if(h->sps.frame_mbs_only_flag){
        s->picture_structure= PICT_FRAME;
    }else{
        if(get_bits1(&s->gb)) //field_pic_flag
            s->picture_structure= PICT_TOP_FIELD + get_bits1(&s->gb); //bottom_field_flag
        else
            s->picture_structure= PICT_FRAME;
    }

    if(s->picture_structure==PICT_FRAME){
        h->curr_pic_num=   h->frame_num;
        h->max_pic_num= 1<< h->sps.log2_max_frame_num;
    }else{
        h->curr_pic_num= 2*h->frame_num;
        h->max_pic_num= 1<<(h->sps.log2_max_frame_num + 1);
    }
        
    if(h->nal_unit_type == NAL_IDR_SLICE){
Fabrice Bellard's avatar
Fabrice Bellard committed
3088
        get_ue_golomb(&s->gb); /* idr_pic_id */
3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127
    }
   
    if(h->sps.poc_type==0){
        h->poc_lsb= get_bits(&s->gb, h->sps.log2_max_poc_lsb);
        
        if(h->pps.pic_order_present==1 && s->picture_structure==PICT_FRAME){
            h->delta_poc_bottom= get_se_golomb(&s->gb);
        }
    }
    
    if(h->sps.poc_type==1 && !h->sps.delta_pic_order_always_zero_flag){
        h->delta_poc[0]= get_se_golomb(&s->gb);
        
        if(h->pps.pic_order_present==1 && s->picture_structure==PICT_FRAME)
            h->delta_poc[1]= get_se_golomb(&s->gb);
    }
    
    init_poc(h);
    
    if(h->pps.redundant_pic_cnt_present){
        h->redundant_pic_count= get_ue_golomb(&s->gb);
    }

    //set defaults, might be overriden a few line later
    h->ref_count[0]= h->pps.ref_count[0];
    h->ref_count[1]= h->pps.ref_count[1];

    if(h->slice_type == P_TYPE || h->slice_type == SP_TYPE || h->slice_type == B_TYPE){
        if(h->slice_type == B_TYPE){
            h->direct_spatial_mv_pred= get_bits1(&s->gb);
        }
        num_ref_idx_active_override_flag= get_bits1(&s->gb);
    
        if(num_ref_idx_active_override_flag){
            h->ref_count[0]= get_ue_golomb(&s->gb) + 1;
            if(h->slice_type==B_TYPE)
                h->ref_count[1]= get_ue_golomb(&s->gb) + 1;

            if(h->ref_count[0] > 32 || h->ref_count[1] > 32){
3128
                av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n");
3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146
                return -1;
            }
        }
    }

    if(first_mb_in_slice == 0){
        fill_default_ref_list(h);
    }

    decode_ref_pic_list_reordering(h);

    if(   (h->pps.weighted_pred          && (h->slice_type == P_TYPE || h->slice_type == SP_TYPE )) 
       || (h->pps.weighted_bipred_idc==1 && h->slice_type==B_TYPE ) )
        pred_weight_table(h);
    
    if(s->current_picture.reference)
        decode_ref_pic_marking(h);

3147 3148 3149 3150 3151
    if( h->slice_type != I_TYPE && h->slice_type != SI_TYPE && h->pps.cabac )
        h->cabac_init_idc = get_ue_golomb(&s->gb);

    h->last_qscale_diff = 0;
    s->qscale = h->pps.init_qp + get_se_golomb(&s->gb);
3152 3153 3154 3155
    if(s->qscale<0 || s->qscale>51){
        av_log(s->avctx, AV_LOG_ERROR, "QP %d out of range\n", s->qscale);
        return -1;
    }
3156 3157
    //FIXME qscale / qp ... stuff
    if(h->slice_type == SP_TYPE){
Fabrice Bellard's avatar
Fabrice Bellard committed
3158
        get_bits1(&s->gb); /* sp_for_switch_flag */
3159 3160
    }
    if(h->slice_type==SP_TYPE || h->slice_type == SI_TYPE){
Fabrice Bellard's avatar
Fabrice Bellard committed
3161
        get_se_golomb(&s->gb); /* slice_qs_delta */
3162 3163
    }

3164
    h->deblocking_filter = 1;
3165 3166
    h->slice_alpha_c0_offset = 0;
    h->slice_beta_offset = 0;
3167
    if( h->pps.deblocking_filter_parameters_present ) {
3168 3169 3170 3171 3172
        h->deblocking_filter= get_ue_golomb(&s->gb);
        if(h->deblocking_filter < 2) 
            h->deblocking_filter^= 1; // 1<->0

        if( h->deblocking_filter ) {
3173 3174
            h->slice_alpha_c0_offset = get_se_golomb(&s->gb) << 1;
            h->slice_beta_offset = get_se_golomb(&s->gb) << 1;
3175
        }
3176
    }
3177 3178 3179 3180 3181 3182 3183

#if 0 //FMO
    if( h->pps.num_slice_groups > 1  && h->pps.mb_slice_group_map_type >= 3 && h->pps.mb_slice_group_map_type <= 5)
        slice_group_change_cycle= get_bits(&s->gb, ?);
#endif

    if(s->avctx->debug&FF_DEBUG_PICT_INFO){
3184
        av_log(h->s.avctx, AV_LOG_DEBUG, "mb:%d %c pps:%d frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d\n", 
3185
               first_mb_in_slice, 
Michael Niedermayer's avatar
Michael Niedermayer committed
3186
               av_get_pict_type_char(h->slice_type),
3187 3188 3189 3190
               pps_id, h->frame_num,
               s->current_picture_ptr->field_poc[0], s->current_picture_ptr->field_poc[1],
               h->ref_count[0], h->ref_count[1],
               s->qscale,
3191
               h->deblocking_filter
3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211
               );
    }

    return 0;
}

/**
 *
 */
static inline int get_level_prefix(GetBitContext *gb){
    unsigned int buf;
    int log;
    
    OPEN_READER(re, gb);
    UPDATE_CACHE(re, gb);
    buf=GET_CACHE(re, gb);
    
    log= 32 - av_log2(buf);
#ifdef TRACE
    print_bin(buf>>(32-log), log);
3212
    av_log(NULL, AV_LOG_DEBUG, "%5d %2d %3d lpr @%5d in %s get_level_prefix\n", buf>>(32-log), log, log-1, get_bits_count(gb), __FILE__);
3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258
#endif

    LAST_SKIP_BITS(re, gb, log);
    CLOSE_READER(re, gb);

    return log-1;
}

/**
 * decodes a residual block.
 * @param n block index
 * @param scantable scantable
 * @param max_coeff number of coefficients in the block
 * @return <0 if an error occured
 */
static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, int n, const uint8_t *scantable, int qp, int max_coeff){
    MpegEncContext * const s = &h->s;
    const uint16_t *qmul= dequant_coeff[qp];
    static const int coeff_token_table_index[17]= {0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3};
    int level[16], run[16];
    int suffix_length, zeros_left, coeff_num, coeff_token, total_coeff, i, trailing_ones;

    //FIXME put trailing_onex into the context

    if(n == CHROMA_DC_BLOCK_INDEX){
        coeff_token= get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1);
        total_coeff= coeff_token>>2;
    }else{    
        if(n == LUMA_DC_BLOCK_INDEX){
            total_coeff= pred_non_zero_count(h, 0);
            coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
            total_coeff= coeff_token>>2;
        }else{
            total_coeff= pred_non_zero_count(h, n);
            coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
            total_coeff= coeff_token>>2;
            h->non_zero_count_cache[ scan8[n] ]= total_coeff;
        }
    }

    //FIXME set last_non_zero?

    if(total_coeff==0)
        return 0;
        
    trailing_ones= coeff_token&3;
3259
    tprintf("trailing:%d, total:%d\n", trailing_ones, total_coeff);
3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285
    assert(total_coeff<=16);
    
    for(i=0; i<trailing_ones; i++){
        level[i]= 1 - 2*get_bits1(gb);
    }

    suffix_length= total_coeff > 10 && trailing_ones < 3;

    for(; i<total_coeff; i++){
        const int prefix= get_level_prefix(gb);
        int level_code, mask;

        if(prefix<14){ //FIXME try to build a large unified VLC table for all this
            if(suffix_length)
                level_code= (prefix<<suffix_length) + get_bits(gb, suffix_length); //part
            else
                level_code= (prefix<<suffix_length); //part
        }else if(prefix==14){
            if(suffix_length)
                level_code= (prefix<<suffix_length) + get_bits(gb, suffix_length); //part
            else
                level_code= prefix + get_bits(gb, 4); //part
        }else if(prefix==15){
            level_code= (prefix<<suffix_length) + get_bits(gb, 12); //part
            if(suffix_length==0) level_code+=15; //FIXME doesnt make (much)sense
        }else{
3286
            av_log(h->s.avctx, AV_LOG_ERROR, "prefix too large at %d %d\n", s->mb_x, s->mb_y);
3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300
            return -1;
        }

        if(i==trailing_ones && i<3) level_code+= 2; //FIXME split first iteration

        mask= -(level_code&1);
        level[i]= (((2+level_code)>>1) ^ mask) - mask;

        if(suffix_length==0) suffix_length=1; //FIXME split first iteration

#if 1
        if(ABS(level[i]) > (3<<(suffix_length-1)) && suffix_length<6) suffix_length++;
#else        
        if((2+level_code)>>1) > (3<<(suffix_length-1)) && suffix_length<6) suffix_length++;
3301
        /* ? == prefix > 2 or sth */
3302
#endif
3303
        tprintf("level: %d suffix_length:%d\n", level[i], suffix_length);
3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326
    }

    if(total_coeff == max_coeff)
        zeros_left=0;
    else{
        if(n == CHROMA_DC_BLOCK_INDEX)
            zeros_left= get_vlc2(gb, chroma_dc_total_zeros_vlc[ total_coeff-1 ].table, CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1);
        else
            zeros_left= get_vlc2(gb, total_zeros_vlc[ total_coeff-1 ].table, TOTAL_ZEROS_VLC_BITS, 1);
    }
    
    for(i=0; i<total_coeff-1; i++){
        if(zeros_left <=0)
            break;
        else if(zeros_left < 7){
            run[i]= get_vlc2(gb, run_vlc[zeros_left-1].table, RUN_VLC_BITS, 1);
        }else{
            run[i]= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2);
        }
        zeros_left -= run[i];
    }

    if(zeros_left<0){
3327
        av_log(h->s.avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", s->mb_x, s->mb_y);
3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364
        return -1;
    }
    
    for(; i<total_coeff-1; i++){
        run[i]= 0;
    }

    run[i]= zeros_left;

    coeff_num=-1;
    if(n > 24){
        for(i=total_coeff-1; i>=0; i--){ //FIXME merge into rundecode?
            int j;

            coeff_num += run[i] + 1; //FIXME add 1 earlier ?
            j= scantable[ coeff_num ];

            block[j]= level[i];
        }
    }else{
        for(i=total_coeff-1; i>=0; i--){ //FIXME merge into  rundecode?
            int j;

            coeff_num += run[i] + 1; //FIXME add 1 earlier ?
            j= scantable[ coeff_num ];

            block[j]= level[i] * qmul[j];
//            printf("%d %d  ", block[j], qmul[j]);
        }
    }
    return 0;
}

/**
 * decodes a macroblock
 * @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
 */
3365
static int decode_mb_cavlc(H264Context *h){
3366
    MpegEncContext * const s = &h->s;
3367
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
3368
    int mb_type, partition_count, cbp;
3369

Michael Niedermayer's avatar
Michael Niedermayer committed
3370
    s->dsp.clear_blocks(h->mb); //FIXME avoid if allready clear (move after skip handlong?    
3371

3372
    tprintf("pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
Fabrice Bellard's avatar
Fabrice Bellard committed
3373 3374
    cbp = 0; /* avoid warning. FIXME: find a solution without slowing
                down the code */
3375 3376 3377 3378 3379
    if(h->slice_type != I_TYPE && h->slice_type != SI_TYPE){
        if(s->mb_skip_run==-1)
            s->mb_skip_run= get_ue_golomb(&s->gb);
        
        if (s->mb_skip_run--) {
3380
            int mx, my;
3381 3382 3383 3384
            /* skip mb */
//FIXME b frame
            mb_type= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P1L0;

3385
            memset(h->non_zero_count[mb_xy], 0, 16);
3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397
            memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui

            if(h->sps.mb_aff && s->mb_skip_run==0 && (s->mb_y&1)==0){
                h->mb_field_decoding_flag= get_bits1(&s->gb);
            }

            if(h->mb_field_decoding_flag)
                mb_type|= MB_TYPE_INTERLACED;
            
            fill_caches(h, mb_type); //FIXME check what is needed and what not ...
            pred_pskip_motion(h, &mx, &my);
            fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, 0, 1);
Michael Niedermayer's avatar
Michael Niedermayer committed
3398
            fill_rectangle(  h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mx,my), 4);
3399 3400 3401
            write_back_motion(h, mb_type);

            s->current_picture.mb_type[mb_xy]= mb_type; //FIXME SKIP type
3402
            s->current_picture.qscale_table[mb_xy]= s->qscale;
3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437
            h->slice_table[ mb_xy ]= h->slice_num;

            h->prev_mb_skiped= 1;
            return 0;
        }
    }
    if(h->sps.mb_aff /* && !field pic FIXME needed? */){
        if((s->mb_y&1)==0)
            h->mb_field_decoding_flag = get_bits1(&s->gb);
    }else
        h->mb_field_decoding_flag=0; //FIXME som ed note ?!
    
    h->prev_mb_skiped= 0;
    
    mb_type= get_ue_golomb(&s->gb);
    if(h->slice_type == B_TYPE){
        if(mb_type < 23){
            partition_count= b_mb_type_info[mb_type].partition_count;
            mb_type=         b_mb_type_info[mb_type].type;
        }else{
            mb_type -= 23;
            goto decode_intra_mb;
        }
    }else if(h->slice_type == P_TYPE /*|| h->slice_type == SP_TYPE */){
        if(mb_type < 5){
            partition_count= p_mb_type_info[mb_type].partition_count;
            mb_type=         p_mb_type_info[mb_type].type;
        }else{
            mb_type -= 5;
            goto decode_intra_mb;
        }
    }else{
       assert(h->slice_type == I_TYPE);
decode_intra_mb:
        if(mb_type > 25){
3438
            av_log(h->s.avctx, AV_LOG_ERROR, "mb_type %d in %c slice to large at %d %d\n", mb_type, av_get_pict_type_char(h->slice_type), s->mb_x, s->mb_y);
3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454
            return -1;
        }
        partition_count=0;
        cbp= i_mb_type_info[mb_type].cbp;
        h->intra16x16_pred_mode= i_mb_type_info[mb_type].pred_mode;
        mb_type= i_mb_type_info[mb_type].type;
    }

    if(h->mb_field_decoding_flag)
        mb_type |= MB_TYPE_INTERLACED;

    s->current_picture.mb_type[mb_xy]= mb_type;
    h->slice_table[ mb_xy ]= h->slice_num;
    
    if(IS_INTRA_PCM(mb_type)){
        const uint8_t *ptr;
3455
        int x, y;
3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482
        
        // we assume these blocks are very rare so we dont optimize it
        align_get_bits(&s->gb);
        
        ptr= s->gb.buffer + get_bits_count(&s->gb);
    
        for(y=0; y<16; y++){
            const int index= 4*(y&3) + 64*(y>>2);
            for(x=0; x<16; x++){
                h->mb[index + (x&3) + 16*(x>>2)]= *(ptr++);
            }
        }
        for(y=0; y<8; y++){
            const int index= 256 + 4*(y&3) + 32*(y>>2);
            for(x=0; x<8; x++){
                h->mb[index + (x&3) + 16*(x>>2)]= *(ptr++);
            }
        }
        for(y=0; y<8; y++){
            const int index= 256 + 64 + 4*(y&3) + 32*(y>>2);
            for(x=0; x<8; x++){
                h->mb[index + (x&3) + 16*(x>>2)]= *(ptr++);
            }
        }
    
        skip_bits(&s->gb, 384); //FIXME check /fix the bitstream readers
        
3483 3484
        //FIXME deblock filter, non_zero_count_cache init ...
        memset(h->non_zero_count[mb_xy], 16, 16);
3485
        s->current_picture.qscale_table[mb_xy]= s->qscale;
3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535
        
        return 0;
    }
        
    fill_caches(h, mb_type);

    //mb_pred
    if(IS_INTRA(mb_type)){
//            init_top_left_availability(h);
            if(IS_INTRA4x4(mb_type)){
                int i;

//                fill_intra4x4_pred_table(h);
                for(i=0; i<16; i++){
                    const int mode_coded= !get_bits1(&s->gb);
                    const int predicted_mode=  pred_intra_mode(h, i);
                    int mode;

                    if(mode_coded){
                        const int rem_mode= get_bits(&s->gb, 3);
                        if(rem_mode<predicted_mode)
                            mode= rem_mode;
                        else
                            mode= rem_mode + 1;
                    }else{
                        mode= predicted_mode;
                    }
                    
                    h->intra4x4_pred_mode_cache[ scan8[i] ] = mode;
                }
                write_back_intra_pred_mode(h);
                if( check_intra4x4_pred_mode(h) < 0)
                    return -1;
            }else{
                h->intra16x16_pred_mode= check_intra_pred_mode(h, h->intra16x16_pred_mode);
                if(h->intra16x16_pred_mode < 0)
                    return -1;
            }
            h->chroma_pred_mode= get_ue_golomb(&s->gb);

            h->chroma_pred_mode= check_intra_pred_mode(h, h->chroma_pred_mode);
            if(h->chroma_pred_mode < 0)
                return -1;
    }else if(partition_count==4){
        int i, j, sub_partition_count[4], list, ref[2][4];
        
        if(h->slice_type == B_TYPE){
            for(i=0; i<4; i++){
                h->sub_mb_type[i]= get_ue_golomb(&s->gb);
                if(h->sub_mb_type[i] >=13){
3536
                    av_log(h->s.avctx, AV_LOG_ERROR, "B sub_mb_type %d out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
3537 3538 3539 3540 3541 3542 3543 3544 3545 3546
                    return -1;
                }
                sub_partition_count[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
                h->sub_mb_type[i]=      b_sub_mb_type_info[ h->sub_mb_type[i] ].type;
            }
        }else{
            assert(h->slice_type == P_TYPE || h->slice_type == SP_TYPE); //FIXME SP correct ?
            for(i=0; i<4; i++){
                h->sub_mb_type[i]= get_ue_golomb(&s->gb);
                if(h->sub_mb_type[i] >=4){
3547
                    av_log(h->s.avctx, AV_LOG_ERROR, "P sub_mb_type %d out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585
                    return -1;
                }
                sub_partition_count[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
                h->sub_mb_type[i]=      p_sub_mb_type_info[ h->sub_mb_type[i] ].type;
            }
        }
        
        for(list=0; list<2; list++){
            const int ref_count= IS_REF0(mb_type) ? 1 : h->ref_count[list];
            if(ref_count == 0) continue;
            for(i=0; i<4; i++){
                if(IS_DIR(h->sub_mb_type[i], 0, list) && !IS_DIRECT(h->sub_mb_type[i])){
                    ref[list][i] = get_te0_golomb(&s->gb, ref_count); //FIXME init to 0 before and skip?
                }else{
                 //FIXME
                    ref[list][i] = -1;
                }
            }
        }
        
        for(list=0; list<2; list++){
            const int ref_count= IS_REF0(mb_type) ? 1 : h->ref_count[list];
            if(ref_count == 0) continue;

            for(i=0; i<4; i++){
                h->ref_cache[list][ scan8[4*i]   ]=h->ref_cache[list][ scan8[4*i]+1 ]=
                h->ref_cache[list][ scan8[4*i]+8 ]=h->ref_cache[list][ scan8[4*i]+9 ]= ref[list][i];

                if(IS_DIR(h->sub_mb_type[i], 0, list) && !IS_DIRECT(h->sub_mb_type[i])){
                    const int sub_mb_type= h->sub_mb_type[i];
                    const int block_width= (sub_mb_type & (MB_TYPE_16x16|MB_TYPE_16x8)) ? 2 : 1;
                    for(j=0; j<sub_partition_count[i]; j++){
                        int mx, my;
                        const int index= 4*i + block_width*j;
                        int16_t (* mv_cache)[2]= &h->mv_cache[list][ scan8[index] ];
                        pred_motion(h, index, block_width, list, h->ref_cache[list][ scan8[index] ], &mx, &my);
                        mx += get_se_golomb(&s->gb);
                        my += get_se_golomb(&s->gb);
3586 3587
                        tprintf("final mv:%d %d\n", mx, my);

3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628
                        if(IS_SUB_8X8(sub_mb_type)){
                            mv_cache[ 0 ][0]= mv_cache[ 1 ][0]= 
                            mv_cache[ 8 ][0]= mv_cache[ 9 ][0]= mx;
                            mv_cache[ 0 ][1]= mv_cache[ 1 ][1]= 
                            mv_cache[ 8 ][1]= mv_cache[ 9 ][1]= my;
                        }else if(IS_SUB_8X4(sub_mb_type)){
                            mv_cache[ 0 ][0]= mv_cache[ 1 ][0]= mx;
                            mv_cache[ 0 ][1]= mv_cache[ 1 ][1]= my;
                        }else if(IS_SUB_4X8(sub_mb_type)){
                            mv_cache[ 0 ][0]= mv_cache[ 8 ][0]= mx;
                            mv_cache[ 0 ][1]= mv_cache[ 8 ][1]= my;
                        }else{
                            assert(IS_SUB_4X4(sub_mb_type));
                            mv_cache[ 0 ][0]= mx;
                            mv_cache[ 0 ][1]= my;
                        }
                    }
                }else{
                    uint32_t *p= (uint32_t *)&h->mv_cache[list][ scan8[4*i] ][0];
                    p[0] = p[1]=
                    p[8] = p[9]= 0;
                }
            }
        }
    }else if(!IS_DIRECT(mb_type)){
        int list, mx, my, i;
         //FIXME we should set ref_idx_l? to 0 if we use that later ...
        if(IS_16X16(mb_type)){
            for(list=0; list<2; list++){
                if(h->ref_count[0]>0){
                    if(IS_DIR(mb_type, 0, list)){
                        const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
                        fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, val, 1);
                    }
                }
            }
            for(list=0; list<2; list++){
                if(IS_DIR(mb_type, 0, list)){
                    pred_motion(h, 0, 4, list, h->ref_cache[list][ scan8[0] ], &mx, &my);
                    mx += get_se_golomb(&s->gb);
                    my += get_se_golomb(&s->gb);
3629 3630
                    tprintf("final mv:%d %d\n", mx, my);

Michael Niedermayer's avatar
Michael Niedermayer committed
3631
                    fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651
                }
            }
        }
        else if(IS_16X8(mb_type)){
            for(list=0; list<2; list++){
                if(h->ref_count[list]>0){
                    for(i=0; i<2; i++){
                        if(IS_DIR(mb_type, i, list)){
                            const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
                            fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, val, 1);
                        }
                    }
                }
            }
            for(list=0; list<2; list++){
                for(i=0; i<2; i++){
                    if(IS_DIR(mb_type, i, list)){
                        pred_16x8_motion(h, 8*i, list, h->ref_cache[list][scan8[0] + 16*i], &mx, &my);
                        mx += get_se_golomb(&s->gb);
                        my += get_se_golomb(&s->gb);
3652 3653
                        tprintf("final mv:%d %d\n", mx, my);

Michael Niedermayer's avatar
Michael Niedermayer committed
3654
                        fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx,my), 4);
3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675
                    }
                }
            }
        }else{
            assert(IS_8X16(mb_type));
            for(list=0; list<2; list++){
                if(h->ref_count[list]>0){
                    for(i=0; i<2; i++){
                        if(IS_DIR(mb_type, i, list)){ //FIXME optimize
                            const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
                            fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, val, 1);
                        }
                    }
                }
            }
            for(list=0; list<2; list++){
                for(i=0; i<2; i++){
                    if(IS_DIR(mb_type, i, list)){
                        pred_8x16_motion(h, i*4, list, h->ref_cache[list][ scan8[0] + 2*i ], &mx, &my);
                        mx += get_se_golomb(&s->gb);
                        my += get_se_golomb(&s->gb);
3676 3677
                        tprintf("final mv:%d %d\n", mx, my);

Michael Niedermayer's avatar
Michael Niedermayer committed
3678
                        fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx,my), 4);
3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690
                    }
                }
            }
        }
    }
    
    if(IS_INTER(mb_type))
        write_back_motion(h, mb_type);
    
    if(!IS_INTRA16x16(mb_type)){
        cbp= get_ue_golomb(&s->gb);
        if(cbp > 47){
3691
            av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%d) at %d %d\n", cbp, s->mb_x, s->mb_y);
3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719
            return -1;
        }
        
        if(IS_INTRA4x4(mb_type))
            cbp= golomb_to_intra4x4_cbp[cbp];
        else
            cbp= golomb_to_inter_cbp[cbp];
    }

    if(cbp || IS_INTRA16x16(mb_type)){
        int i8x8, i4x4, chroma_idx;
        int chroma_qp, dquant;
        GetBitContext *gb= IS_INTRA(mb_type) ? h->intra_gb_ptr : h->inter_gb_ptr;
        const uint8_t *scan, *dc_scan;
        
//        fill_non_zero_count_cache(h);

        if(IS_INTERLACED(mb_type)){
            scan= field_scan;
            dc_scan= luma_dc_field_scan;
        }else{
            scan= zigzag_scan;
            dc_scan= luma_dc_zigzag_scan;
        }

        dquant= get_se_golomb(&s->gb);

        if( dquant > 25 || dquant < -26 ){
3720
            av_log(h->s.avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, s->mb_x, s->mb_y);
3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747
            return -1;
        }
        
        s->qscale += dquant;
        if(((unsigned)s->qscale) > 51){
            if(s->qscale<0) s->qscale+= 52;
            else            s->qscale-= 52;
        }
        
        h->chroma_qp= chroma_qp= get_chroma_qp(h, s->qscale);
        if(IS_INTRA16x16(mb_type)){
            if( decode_residual(h, h->intra_gb_ptr, h->mb, LUMA_DC_BLOCK_INDEX, dc_scan, s->qscale, 16) < 0){
                return -1; //FIXME continue if partotioned and other retirn -1 too
            }

            assert((cbp&15) == 0 || (cbp&15) == 15);

            if(cbp&15){
                for(i8x8=0; i8x8<4; i8x8++){
                    for(i4x4=0; i4x4<4; i4x4++){
                        const int index= i4x4 + 4*i8x8;
                        if( decode_residual(h, h->intra_gb_ptr, h->mb + 16*index, index, scan + 1, s->qscale, 15) < 0 ){
                            return -1;
                        }
                    }
                }
            }else{
Michael Niedermayer's avatar
Michael Niedermayer committed
3748
                fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1);
3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788
            }
        }else{
            for(i8x8=0; i8x8<4; i8x8++){
                if(cbp & (1<<i8x8)){
                    for(i4x4=0; i4x4<4; i4x4++){
                        const int index= i4x4 + 4*i8x8;
                        
                        if( decode_residual(h, gb, h->mb + 16*index, index, scan, s->qscale, 16) <0 ){
                            return -1;
                        }
                    }
                }else{
                    uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ];
                    nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;
                }
            }
        }
        
        if(cbp&0x30){
            for(chroma_idx=0; chroma_idx<2; chroma_idx++)
                if( decode_residual(h, gb, h->mb + 256 + 16*4*chroma_idx, CHROMA_DC_BLOCK_INDEX, chroma_dc_scan, chroma_qp, 4) < 0){
                    return -1;
                }
        }

        if(cbp&0x20){
            for(chroma_idx=0; chroma_idx<2; chroma_idx++){
                for(i4x4=0; i4x4<4; i4x4++){
                    const int index= 16 + 4*chroma_idx + i4x4;
                    if( decode_residual(h, gb, h->mb + 16*index, index, scan + 1, chroma_qp, 15) < 0){
                        return -1;
                    }
                }
            }
        }else{
            uint8_t * const nnz= &h->non_zero_count_cache[0];
            nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
            nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
        }
    }else{
3789 3790 3791 3792
        uint8_t * const nnz= &h->non_zero_count_cache[0];
        fill_rectangle(&nnz[scan8[0]], 4, 4, 8, 0, 1);
        nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
        nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
3793
    }
3794
    s->current_picture.qscale_table[mb_xy]= s->qscale;
3795 3796 3797 3798 3799
    write_back_non_zero_count(h);

    return 0;
}

3800 3801 3802 3803 3804 3805
static int decode_cabac_intra_mb_type(H264Context *h, int ctx_base, int intra_slice) {
    uint8_t *state= &h->cabac_state[ctx_base];
    int mb_type;
    
    if(intra_slice){
        MpegEncContext * const s = &h->s;
3806
        const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
3807
        int ctx=0;
3808 3809 3810 3811
        if( s->mb_x > 0 && !IS_INTRA4x4( s->current_picture.mb_type[mb_xy-1] ) )
            ctx++;
        if( s->mb_y > 0 && !IS_INTRA4x4( s->current_picture.mb_type[mb_xy-s->mb_stride] ) )
            ctx++;
3812 3813 3814 3815 3816
        if( get_cabac( &h->cabac, &state[ctx] ) == 0 )
            return 0;   /* I4x4 */
        state += 2;
    }else{
        if( get_cabac( &h->cabac, &state[0] ) == 0 )
3817
            return 0;   /* I4x4 */
3818
    }
3819

3820 3821
    if( get_cabac_terminate( &h->cabac ) )
        return 25;  /* PCM */
3822

3823 3824 3825
    mb_type = 1; /* I16x16 */
    if( get_cabac( &h->cabac, &state[1] ) )
        mb_type += 12;  /* cbp_luma != 0 */
3826

3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838
    if( get_cabac( &h->cabac, &state[2] ) ) {
        if( get_cabac( &h->cabac, &state[2+intra_slice] ) )
            mb_type += 4 * 2;   /* cbp_chroma == 2 */
        else
            mb_type += 4 * 1;   /* cbp_chroma == 1 */
    }
    if( get_cabac( &h->cabac, &state[3+intra_slice] ) )
        mb_type += 2;
    if( get_cabac( &h->cabac, &state[3+2*intra_slice] ) )
        mb_type += 1;
    return mb_type;
}
3839

3840 3841 3842 3843 3844
static int decode_cabac_mb_type( H264Context *h ) {
    MpegEncContext * const s = &h->s;

    if( h->slice_type == I_TYPE ) {
        return decode_cabac_intra_mb_type(h, 3, 1);
3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855
    } else if( h->slice_type == P_TYPE ) {
        if( get_cabac( &h->cabac, &h->cabac_state[14] ) == 0 ) {
            /* P-type */
            if( get_cabac( &h->cabac, &h->cabac_state[15] ) == 0 ) {
                if( get_cabac( &h->cabac, &h->cabac_state[16] ) == 0 )
                    return 0; /* P_L0_D16x16; */
                else
                    return 3; /* P_8x8; */
            } else {
                if( get_cabac( &h->cabac, &h->cabac_state[17] ) == 0 )
                    return 2; /* P_L0_D8x16; */
3856 3857
                else
                    return 1; /* P_L0_D16x8; */
3858 3859
            }
        } else {
3860
            return decode_cabac_intra_mb_type(h, 17, 0) + 5;
3861
        }
3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887
    } else if( h->slice_type == B_TYPE ) {
        const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
        int ctx = 0;
        int bits;

        if( s->mb_x > 0 && !IS_SKIP( s->current_picture.mb_type[mb_xy-1] )
                      && !IS_DIRECT( s->current_picture.mb_type[mb_xy-1] ) )
            ctx++;
        if( s->mb_y > 0 && !IS_SKIP( s->current_picture.mb_type[mb_xy-s->mb_stride] )
                      && !IS_DIRECT( s->current_picture.mb_type[mb_xy-s->mb_stride] ) )
            ctx++;

        if( !get_cabac( &h->cabac, &h->cabac_state[27+ctx] ) )
            return 0; /* B_Direct_16x16 */

        if( !get_cabac( &h->cabac, &h->cabac_state[27+3] ) ) {
            return 1 + get_cabac( &h->cabac, &h->cabac_state[27+5] ); /* B_L[01]_16x16 */
        }

        bits = get_cabac( &h->cabac, &h->cabac_state[27+4] ) << 3;
        bits|= get_cabac( &h->cabac, &h->cabac_state[27+5] ) << 2;
        bits|= get_cabac( &h->cabac, &h->cabac_state[27+5] ) << 1;
        bits|= get_cabac( &h->cabac, &h->cabac_state[27+5] );
        if( bits < 8 )
            return bits + 3; /* B_Bi_16x16 through B_L1_L0_16x8 */
        else if( bits == 13 ) {
3888
            return decode_cabac_intra_mb_type(h, 32, 0) + 23;
3889 3890 3891 3892 3893 3894 3895
        } else if( bits == 14 )
            return 11; /* B_L1_L0_8x16 */
        else if( bits == 15 )
            return 22; /* B_8x8 */

        bits= ( bits<<1 ) | get_cabac( &h->cabac, &h->cabac_state[27+5] );
        return bits - 4; /* B_L0_Bi_* through B_Bi_Bi_* */
3896
    } else {
3897
        /* TODO SI/SP frames? */
3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945
        return -1;
    }
}

static int decode_cabac_mb_skip( H264Context *h) {
    MpegEncContext * const s = &h->s;
    const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
    const int mba_xy = mb_xy - 1;
    const int mbb_xy = mb_xy - s->mb_stride;
    int ctx = 0;

    if( s->mb_x > 0 && !IS_SKIP( s->current_picture.mb_type[mba_xy] ) )
        ctx++;
    if( s->mb_y > 0 && !IS_SKIP( s->current_picture.mb_type[mbb_xy] ) )
        ctx++;

    if( h->slice_type == P_TYPE || h->slice_type == SP_TYPE)
        return get_cabac( &h->cabac, &h->cabac_state[11+ctx] );
    else /* B-frame */
        return get_cabac( &h->cabac, &h->cabac_state[24+ctx] );
}

static int decode_cabac_mb_intra4x4_pred_mode( H264Context *h, int pred_mode ) {
    int mode = 0;

    if( get_cabac( &h->cabac, &h->cabac_state[68] ) )
        return pred_mode;

    if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
        mode += 1;
    if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
        mode += 2;
    if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
        mode += 4;
    if( mode >= pred_mode )
        return mode + 1;
    else
        return mode;
}

static int decode_cabac_mb_chroma_pre_mode( H264Context *h) {
    MpegEncContext * const s = &h->s;
    const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
    const int mba_xy = mb_xy - 1;
    const int mbb_xy = mb_xy - s->mb_stride;

    int ctx = 0;

3946 3947
    /* No need to test for IS_INTRA4x4 and IS_INTRA16x16, as we set chroma_pred_mode_table to 0 */
    if( s->mb_x > 0 && h->chroma_pred_mode_table[mba_xy] != 0 )
3948
        ctx++;
3949 3950

    if( s->mb_y > 0 && h->chroma_pred_mode_table[mbb_xy] != 0 )
3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004
        ctx++;

    if( get_cabac( &h->cabac, &h->cabac_state[64+ctx] ) == 0 )
        return 0;

    if( get_cabac( &h->cabac, &h->cabac_state[64+3] ) == 0 )
        return 1;
    if( get_cabac( &h->cabac, &h->cabac_state[64+3] ) == 0 )
        return 2;
    else
        return 3;
}

static const uint8_t block_idx_x[16] = {
    0, 1, 0, 1, 2, 3, 2, 3, 0, 1, 0, 1, 2, 3, 2, 3
};
static const uint8_t block_idx_y[16] = {
    0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3, 2, 2, 3, 3
};
static const uint8_t block_idx_xy[4][4] = {
    { 0, 2, 8,  10},
    { 1, 3, 9,  11},
    { 4, 6, 12, 14},
    { 5, 7, 13, 15}
};

static int decode_cabac_mb_cbp_luma( H264Context *h) {
    MpegEncContext * const s = &h->s;
    const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;

    int cbp = 0;
    int i8x8;

    h->cbp_table[mb_xy] = 0;  /* FIXME aaahahahah beurk */

    for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
        int mba_xy = -1;
        int mbb_xy = -1;
        int x, y;
        int ctx = 0;

        x = block_idx_x[4*i8x8];
        y = block_idx_y[4*i8x8];

        if( x > 0 )
            mba_xy = mb_xy;
        else if( s->mb_x > 0 )
            mba_xy = mb_xy - 1;

        if( y > 0 )
            mbb_xy = mb_xy;
        else if( s->mb_y > 0 )
            mbb_xy = mb_xy - s->mb_stride;

4005
        /* No need to test for skip as we put 0 for skip block */
4006 4007
        if( mba_xy >= 0 ) {
            int i8x8a = block_idx_xy[(x-1)&0x03][y]/4;
4008
            if( ((h->cbp_table[mba_xy] >> i8x8a)&0x01) == 0 )
4009 4010 4011 4012 4013
                ctx++;
        }

        if( mbb_xy >= 0 ) {
            int i8x8b = block_idx_xy[x][(y-1)&0x03]/4;
4014
            if( ((h->cbp_table[mbb_xy] >> i8x8b)&0x01) == 0 )
4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028
                ctx += 2;
        }

        if( get_cabac( &h->cabac, &h->cabac_state[73 + ctx] ) ) {
            cbp |= 1 << i8x8;
            h->cbp_table[mb_xy] = cbp;  /* FIXME aaahahahah beurk */
        }
    }
    return cbp;
}
static int decode_cabac_mb_cbp_chroma( H264Context *h) {
    int ctx;
    int cbp_a, cbp_b;

Michael Niedermayer's avatar
Michael Niedermayer committed
4029 4030
    cbp_a = (h->left_cbp>>4)&0x03;
    cbp_b = (h-> top_cbp>>4)&0x03;
4031 4032 4033 4034 4035 4036 4037 4038 4039 4040

    ctx = 0;
    if( cbp_a > 0 ) ctx++;
    if( cbp_b > 0 ) ctx += 2;
    if( get_cabac( &h->cabac, &h->cabac_state[77 + ctx] ) == 0 )
        return 0;

    ctx = 4;
    if( cbp_a == 2 ) ctx++;
    if( cbp_b == 2 ) ctx += 2;
Michael Niedermayer's avatar
Michael Niedermayer committed
4041
    return 1 + get_cabac( &h->cabac, &h->cabac_state[77 + ctx] );
4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069
}
static int decode_cabac_mb_dqp( H264Context *h) {
    MpegEncContext * const s = &h->s;
    int mbn_xy;
    int   ctx = 0;
    int   val = 0;

    if( s->mb_x > 0 )
        mbn_xy = s->mb_x + s->mb_y*s->mb_stride - 1;
    else
        mbn_xy = s->mb_width - 1 + (s->mb_y-1)*s->mb_stride;

    if( mbn_xy >= 0 && h->last_qscale_diff != 0 && ( IS_INTRA16x16(s->current_picture.mb_type[mbn_xy] ) || (h->cbp_table[mbn_xy]&0x3f) ) )
        ctx++;

    while( get_cabac( &h->cabac, &h->cabac_state[60 + ctx] ) ) {
        if( ctx < 2 )
            ctx = 2;
        else
            ctx = 3;
        val++;
    }

    if( val&0x01 )
        return (val + 1)/2;
    else
        return -(val + 1)/2;
}
4070
static int decode_cabac_p_mb_sub_type( H264Context *h ) {
4071 4072 4073 4074 4075 4076 4077 4078
    if( get_cabac( &h->cabac, &h->cabac_state[21] ) )
        return 0;   /* 8x8 */
    if( !get_cabac( &h->cabac, &h->cabac_state[22] ) )
        return 1;   /* 8x4 */
    if( get_cabac( &h->cabac, &h->cabac_state[23] ) )
        return 2;   /* 4x8 */
    return 3;       /* 4x4 */
}
4079 4080 4081 4082 4083
static int decode_cabac_b_mb_sub_type( H264Context *h ) {
    int type;
    if( !get_cabac( &h->cabac, &h->cabac_state[36] ) )
        return 0;   /* B_Direct_8x8 */
    if( !get_cabac( &h->cabac, &h->cabac_state[37] ) )
4084
        return 1 + get_cabac( &h->cabac, &h->cabac_state[39] ); /* B_L0_8x8, B_L1_8x8 */
4085 4086 4087 4088 4089 4090 4091 4092 4093 4094
    type = 3;
    if( get_cabac( &h->cabac, &h->cabac_state[38] ) ) {
        if( get_cabac( &h->cabac, &h->cabac_state[39] ) )
            return 11 + get_cabac( &h->cabac, &h->cabac_state[39] ); /* B_L1_4x4, B_Bi_4x4 */
        type += 4;
    }
    type += 2*get_cabac( &h->cabac, &h->cabac_state[39] );
    type +=   get_cabac( &h->cabac, &h->cabac_state[39] );
    return type;
}
4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120

static int decode_cabac_mb_ref( H264Context *h, int list, int n ) {
    int refa = h->ref_cache[list][scan8[n] - 1];
    int refb = h->ref_cache[list][scan8[n] - 8];
    int ref  = 0;
    int ctx  = 0;

    if( refa > 0 )
        ctx++;
    if( refb > 0 )
        ctx += 2;

    while( get_cabac( &h->cabac, &h->cabac_state[54+ctx] ) ) {
        ref++;
        if( ctx < 4 )
            ctx = 4;
        else
            ctx = 5;
    }
    return ref;
}

static int decode_cabac_mb_mvd( H264Context *h, int list, int n, int l ) {
    int amvd = abs( h->mvd_cache[list][scan8[n] - 1][l] ) +
               abs( h->mvd_cache[list][scan8[n] - 8][l] );
    int ctxbase = (l == 0) ? 40 : 47;
Michael Niedermayer's avatar
Michael Niedermayer committed
4121
    int ctx, mvd;
4122 4123 4124 4125 4126 4127 4128 4129

    if( amvd < 3 )
        ctx = 0;
    else if( amvd > 32 )
        ctx = 2;
    else
        ctx = 1;

Michael Niedermayer's avatar
Michael Niedermayer committed
4130 4131 4132 4133 4134
    if(!get_cabac(&h->cabac, &h->cabac_state[ctxbase+ctx]))
        return 0;

    mvd= 1;
    ctx= 3;
4135 4136
    while( mvd < 9 && get_cabac( &h->cabac, &h->cabac_state[ctxbase+ctx] ) ) {
        mvd++;
Michael Niedermayer's avatar
Michael Niedermayer committed
4137
        if( ctx < 6 )
4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151
            ctx++;
    }

    if( mvd >= 9 ) {
        int k = 3;
        while( get_cabac_bypass( &h->cabac ) ) {
            mvd += 1 << k;
            k++;
        }
        while( k-- ) {
            if( get_cabac_bypass( &h->cabac ) )
                mvd += 1 << k;
        }
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
4152 4153
    if( get_cabac_bypass( &h->cabac ) )  return -mvd;
    else                                 return  mvd;
4154 4155
}

Michael Niedermayer's avatar
Michael Niedermayer committed
4156
static int inline get_cabac_cbf_ctx( H264Context *h, int cat, int idx ) {
Michael Niedermayer's avatar
Michael Niedermayer committed
4157
    int nza, nzb;
4158 4159 4160
    int ctx = 0;

    if( cat == 0 ) {
Michael Niedermayer's avatar
Michael Niedermayer committed
4161 4162
        nza = h->left_cbp&0x100;
        nzb = h-> top_cbp&0x100;
4163
    } else if( cat == 1 || cat == 2 ) {
Michael Niedermayer's avatar
Michael Niedermayer committed
4164 4165
        nza = h->non_zero_count_cache[scan8[idx] - 1];
        nzb = h->non_zero_count_cache[scan8[idx] - 8];
4166
    } else if( cat == 3 ) {
Michael Niedermayer's avatar
Michael Niedermayer committed
4167 4168 4169 4170 4171 4172
        nza = (h->left_cbp>>(6+idx))&0x01;
        nzb = (h-> top_cbp>>(6+idx))&0x01;
    } else {
        assert(cat == 4);
        nza = h->non_zero_count_cache[scan8[16+idx] - 1];
        nzb = h->non_zero_count_cache[scan8[16+idx] - 8];
4173 4174
    }

Michael Niedermayer's avatar
Michael Niedermayer committed
4175
    if( nza > 0 )
4176 4177
        ctx++;

Michael Niedermayer's avatar
Michael Niedermayer committed
4178
    if( nzb > 0 )
4179 4180 4181 4182 4183
        ctx += 2;

    return ctx + 4 * cat;
}

Michael Niedermayer's avatar
Michael Niedermayer committed
4184
static int inline decode_cabac_residual( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, int qp, int max_coeff) {
4185 4186 4187
    const int mb_xy  = h->s.mb_x + h->s.mb_y*h->s.mb_stride;
    const uint16_t *qmul= dequant_coeff[qp];
    static const int significant_coeff_flag_offset[5] = { 0, 15, 29, 44, 47 };
4188
    static const int coeff_abs_level_m1_offset[5] = {227+ 0, 227+10, 227+20, 227+30, 227+39 };
4189

4190
    int index[16];
4191

4192
    int i, last;
4193 4194
    int coeff_count = 0;

Michael Niedermayer's avatar
Michael Niedermayer committed
4195
    int abslevel1 = 1;
4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214
    int abslevelgt1 = 0;

    /* cat: 0-> DC 16x16  n = 0
     *      1-> AC 16x16  n = luma4x4idx
     *      2-> Luma4x4   n = luma4x4idx
     *      3-> DC Chroma n = iCbCr
     *      4-> AC Chroma n = 4 * iCbCr + chroma4x4idx
     */

    /* read coded block flag */
    if( get_cabac( &h->cabac, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n ) ] ) == 0 ) {
        if( cat == 1 || cat == 2 )
            h->non_zero_count_cache[scan8[n]] = 0;
        else if( cat == 4 )
            h->non_zero_count_cache[scan8[16+n]] = 0;

        return 0;
    }

4215 4216 4217 4218 4219
    for(last= 0; last < max_coeff - 1; last++) {
        if( get_cabac( &h->cabac, &h->cabac_state[105+significant_coeff_flag_offset[cat]+last] )) {
            index[coeff_count++] = last;
            if( get_cabac( &h->cabac, &h->cabac_state[166+significant_coeff_flag_offset[cat]+last] ) ) {
                last= max_coeff;
4220 4221 4222 4223 4224
                break;
            }
        }
    }
    if( last == max_coeff -1 ) {
4225
        index[coeff_count++] = last;
4226
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
4227
    assert(coeff_count > 0);
4228

Michael Niedermayer's avatar
Michael Niedermayer committed
4229
    if( cat == 0 )
4230 4231 4232
        h->cbp_table[mb_xy] |= 0x100;
    else if( cat == 1 || cat == 2 )
        h->non_zero_count_cache[scan8[n]] = coeff_count;
Michael Niedermayer's avatar
Michael Niedermayer committed
4233
    else if( cat == 3 )
4234
        h->cbp_table[mb_xy] |= 0x40 << n;
Michael Niedermayer's avatar
Michael Niedermayer committed
4235 4236
    else {
        assert( cat == 4 );
4237
        h->non_zero_count_cache[scan8[16+n]] = coeff_count;
Michael Niedermayer's avatar
Michael Niedermayer committed
4238
    }
4239 4240

    for( i = coeff_count - 1; i >= 0; i-- ) {
Michael Niedermayer's avatar
Michael Niedermayer committed
4241 4242
        int ctx = (abslevelgt1 != 0 ? 0 : FFMIN( 4, abslevel1 )) + coeff_abs_level_m1_offset[cat];
        int j= scantable[index[i]];
4243

4244
        if( get_cabac( &h->cabac, &h->cabac_state[ctx] ) == 0 ) {
Michael Niedermayer's avatar
Michael Niedermayer committed
4245 4246 4247 4248 4249 4250 4251
            if( cat == 0 || cat == 3 ) {
                if( get_cabac_bypass( &h->cabac ) ) block[j] = -1;
                else                                block[j] =  1;
            }else{
                if( get_cabac_bypass( &h->cabac ) ) block[j] = -qmul[j];
                else                                block[j] =  qmul[j];
            }
4252 4253
    
            abslevel1++;
4254
        } else {
4255
            int coeff_abs = 2;
4256
            ctx = 5 + FFMIN( 4, abslevelgt1 ) + coeff_abs_level_m1_offset[cat];
4257 4258
            while( coeff_abs < 15 && get_cabac( &h->cabac, &h->cabac_state[ctx] ) ) {
                coeff_abs++;
4259 4260
            }

4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271
            if( coeff_abs >= 15 ) {
                int j = 0;
                while( get_cabac_bypass( &h->cabac ) ) {
                    coeff_abs += 1 << j;
                    j++;
                }
    
                while( j-- ) {
                    if( get_cabac_bypass( &h->cabac ) )
                        coeff_abs += 1 << j ;
                }
4272 4273
            }

Michael Niedermayer's avatar
Michael Niedermayer committed
4274 4275 4276 4277 4278 4279 4280
            if( cat == 0 || cat == 3 ) {
                if( get_cabac_bypass( &h->cabac ) ) block[j] = -coeff_abs;
                else                                block[j] =  coeff_abs;
            }else{
                if( get_cabac_bypass( &h->cabac ) ) block[j] = -coeff_abs * qmul[j];
                else                                block[j] =  coeff_abs * qmul[j];
            }
4281
    
4282
            abslevelgt1++;
4283
        }
4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308
    }
    return 0;
}

/**
 * decodes a macroblock
 * @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
 */
static int decode_mb_cabac(H264Context *h) {
    MpegEncContext * const s = &h->s;
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
    int mb_type, partition_count, cbp = 0;

    s->dsp.clear_blocks(h->mb); //FIXME avoid if allready clear (move after skip handlong?)

    if( h->sps.mb_aff ) {
        av_log( h->s.avctx, AV_LOG_ERROR, "Fields not supported with CABAC\n" );
        return -1;
    }

    if( h->slice_type != I_TYPE && h->slice_type != SI_TYPE ) {
        /* read skip flags */
        if( decode_cabac_mb_skip( h ) ) {
            int mx, my;

4309
//FIXME b frame
4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325
            /* skip mb */
            mb_type= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P1L0|MB_TYPE_SKIP;

            memset(h->non_zero_count[mb_xy], 0, 16);
            memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui
#if 0
            if(h->sps.mb_aff && s->mb_skip_run==0 && (s->mb_y&1)==0){
                h->mb_field_decoding_flag= get_bits1(&s->gb);
            }
            if(h->mb_field_decoding_flag)
                mb_type|= MB_TYPE_INTERLACED;
#endif

            fill_caches(h, mb_type); //FIXME check what is needed and what not ...
            pred_pskip_motion(h, &mx, &my);
            fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, 0, 1);
4326
            fill_rectangle(  h->mvd_cache[0][scan8[0]], 4, 4, 8, pack16to32(0,0), 4);
4327 4328 4329 4330 4331 4332 4333
            fill_rectangle(  h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mx,my), 4);
            write_back_motion(h, mb_type);

            s->current_picture.mb_type[mb_xy]= mb_type; //FIXME SKIP type
            s->current_picture.qscale_table[mb_xy]= s->qscale;
            h->slice_table[ mb_xy ]= h->slice_num;
            h->cbp_table[mb_xy] = 0;
4334
            h->chroma_pred_mode_table[mb_xy] = 0;
4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349
            h->last_qscale_diff = 0;

            h->prev_mb_skiped= 1;

            return 0;

        }
    }
    h->prev_mb_skiped = 0;

    if( ( mb_type = decode_cabac_mb_type( h ) ) < 0 ) {
        av_log( h->s.avctx, AV_LOG_ERROR, "decode_cabac_mb_type failed\n" );
        return -1;
    }

4350 4351 4352 4353 4354 4355 4356 4357 4358
    if( h->slice_type == B_TYPE ) {
        if( mb_type < 23 ){
            partition_count= b_mb_type_info[mb_type].partition_count;
            mb_type=         b_mb_type_info[mb_type].type;
        }else{
            mb_type -= 23;
            goto decode_intra_mb;
        }
    } else if( h->slice_type == P_TYPE ) {
4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383
        if( mb_type < 5) {
            partition_count= p_mb_type_info[mb_type].partition_count;
            mb_type=         p_mb_type_info[mb_type].type;
        } else {
            mb_type -= 5;
            goto decode_intra_mb;
        }
    } else {
       assert(h->slice_type == I_TYPE);
decode_intra_mb:
        partition_count = 0;
        cbp= i_mb_type_info[mb_type].cbp;
        h->intra16x16_pred_mode= i_mb_type_info[mb_type].pred_mode;
        mb_type= i_mb_type_info[mb_type].type;
    }
#if 0
    if(h->mb_field_decoding_flag)
        mb_type |= MB_TYPE_INTERLACED;
#endif

    s->current_picture.mb_type[mb_xy]= mb_type;
    h->slice_table[ mb_xy ]= h->slice_num;

    if(IS_INTRA_PCM(mb_type)) {
        /* TODO */
Michael Niedermayer's avatar
Michael Niedermayer committed
4384 4385 4386
        assert(0);
        h->cbp_table[mb_xy] = 0xf +4*2; //FIXME ?!
        h->cbp_table[mb_xy] |= 0x1C0;
4387
        h->chroma_pred_mode_table[mb_xy] = 0;
4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414
        s->current_picture.qscale_table[mb_xy]= s->qscale;
        return -1;
    }

    fill_caches(h, mb_type);

    if( IS_INTRA( mb_type ) ) {
        if( IS_INTRA4x4( mb_type ) ) {
            int i;
            for( i = 0; i < 16; i++ ) {
                int pred = pred_intra_mode( h, i );
                h->intra4x4_pred_mode_cache[ scan8[i] ] = decode_cabac_mb_intra4x4_pred_mode( h, pred );

                //av_log( s->avctx, AV_LOG_ERROR, "i4x4 pred=%d mode=%d\n", pred, h->intra4x4_pred_mode_cache[ scan8[i] ] );
            }
            write_back_intra_pred_mode(h);
            if( check_intra4x4_pred_mode(h) < 0 ) return -1;
        } else {
            h->intra16x16_pred_mode= check_intra_pred_mode( h, h->intra16x16_pred_mode );
            if( h->intra16x16_pred_mode < 0 ) return -1;
        }
        h->chroma_pred_mode_table[mb_xy] =
            h->chroma_pred_mode          = decode_cabac_mb_chroma_pre_mode( h );

        h->chroma_pred_mode= check_intra_pred_mode( h, h->chroma_pred_mode );
        if( h->chroma_pred_mode < 0 ) return -1;
    } else if( partition_count == 4 ) {
4415 4416
        int i, j, sub_partition_count[4], list, ref[2][4];

4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428
        if( h->slice_type == B_TYPE ) {
            for( i = 0; i < 4; i++ ) {
                h->sub_mb_type[i] = decode_cabac_b_mb_sub_type( h );
                sub_partition_count[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
                h->sub_mb_type[i]=      b_sub_mb_type_info[ h->sub_mb_type[i] ].type;
            }
        } else {
            for( i = 0; i < 4; i++ ) {
                h->sub_mb_type[i] = decode_cabac_p_mb_sub_type( h );
                sub_partition_count[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
                h->sub_mb_type[i]=      p_sub_mb_type_info[ h->sub_mb_type[i] ].type;
            }
4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441
        }

        for( list = 0; list < 2; list++ ) {
            if( h->ref_count[list] > 0 ) {
                for( i = 0; i < 4; i++ ) {
                    if(IS_DIR(h->sub_mb_type[i], 0, list) && !IS_DIRECT(h->sub_mb_type[i])){
                        if( h->ref_count[list] > 1 )
                            ref[list][i] = decode_cabac_mb_ref( h, list, 4*i );
                        else
                            ref[list][i] = 0;
                    } else {
                        ref[list][i] = -1;
                    }
Michael Niedermayer's avatar
Michael Niedermayer committed
4442
                                                       h->ref_cache[list][ scan8[4*i]+1 ]=
4443 4444 4445 4446 4447 4448 4449
                    h->ref_cache[list][ scan8[4*i]+8 ]=h->ref_cache[list][ scan8[4*i]+9 ]= ref[list][i];
                }
            }
        }

        for(list=0; list<2; list++){
            for(i=0; i<4; i++){
Michael Niedermayer's avatar
Michael Niedermayer committed
4450
                h->ref_cache[list][ scan8[4*i]   ]=h->ref_cache[list][ scan8[4*i]+1 ];
4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505

                if(IS_DIR(h->sub_mb_type[i], 0, list) && !IS_DIRECT(h->sub_mb_type[i])){
                    const int sub_mb_type= h->sub_mb_type[i];
                    const int block_width= (sub_mb_type & (MB_TYPE_16x16|MB_TYPE_16x8)) ? 2 : 1;
                    for(j=0; j<sub_partition_count[i]; j++){
                        int mpx, mpy;
                        int mx, my;
                        const int index= 4*i + block_width*j;
                        int16_t (* mv_cache)[2]= &h->mv_cache[list][ scan8[index] ];
                        int16_t (* mvd_cache)[2]= &h->mvd_cache[list][ scan8[index] ];
                        pred_motion(h, index, block_width, list, h->ref_cache[list][ scan8[index] ], &mpx, &mpy);

                        mx = mpx + decode_cabac_mb_mvd( h, list, index, 0 );
                        my = mpy + decode_cabac_mb_mvd( h, list, index, 1 );
                        tprintf("final mv:%d %d\n", mx, my);

                        if(IS_SUB_8X8(sub_mb_type)){
                            mv_cache[ 0 ][0]= mv_cache[ 1 ][0]=
                            mv_cache[ 8 ][0]= mv_cache[ 9 ][0]= mx;
                            mv_cache[ 0 ][1]= mv_cache[ 1 ][1]=
                            mv_cache[ 8 ][1]= mv_cache[ 9 ][1]= my;

                            mvd_cache[ 0 ][0]= mvd_cache[ 1 ][0]=
                            mvd_cache[ 8 ][0]= mvd_cache[ 9 ][0]= mx - mpx;
                            mvd_cache[ 0 ][1]= mvd_cache[ 1 ][1]=
                            mvd_cache[ 8 ][1]= mvd_cache[ 9 ][1]= my - mpy;
                        }else if(IS_SUB_8X4(sub_mb_type)){
                            mv_cache[ 0 ][0]= mv_cache[ 1 ][0]= mx;
                            mv_cache[ 0 ][1]= mv_cache[ 1 ][1]= my;

                            mvd_cache[ 0 ][0]= mvd_cache[ 1 ][0]= mx- mpx;
                            mvd_cache[ 0 ][1]= mvd_cache[ 1 ][1]= my - mpy;
                        }else if(IS_SUB_4X8(sub_mb_type)){
                            mv_cache[ 0 ][0]= mv_cache[ 8 ][0]= mx;
                            mv_cache[ 0 ][1]= mv_cache[ 8 ][1]= my;

                            mvd_cache[ 0 ][0]= mvd_cache[ 8 ][0]= mx - mpx;
                            mvd_cache[ 0 ][1]= mvd_cache[ 8 ][1]= my - mpy;
                        }else{
                            assert(IS_SUB_4X4(sub_mb_type));
                            mv_cache[ 0 ][0]= mx;
                            mv_cache[ 0 ][1]= my;

                            mvd_cache[ 0 ][0]= mx - mpx;
                            mvd_cache[ 0 ][1]= my - mpy;
                        }
                    }
                }else{
                    uint32_t *p= (uint32_t *)&h->mv_cache[list][ scan8[4*i] ][0];
                    uint32_t *pd= (uint32_t *)&h->mvd_cache[list][ scan8[4*i] ][0];
                    p[0] = p[1] = p[8] = p[9] = 0;
                    pd[0]= pd[1]= pd[8]= pd[9]= 0;
                }
            }
        }
4506
    } else if( !IS_DIRECT(mb_type) ) {
4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579
        int list, mx, my, i, mpx, mpy;
        if(IS_16X16(mb_type)){
            for(list=0; list<2; list++){
                if(IS_DIR(mb_type, 0, list)){
                    if(h->ref_count[list] > 0 ){
                        const int ref = h->ref_count[list] > 1 ? decode_cabac_mb_ref( h, list, 0 ) : 0;
                        fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, ref, 1);
                    }
                }
            }
            for(list=0; list<2; list++){
                if(IS_DIR(mb_type, 0, list)){
                    pred_motion(h, 0, 4, list, h->ref_cache[list][ scan8[0] ], &mpx, &mpy);

                    mx = mpx + decode_cabac_mb_mvd( h, list, 0, 0 );
                    my = mpy + decode_cabac_mb_mvd( h, list, 0, 1 );
                    tprintf("final mv:%d %d\n", mx, my);

                    fill_rectangle(h->mvd_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx-mpx,my-mpy), 4);
                    fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
                }
            }
        }
        else if(IS_16X8(mb_type)){
            for(list=0; list<2; list++){
                if(h->ref_count[list]>0){
                    for(i=0; i<2; i++){
                        if(IS_DIR(mb_type, i, list)){
                            const int ref= h->ref_count[list] > 1 ? decode_cabac_mb_ref( h, list, 8*i ) : 0;
                            fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, ref, 1);
                        }
                    }
                }
            }
            for(list=0; list<2; list++){
                for(i=0; i<2; i++){
                    if(IS_DIR(mb_type, i, list)){
                        pred_16x8_motion(h, 8*i, list, h->ref_cache[list][scan8[0] + 16*i], &mpx, &mpy);
                        mx = mpx + decode_cabac_mb_mvd( h, list, 8*i, 0 );
                        my = mpy + decode_cabac_mb_mvd( h, list, 8*i, 1 );
                        tprintf("final mv:%d %d\n", mx, my);

                        fill_rectangle(h->mvd_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx-mpx,my-mpy), 4);
                        fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx,my), 4);
                    }
                }
            }
        }else{
            assert(IS_8X16(mb_type));
            for(list=0; list<2; list++){
                if(h->ref_count[list]>0){
                    for(i=0; i<2; i++){
                        if(IS_DIR(mb_type, i, list)){ //FIXME optimize
                            const int ref= h->ref_count[list] > 1 ? decode_cabac_mb_ref( h, list, 4*i ) : 0;
                            fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, ref, 1);
                        }
                    }
                }
            }
            for(list=0; list<2; list++){
                for(i=0; i<2; i++){
                    if(IS_DIR(mb_type, i, list)){
                        pred_8x16_motion(h, i*4, list, h->ref_cache[list][ scan8[0] + 2*i ], &mpx, &mpy);
                        mx = mpx + decode_cabac_mb_mvd( h, list, 4*i, 0 );
                        my = mpy + decode_cabac_mb_mvd( h, list, 4*i, 1 );

                        tprintf("final mv:%d %d\n", mx, my);
                        fill_rectangle(h->mvd_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx-mpx,my-mpy), 4);
                        fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx,my), 4);
                    }
                }
            }
        }
4580 4581
    }

4582 4583
   if( IS_INTER( mb_type ) ) {
        h->chroma_pred_mode_table[mb_xy] = 0;
4584
        write_back_motion( h, mb_type );
4585
   }
4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669

    if( !IS_INTRA16x16( mb_type ) ) {
        cbp  = decode_cabac_mb_cbp_luma( h );
        cbp |= decode_cabac_mb_cbp_chroma( h ) << 4;
    }

    h->cbp_table[mb_xy] = cbp;

    if( cbp || IS_INTRA16x16( mb_type ) ) {
        const uint8_t *scan, *dc_scan;
        int dqp;

        if(IS_INTERLACED(mb_type)){
            scan= field_scan;
            dc_scan= luma_dc_field_scan;
        }else{
            scan= zigzag_scan;
            dc_scan= luma_dc_zigzag_scan;
        }

        h->last_qscale_diff = dqp = decode_cabac_mb_dqp( h );
        s->qscale += dqp;
        if(((unsigned)s->qscale) > 51){
            if(s->qscale<0) s->qscale+= 52;
            else            s->qscale-= 52;
        }
        h->chroma_qp = get_chroma_qp(h, s->qscale);

        if( IS_INTRA16x16( mb_type ) ) {
            int i;
            //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 DC\n" );
            if( decode_cabac_residual( h, h->mb, 0, 0, dc_scan, s->qscale, 16) < 0)
                return -1;
            if( cbp&15 ) {
                for( i = 0; i < 16; i++ ) {
                    //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 AC:%d\n", i );
                    if( decode_cabac_residual(h, h->mb + 16*i, 1, i, scan + 1, s->qscale, 15) < 0 )
                        return -1;
                }
            } else {
                fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1);
            }
        } else {
            int i8x8, i4x4;
            for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
                if( cbp & (1<<i8x8) ) {
                    for( i4x4 = 0; i4x4 < 4; i4x4++ ) {
                        const int index = 4*i8x8 + i4x4;
                        //av_log( s->avctx, AV_LOG_ERROR, "Luma4x4: %d\n", index );
                        if( decode_cabac_residual(h, h->mb + 16*index, 2, index, scan, s->qscale, 16) < 0 )
                            return -1;
                    }
                } else {
                    uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ];
                    nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;
                }
            }
        }

        if( cbp&0x30 ){
            int c;
            for( c = 0; c < 2; c++ ) {
                //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-DC\n",c );
                if( decode_cabac_residual(h, h->mb + 256 + 16*4*c, 3, c, chroma_dc_scan, h->chroma_qp, 4) < 0)
                    return -1;
            }
        }

        if( cbp&0x20 ) {
            int c, i;
            for( c = 0; c < 2; c++ ) {
                for( i = 0; i < 4; i++ ) {
                    const int index = 16 + 4 * c + i;
                    //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16 );
                    if( decode_cabac_residual(h, h->mb + 16*index, 4, index - 16, scan + 1, h->chroma_qp, 15) < 0)
                        return -1;
                }
            }
        } else {
            uint8_t * const nnz= &h->non_zero_count_cache[0];
            nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
            nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
        }
    } else {
4670 4671 4672 4673
        uint8_t * const nnz= &h->non_zero_count_cache[0];
        fill_rectangle(&nnz[scan8[0]], 4, 4, 8, 0, 1);
        nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
        nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
4674 4675 4676 4677 4678 4679 4680 4681 4682
    }

    s->current_picture.qscale_table[mb_xy]= s->qscale;
    write_back_non_zero_count(h);

    return 0;
}


4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694
static void filter_mb_edgev( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
    int i, d;
    const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];

    for( i = 0; i < 4; i++ ) {
        if( bS[i] == 0 ) {
            pix += 4 * stride;
            continue;
        }

4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719
        if( bS[i] < 4 ) {
            const int tc0 = tc0_table[index_a][bS[i] - 1];
            /* 4px edge length */
            for( d = 0; d < 4; d++ ) {
                const int p0 = pix[-1];
                const int p1 = pix[-2];
                const int p2 = pix[-3];
                const int q0 = pix[0];
                const int q1 = pix[1];
                const int q2 = pix[2];

                if( ABS( p0 - q0 ) < alpha &&
                    ABS( p1 - p0 ) < beta &&
                    ABS( q1 - q0 ) < beta ) {
                    int tc = tc0;
                    int i_delta;

                    if( ABS( p2 - p0 ) < beta ) {
                        pix[-2] = p1 + clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
                        tc++;
                    }
                    if( ABS( q2 - q0 ) < beta ) {
                        pix[1] = q1 + clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
                        tc++;
                    }
4720

4721 4722 4723
                    i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
                    pix[-1] = clip_uint8( p0 + i_delta );    /* p0' */
                    pix[0]  = clip_uint8( q0 - i_delta );    /* q0' */
4724
                }
4725
                pix += stride;
4726
            }
4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769
        }else{
            /* 4px edge length */
            for( d = 0; d < 4; d++ ) {
                const int p0 = pix[-1];
                const int p1 = pix[-2];
                const int p2 = pix[-3];

                const int q0 = pix[0];
                const int q1 = pix[1];
                const int q2 = pix[2];

                if( ABS( p0 - q0 ) < alpha &&
                    ABS( p1 - p0 ) < beta &&
                    ABS( q1 - q0 ) < beta ) {

                    if(ABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
                        if( ABS( p2 - p0 ) < beta)
                        {
                            const int p3 = pix[-4];
                            /* p0', p1', p2' */
                            pix[-1] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
                            pix[-2] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
                            pix[-3] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
                        } else {
                            /* p0' */
                            pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                        }
                        if( ABS( q2 - q0 ) < beta)
                        {
                            const int q3 = pix[3];
                            /* q0', q1', q2' */
                            pix[0] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
                            pix[1] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
                            pix[2] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
                        } else {
                            /* q0' */
                            pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                        }
                    }else{
                        /* p0', q0' */
                        pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                        pix[ 0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                    }
4770
                }
4771
                pix += stride;
4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787
            }
        }
    }
}
static void filter_mb_edgecv( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
    int i, d;
    const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];

    for( i = 0; i < 4; i++ ) {
        if( bS[i] == 0 ) {
            pix += 2 * stride;
            continue;
        }

4788 4789 4790 4791 4792 4793 4794 4795
        if( bS[i] < 4 ) {
            const int tc = tc0_table[index_a][bS[i] - 1] + 1;
            /* 2px edge length (because we use same bS than the one for luma) */
            for( d = 0; d < 2; d++ ){
                const int p0 = pix[-1];
                const int p1 = pix[-2];
                const int q0 = pix[0];
                const int q1 = pix[1];
4796

4797 4798 4799 4800 4801 4802 4803 4804
                if( ABS( p0 - q0 ) < alpha &&
                    ABS( p1 - p0 ) < beta &&
                    ABS( q1 - q0 ) < beta ) {
                    const int i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );

                    pix[-1] = clip_uint8( p0 + i_delta );    /* p0' */
                    pix[0]  = clip_uint8( q0 - i_delta );    /* q0' */
                }
4805 4806
                pix += stride;
            }
4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822
        }else{
            /* 2px edge length (because we use same bS than the one for luma) */
            for( d = 0; d < 2; d++ ){
                const int p0 = pix[-1];
                const int p1 = pix[-2];
                const int q0 = pix[0];
                const int q1 = pix[1];

                if( ABS( p0 - q0 ) < alpha &&
                    ABS( p1 - p0 ) < beta &&
                    ABS( q1 - q0 ) < beta ) {

                    pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
                    pix[0]  = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
                }
                pix += stride;
4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840
            }
        }
    }
}

static void filter_mb_edgeh( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
    int i, d;
    const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
    const int pix_next  = stride;

    for( i = 0; i < 4; i++ ) {
        if( bS[i] == 0 ) {
            pix += 4;
            continue;
        }

4841 4842 4843 4844 4845 4846 4847 4848 4849 4850
        if( bS[i] < 4 ) {
            const int tc0 = tc0_table[index_a][bS[i] - 1];
            /* 4px edge length */
            for( d = 0; d < 4; d++ ) {
                const int p0 = pix[-1*pix_next];
                const int p1 = pix[-2*pix_next];
                const int p2 = pix[-3*pix_next];
                const int q0 = pix[0];
                const int q1 = pix[1*pix_next];
                const int q2 = pix[2*pix_next];
4851

4852 4853 4854
                if( ABS( p0 - q0 ) < alpha &&
                    ABS( p1 - p0 ) < beta &&
                    ABS( q1 - q0 ) < beta ) {
4855

4856 4857
                    int tc = tc0;
                    int i_delta;
4858

4859 4860 4861 4862 4863 4864 4865 4866
                    if( ABS( p2 - p0 ) < beta ) {
                        pix[-2*pix_next] = p1 + clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
                        tc++;
                    }
                    if( ABS( q2 - q0 ) < beta ) {
                        pix[pix_next] = q1 + clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
                        tc++;
                    }
4867

4868 4869 4870
                    i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
                    pix[-pix_next] = clip_uint8( p0 + i_delta );    /* p0' */
                    pix[0]         = clip_uint8( q0 - i_delta );    /* q0' */
4871
                }
4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914
                pix++;
            }
        }else{
            /* 4px edge length */
            for( d = 0; d < 4; d++ ) {
                const int p0 = pix[-1*pix_next];
                const int p1 = pix[-2*pix_next];
                const int p2 = pix[-3*pix_next];
                const int q0 = pix[0];
                const int q1 = pix[1*pix_next];
                const int q2 = pix[2*pix_next];

                if( ABS( p0 - q0 ) < alpha &&
                    ABS( p1 - p0 ) < beta &&
                    ABS( q1 - q0 ) < beta ) {

                    const int p3 = pix[-4*pix_next];
                    const int q3 = pix[ 3*pix_next];

                    if(ABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
                        if( ABS( p2 - p0 ) < beta) {
                            /* p0', p1', p2' */
                            pix[-1*pix_next] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
                            pix[-2*pix_next] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
                            pix[-3*pix_next] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
                        } else {
                            /* p0' */
                            pix[-1*pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                        }
                        if( ABS( q2 - q0 ) < beta) {
                            /* q0', q1', q2' */
                            pix[0*pix_next] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
                            pix[1*pix_next] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
                            pix[2*pix_next] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
                        } else {
                            /* q0' */
                            pix[0*pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                        }
                    }else{
                        /* p0', q0' */
                        pix[-1*pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                        pix[ 0*pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                    }
4915
                }
4916
                pix++;
4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935
            }
        }
    }
}

static void filter_mb_edgech( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
    int i, d;
    const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
    const int pix_next  = stride;

    for( i = 0; i < 4; i++ )
    {
        if( bS[i] == 0 ) {
            pix += 2;
            continue;
        }

4936 4937 4938 4939 4940 4941 4942 4943
        if( bS[i] < 4 ) {
            int tc = tc0_table[index_a][bS[i] - 1] + 1;
            /* 2px edge length (see deblocking_filter_edgecv) */
            for( d = 0; d < 2; d++ ) {
                const int p0 = pix[-1*pix_next];
                const int p1 = pix[-2*pix_next];
                const int q0 = pix[0];
                const int q1 = pix[1*pix_next];
4944

4945 4946 4947
                if( ABS( p0 - q0 ) < alpha &&
                    ABS( p1 - p0 ) < beta &&
                    ABS( q1 - q0 ) < beta ) {
4948

4949
                    int i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
4950

4951 4952 4953 4954
                    pix[-pix_next] = clip_uint8( p0 + i_delta );    /* p0' */
                    pix[0]         = clip_uint8( q0 - i_delta );    /* q0' */
                }
                pix++;
4955
            }
4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971
        }else{
            /* 2px edge length (see deblocking_filter_edgecv) */
            for( d = 0; d < 2; d++ ) {
                const int p0 = pix[-1*pix_next];
                const int p1 = pix[-2*pix_next];
                const int q0 = pix[0];
                const int q1 = pix[1*pix_next];

                if( ABS( p0 - q0 ) < alpha &&
                    ABS( p1 - p0 ) < beta &&
                    ABS( q1 - q0 ) < beta ) {

                    pix[-pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
                    pix[0]         = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
                }
                pix++;
4972 4973 4974 4975 4976
            }
        }
    }
}

4977
static void filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr) {
4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000
    MpegEncContext * const s = &h->s;
    const int mb_xy= mb_x + mb_y*s->mb_stride;
    int linesize, uvlinesize;
    int dir;

    /* FIXME Implement deblocking filter for field MB */
    if( h->sps.mb_aff ) {
        return;
    }
    linesize = s->linesize;
    uvlinesize = s->uvlinesize;

    /* dir : 0 -> vertical edge, 1 -> horizontal edge */
    for( dir = 0; dir < 2; dir++ )
    {
        int start = 0;
        int edge;

        /* test picture boundary */
        if( ( dir == 0 && mb_x == 0 ) || ( dir == 1 && mb_y == 0 ) ) {
            start = 1;
        }
        /* FIXME test slice boundary */
5001
        if( h->deblocking_filter == 2 ) {
5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018
        }

        /* Calculate bS */
        for( edge = start; edge < 4; edge++ ) {
            /* mbn_xy: neighbour macroblock (how that works for field ?) */
            int mbn_xy = edge > 0 ? mb_xy : ( dir == 0 ? mb_xy -1 : mb_xy - s->mb_stride );
            int bS[4];
            int qp;

            if( IS_INTRA( s->current_picture.mb_type[mb_xy] ) ||
                IS_INTRA( s->current_picture.mb_type[mbn_xy] ) ) {
                bS[0] = bS[1] = bS[2] = bS[3] = ( edge == 0 ? 4 : 3 );
            } else {
                int i;
                for( i = 0; i < 4; i++ ) {
                    int x = dir == 0 ? edge : i;
                    int y = dir == 0 ? i    : edge;
5019 5020
                    int b_idx= 8 + 4 + x + 8*y;
                    int bn_idx= b_idx - (dir ? 8:1);
5021

5022 5023
                    if( h->non_zero_count_cache[b_idx] != 0 ||
                        h->non_zero_count_cache[bn_idx] != 0 ) {
5024 5025 5026
                        bS[i] = 2;
                    }
                    else if( h->slice_type == P_TYPE ) {
5027 5028 5029
                        if( h->ref_cache[0][b_idx] != h->ref_cache[0][bn_idx] ||
                            ABS( h->mv_cache[0][b_idx][0] - h->mv_cache[0][bn_idx][0] ) >= 4 ||
                            ABS( h->mv_cache[0][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= 4 )
5030 5031 5032 5033 5034 5035 5036 5037 5038
                            bS[i] = 1;
                        else
                            bS[i] = 0;
                    }
                    else {
                        /* FIXME Add support for B frame */
                        return;
                    }
                }
5039 5040 5041

                if(bS[0]+bS[1]+bS[2]+bS[3] == 0)
                    continue;
5042 5043 5044
            }

            /* Filter edge */
5045
            qp = ( s->qscale + s->current_picture.qscale_table[mbn_xy] + 1 ) >> 1;
5046 5047
            if( dir == 0 ) {
                filter_mb_edgev( h, &img_y[4*edge], linesize, bS, qp );
5048
                if( (edge&1) == 0 ) {
5049
                    int chroma_qp = ( h->chroma_qp +
5050 5051 5052 5053 5054 5055
                                      get_chroma_qp( h, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
                    filter_mb_edgecv( h, &img_cb[2*edge], uvlinesize, bS, chroma_qp );
                    filter_mb_edgecv( h, &img_cr[2*edge], uvlinesize, bS, chroma_qp );
                }
            } else {
                filter_mb_edgeh( h, &img_y[4*edge*linesize], linesize, bS, qp );
5056
                if( (edge&1) == 0 ) {
5057
                    int chroma_qp = ( h->chroma_qp +
5058 5059 5060 5061 5062 5063 5064 5065 5066
                                      get_chroma_qp( h, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
                    filter_mb_edgech( h, &img_cb[2*edge*uvlinesize], uvlinesize, bS, chroma_qp );
                    filter_mb_edgech( h, &img_cr[2*edge*uvlinesize], uvlinesize, bS, chroma_qp );
                }
            }
        }
    }
}

5067 5068 5069 5070 5071 5072
static int decode_slice(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int part_mask= s->partitioned_frame ? (AC_END|AC_ERROR) : 0x7F;

    s->mb_skip_run= -1;

5073 5074
    if( h->pps.cabac ) {
        int i;
5075

5076 5077 5078 5079 5080 5081 5082
        /* realign */
        align_get_bits( &s->gb );

        /* init cabac */
        ff_init_cabac_states( &h->cabac, ff_h264_lps_range, ff_h264_mps_state, ff_h264_lps_state, 64 );
        ff_init_cabac_decoder( &h->cabac,
                               s->gb.buffer + get_bits_count(&s->gb)/8,
Michael Niedermayer's avatar
Michael Niedermayer committed
5083
                               ( s->gb.size_in_bits - get_bits_count(&s->gb) + 7)/8);
5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095
        /* calculate pre-state */
        for( i= 0; i < 399; i++ ) {
            int pre;
            if( h->slice_type == I_TYPE )
                pre = clip( ((cabac_context_init_I[i][0] * s->qscale) >>4 ) + cabac_context_init_I[i][1], 1, 126 );
            else
                pre = clip( ((cabac_context_init_PB[h->cabac_init_idc][i][0] * s->qscale) >>4 ) + cabac_context_init_PB[h->cabac_init_idc][i][1], 1, 126 );

            if( pre <= 63 )
                h->cabac_state[i] = 2 * ( 63 - pre ) + 0;
            else
                h->cabac_state[i] = 2 * ( pre - 64 ) + 1;
5096
        }
5097

5098 5099 5100
        for(;;){
            int ret = decode_mb_cabac(h);
            int eos = get_cabac_terminate( &h->cabac ); /* End of Slice flag */
5101

Michael Niedermayer's avatar
Michael Niedermayer committed
5102
            if(ret>=0) hl_decode_mb(h);
5103

5104 5105 5106 5107
            /* XXX: useless as decode_mb_cabac it doesn't support that ... */
            if( ret >= 0 && h->sps.mb_aff ) { //FIXME optimal? or let mb_decode decode 16x32 ?
                s->mb_y++;

Michael Niedermayer's avatar
Michael Niedermayer committed
5108
                if(ret>=0) ret = decode_mb_cabac(h);
5109 5110 5111 5112 5113 5114
                eos = get_cabac_terminate( &h->cabac );

                hl_decode_mb(h);
                s->mb_y--;
            }

Michael Niedermayer's avatar
Michael Niedermayer committed
5115
            if( ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 1) {
5116 5117 5118 5119 5120 5121 5122 5123 5124 5125
                av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
                ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
                return -1;
            }

            if( ++s->mb_x >= s->mb_width ) {
                s->mb_x = 0;
                ff_draw_horiz_band(s, 16*s->mb_y, 16);
                if( ++s->mb_y >= s->mb_height ) {
                    tprintf("slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
5126 5127 5128
                }
            }

5129 5130
            if( eos || s->mb_y >= s->mb_height ) {
                ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
5131
                return 0;
5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145
            }
#if 0
            /* TODO test over-reading in cabac code */
            else if( read too much in h->cabac ) {
                ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
                return -1;
            }
#endif
        }

    } else {
        for(;;){
            int ret = decode_mb_cavlc(h);

Michael Niedermayer's avatar
Michael Niedermayer committed
5146
            if(ret>=0) hl_decode_mb(h);
5147 5148 5149 5150 5151

            if(ret>=0 && h->sps.mb_aff){ //FIXME optimal? or let mb_decode decode 16x32 ?
                s->mb_y++;
                ret = decode_mb_cavlc(h);

Michael Niedermayer's avatar
Michael Niedermayer committed
5152
                if(ret>=0) hl_decode_mb(h);
5153 5154 5155 5156 5157
                s->mb_y--;
            }

            if(ret<0){
                av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
5158 5159 5160 5161
                ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);

                return -1;
            }
5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191

            if(++s->mb_x >= s->mb_width){
                s->mb_x=0;
                ff_draw_horiz_band(s, 16*s->mb_y, 16);
                if(++s->mb_y >= s->mb_height){
                    tprintf("slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);

                    if(get_bits_count(&s->gb) == s->gb.size_in_bits ) {
                        ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);

                        return 0;
                    }else{
                        ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);

                        return -1;
                    }
                }
            }

            if(get_bits_count(&s->gb) >= s->gb.size_in_bits && s->mb_skip_run<=0){
                if(get_bits_count(&s->gb) == s->gb.size_in_bits ){
                    ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);

                    return 0;
                }else{
                    ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);

                    return -1;
                }
            }
5192 5193
        }
    }
5194

5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251
#if 0
    for(;s->mb_y < s->mb_height; s->mb_y++){
        for(;s->mb_x < s->mb_width; s->mb_x++){
            int ret= decode_mb(h);
            
            hl_decode_mb(h);

            if(ret<0){
                fprintf(stderr, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
                ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);

                return -1;
            }
        
            if(++s->mb_x >= s->mb_width){
                s->mb_x=0;
                if(++s->mb_y >= s->mb_height){
                    if(get_bits_count(s->gb) == s->gb.size_in_bits){
                        ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);

                        return 0;
                    }else{
                        ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);

                        return -1;
                    }
                }
            }
        
            if(get_bits_count(s->?gb) >= s->gb?.size_in_bits){
                if(get_bits_count(s->gb) == s->gb.size_in_bits){
                    ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);

                    return 0;
                }else{
                    ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);

                    return -1;
                }
            }
        }
        s->mb_x=0;
        ff_draw_horiz_band(s, 16*s->mb_y, 16);
    }
#endif
    return -1; //not reached
}

static inline int decode_vui_parameters(H264Context *h, SPS *sps){
    MpegEncContext * const s = &h->s;
    int aspect_ratio_info_present_flag, aspect_ratio_idc;

    aspect_ratio_info_present_flag= get_bits1(&s->gb);
    
    if( aspect_ratio_info_present_flag ) {
        aspect_ratio_idc= get_bits(&s->gb, 8);
        if( aspect_ratio_idc == EXTENDED_SAR ) {
Michael Niedermayer's avatar
Michael Niedermayer committed
5252 5253
            sps->sar.num= get_bits(&s->gb, 16);
            sps->sar.den= get_bits(&s->gb, 16);
5254
        }else if(aspect_ratio_idc < 16){
Michael Niedermayer's avatar
Michael Niedermayer committed
5255
            sps->sar=  pixel_aspect[aspect_ratio_idc];
5256
        }else{
5257
            av_log(h->s.avctx, AV_LOG_ERROR, "illegal aspect ratio\n");
5258 5259 5260
            return -1;
        }
    }else{
Michael Niedermayer's avatar
Michael Niedermayer committed
5261 5262
        sps->sar.num= 
        sps->sar.den= 0;
5263 5264
    }
//            s->avctx->aspect_ratio= sar_width*s->width / (float)(s->height*sar_height);
5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291

    if(get_bits1(&s->gb)){      /* overscan_info_present_flag */
        get_bits1(&s->gb);      /* overscan_appropriate_flag */
    }

    if(get_bits1(&s->gb)){      /* video_signal_type_present_flag */
        get_bits(&s->gb, 3);    /* video_format */
        get_bits1(&s->gb);      /* video_full_range_flag */
        if(get_bits1(&s->gb)){  /* colour_description_present_flag */
            get_bits(&s->gb, 8); /* colour_primaries */
            get_bits(&s->gb, 8); /* transfer_characteristics */
            get_bits(&s->gb, 8); /* matrix_coefficients */
        }
    }

    if(get_bits1(&s->gb)){      /* chroma_location_info_present_flag */
        get_ue_golomb(&s->gb);  /* chroma_sample_location_type_top_field */
        get_ue_golomb(&s->gb);  /* chroma_sample_location_type_bottom_field */
    }

    sps->timing_info_present_flag = get_bits1(&s->gb);
    if(sps->timing_info_present_flag){
        sps->num_units_in_tick = get_bits_long(&s->gb, 32);
        sps->time_scale = get_bits_long(&s->gb, 32);
        sps->fixed_frame_rate_flag = get_bits1(&s->gb);
    }

5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319
#if 0
| nal_hrd_parameters_present_flag                   |0  |u(1)    |
| if( nal_hrd_parameters_present_flag  = =  1)      |   |        |
|  hrd_parameters( )                                |   |        |
| vcl_hrd_parameters_present_flag                   |0  |u(1)    |
| if( vcl_hrd_parameters_present_flag  = =  1)      |   |        |
|  hrd_parameters( )                                |   |        |
| if( ( nal_hrd_parameters_present_flag  = =  1  | ||   |        |
|                                                   |   |        |
|( vcl_hrd_parameters_present_flag  = =  1 ) )      |   |        |
|  low_delay_hrd_flag                               |0  |u(1)    |
| bitstream_restriction_flag                        |0  |u(1)    |
| if( bitstream_restriction_flag ) {                |0  |u(1)    |
|  motion_vectors_over_pic_boundaries_flag          |0  |u(1)    |
|  max_bytes_per_pic_denom                          |0  |ue(v)   |
|  max_bits_per_mb_denom                            |0  |ue(v)   |
|  log2_max_mv_length_horizontal                    |0  |ue(v)   |
|  log2_max_mv_length_vertical                      |0  |ue(v)   |
|  num_reorder_frames                               |0  |ue(v)   |
|  max_dec_frame_buffering                          |0  |ue(v)   |
| }                                                 |   |        |
|}                                                  |   |        |
#endif
    return 0;
}

static inline int decode_seq_parameter_set(H264Context *h){
    MpegEncContext * const s = &h->s;
5320
    int profile_idc, level_idc;
5321 5322 5323 5324
    int sps_id, i;
    SPS *sps;
    
    profile_idc= get_bits(&s->gb, 8);
5325 5326 5327
    get_bits1(&s->gb);   //constraint_set0_flag
    get_bits1(&s->gb);   //constraint_set1_flag
    get_bits1(&s->gb);   //constraint_set2_flag
5328 5329
    get_bits1(&s->gb);   //constraint_set3_flag
    get_bits(&s->gb, 4); // reserved
5330 5331 5332 5333 5334 5335
    level_idc= get_bits(&s->gb, 8);
    sps_id= get_ue_golomb(&s->gb);
    
    sps= &h->sps_buffer[ sps_id ];
    sps->profile_idc= profile_idc;
    sps->level_idc= level_idc;
5336

5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351
    sps->log2_max_frame_num= get_ue_golomb(&s->gb) + 4;
    sps->poc_type= get_ue_golomb(&s->gb);
    
    if(sps->poc_type == 0){ //FIXME #define
        sps->log2_max_poc_lsb= get_ue_golomb(&s->gb) + 4;
    } else if(sps->poc_type == 1){//FIXME #define
        sps->delta_pic_order_always_zero_flag= get_bits1(&s->gb);
        sps->offset_for_non_ref_pic= get_se_golomb(&s->gb);
        sps->offset_for_top_to_bottom_field= get_se_golomb(&s->gb);
        sps->poc_cycle_length= get_ue_golomb(&s->gb);
        
        for(i=0; i<sps->poc_cycle_length; i++)
            sps->offset_for_ref_frame[i]= get_se_golomb(&s->gb);
    }
    if(sps->poc_type > 2){
5352
        av_log(h->s.avctx, AV_LOG_ERROR, "illegal POC type %d\n", sps->poc_type);
5353 5354 5355 5356
        return -1;
    }

    sps->ref_frame_count= get_ue_golomb(&s->gb);
5357 5358 5359
    if(sps->ref_frame_count > MAX_PICTURE_COUNT-2){
        av_log(h->s.avctx, AV_LOG_ERROR, "too many reference frames\n");
    }
5360
    sps->gaps_in_frame_num_allowed_flag= get_bits1(&s->gb);
5361 5362 5363 5364 5365 5366 5367 5368 5369 5370
    sps->mb_width= get_ue_golomb(&s->gb) + 1;
    sps->mb_height= get_ue_golomb(&s->gb) + 1;
    sps->frame_mbs_only_flag= get_bits1(&s->gb);
    if(!sps->frame_mbs_only_flag)
        sps->mb_aff= get_bits1(&s->gb);
    else
        sps->mb_aff= 0;

    sps->direct_8x8_inference_flag= get_bits1(&s->gb);

5371 5372 5373 5374 5375 5376 5377
    sps->crop= get_bits1(&s->gb);
    if(sps->crop){
        sps->crop_left  = get_ue_golomb(&s->gb);
        sps->crop_right = get_ue_golomb(&s->gb);
        sps->crop_top   = get_ue_golomb(&s->gb);
        sps->crop_bottom= get_ue_golomb(&s->gb);
        if(sps->crop_left || sps->crop_top){
5378
            av_log(h->s.avctx, AV_LOG_ERROR, "insane cropping not completly supported, this could look slightly wrong ...\n");
5379 5380 5381 5382 5383 5384 5385 5386
        }
    }else{
        sps->crop_left  = 
        sps->crop_right = 
        sps->crop_top   = 
        sps->crop_bottom= 0;
    }

5387 5388 5389 5390 5391
    sps->vui_parameters_present_flag= get_bits1(&s->gb);
    if( sps->vui_parameters_present_flag )
        decode_vui_parameters(h, sps);
    
    if(s->avctx->debug&FF_DEBUG_PICT_INFO){
5392
        av_log(h->s.avctx, AV_LOG_DEBUG, "sps:%d profile:%d/%d poc:%d ref:%d %dx%d %s %s crop:%d/%d/%d/%d %s\n", 
5393 5394 5395 5396 5397 5398
               sps_id, sps->profile_idc, sps->level_idc,
               sps->poc_type,
               sps->ref_frame_count,
               sps->mb_width, sps->mb_height,
               sps->frame_mbs_only_flag ? "FRM" : (sps->mb_aff ? "MB-AFF" : "PIC-AFF"),
               sps->direct_8x8_inference_flag ? "8B8" : "",
5399 5400
               sps->crop_left, sps->crop_right, 
               sps->crop_top, sps->crop_bottom, 
5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417
               sps->vui_parameters_present_flag ? "VUI" : ""
               );
    }
    return 0;
}

static inline int decode_picture_parameter_set(H264Context *h){
    MpegEncContext * const s = &h->s;
    int pps_id= get_ue_golomb(&s->gb);
    PPS *pps= &h->pps_buffer[pps_id];
    
    pps->sps_id= get_ue_golomb(&s->gb);
    pps->cabac= get_bits1(&s->gb);
    pps->pic_order_present= get_bits1(&s->gb);
    pps->slice_group_count= get_ue_golomb(&s->gb) + 1;
    if(pps->slice_group_count > 1 ){
        pps->mb_slice_group_map_type= get_ue_golomb(&s->gb);
5418
        av_log(h->s.avctx, AV_LOG_ERROR, "FMO not supported\n");
5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449
        switch(pps->mb_slice_group_map_type){
        case 0:
#if 0
|   for( i = 0; i <= num_slice_groups_minus1; i++ ) |   |        |
|    run_length[ i ]                                |1  |ue(v)   |
#endif
            break;
        case 2:
#if 0
|   for( i = 0; i < num_slice_groups_minus1; i++ )  |   |        |
|{                                                  |   |        |
|    top_left_mb[ i ]                               |1  |ue(v)   |
|    bottom_right_mb[ i ]                           |1  |ue(v)   |
|   }                                               |   |        |
#endif
            break;
        case 3:
        case 4:
        case 5:
#if 0
|   slice_group_change_direction_flag               |1  |u(1)    |
|   slice_group_change_rate_minus1                  |1  |ue(v)   |
#endif
            break;
        case 6:
#if 0
|   slice_group_id_cnt_minus1                       |1  |ue(v)   |
|   for( i = 0; i <= slice_group_id_cnt_minus1; i++ |   |        |
|)                                                  |   |        |
|    slice_group_id[ i ]                            |1  |u(v)    |
#endif
5450
            break;
5451 5452 5453 5454 5455
        }
    }
    pps->ref_count[0]= get_ue_golomb(&s->gb) + 1;
    pps->ref_count[1]= get_ue_golomb(&s->gb) + 1;
    if(pps->ref_count[0] > 32 || pps->ref_count[1] > 32){
5456
        av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow (pps)\n");
5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469
        return -1;
    }
    
    pps->weighted_pred= get_bits1(&s->gb);
    pps->weighted_bipred_idc= get_bits(&s->gb, 2);
    pps->init_qp= get_se_golomb(&s->gb) + 26;
    pps->init_qs= get_se_golomb(&s->gb) + 26;
    pps->chroma_qp_index_offset= get_se_golomb(&s->gb);
    pps->deblocking_filter_parameters_present= get_bits1(&s->gb);
    pps->constrained_intra_pred= get_bits1(&s->gb);
    pps->redundant_pic_cnt_present = get_bits1(&s->gb);
    
    if(s->avctx->debug&FF_DEBUG_PICT_INFO){
5470
        av_log(h->s.avctx, AV_LOG_DEBUG, "pps:%d sps:%d %s slice_groups:%d ref:%d/%d %s qp:%d/%d/%d %s %s %s\n", 
5471 5472 5473 5474 5475 5476 5477 5478
               pps_id, pps->sps_id,
               pps->cabac ? "CABAC" : "CAVLC",
               pps->slice_group_count,
               pps->ref_count[0], pps->ref_count[1],
               pps->weighted_pred ? "weighted" : "",
               pps->init_qp, pps->init_qs, pps->chroma_qp_index_offset,
               pps->deblocking_filter_parameters_present ? "LPAR" : "",
               pps->constrained_intra_pred ? "CONSTR" : "",
5479
               pps->redundant_pic_cnt_present ? "REDU" : ""
5480 5481 5482 5483 5484 5485 5486 5487 5488 5489
               );
    }
    
    return 0;
}

/**
 * finds the end of the current frame in the bitstream.
 * @return the position of the first byte of the next frame, or -1
 */
5490
static int find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size){
5491
    int i;
5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509
    uint32_t state;
//printf("first %02X%02X%02X%02X\n", buf[0], buf[1],buf[2],buf[3]);
//    mb_addr= pc->mb_addr - 1;
    state= pc->state;
    //FIXME this will fail with slices
    for(i=0; i<buf_size; i++){
        state= (state<<8) | buf[i];
        if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
            if(pc->frame_start_found){
                pc->state=-1; 
                pc->frame_start_found= 0;
                return i-3;
            }
            pc->frame_start_found= 1;
        }
    }
    
    pc->state= state;
Michael Niedermayer's avatar
Michael Niedermayer committed
5510
    return END_NOT_FOUND;
5511 5512
}

5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533
static int h264_parse(AVCodecParserContext *s,
                      AVCodecContext *avctx,
                      uint8_t **poutbuf, int *poutbuf_size, 
                      const uint8_t *buf, int buf_size)
{
    ParseContext *pc = s->priv_data;
    int next;
    
    next= find_frame_end(pc, buf, buf_size);

    if (ff_combine_frame(pc, next, (uint8_t **)&buf, &buf_size) < 0) {
        *poutbuf = NULL;
        *poutbuf_size = 0;
        return buf_size;
    }

    *poutbuf = (uint8_t *)buf;
    *poutbuf_size = buf_size;
    return next;
}

5534 5535 5536 5537
static int decode_nal_units(H264Context *h, uint8_t *buf, int buf_size){
    MpegEncContext * const s = &h->s;
    AVCodecContext * const avctx= s->avctx;
    int buf_index=0;
Michael Niedermayer's avatar
Michael Niedermayer committed
5538
#if 0
Alex Beregszaszi's avatar
Alex Beregszaszi committed
5539
    int i;
5540 5541 5542 5543 5544 5545 5546 5547 5548
    for(i=0; i<32; i++){
        printf("%X ", buf[i]);
    }
#endif
    for(;;){
        int consumed;
        int dst_length;
        int bit_length;
        uint8_t *ptr;
5549
        int i, nalsize = 0;
5550
        
5551 5552 5553 5554 5555 5556
      if(h->is_avc) {
        if(buf_index >= buf_size) break;
        nalsize = 0;
        for(i = 0; i < h->nal_length_size; i++)
            nalsize = (nalsize << 8) | buf[buf_index++];
      } else {
5557 5558 5559 5560 5561 5562 5563 5564 5565 5566
        // start code prefix search
        for(; buf_index + 3 < buf_size; buf_index++){
            // this should allways succeed in the first iteration
            if(buf[buf_index] == 0 && buf[buf_index+1] == 0 && buf[buf_index+2] == 1)
                break;
        }
        
        if(buf_index+3 >= buf_size) break;
        
        buf_index+=3;
5567
      }  
5568 5569 5570 5571 5572 5573
        
        ptr= decode_nal(h, buf + buf_index, &dst_length, &consumed, buf_size - buf_index);
        if(ptr[dst_length - 1] == 0) dst_length--;
        bit_length= 8*dst_length - decode_rbsp_trailing(ptr + dst_length - 1);

        if(s->avctx->debug&FF_DEBUG_STARTCODE){
5574
            av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d at %d length %d\n", h->nal_unit_type, buf_index, dst_length);
5575 5576
        }
        
5577 5578 5579
        if (h->is_avc && (nalsize != consumed))
            av_log(h->s.avctx, AV_LOG_ERROR, "AVC: Consumed only %d bytes instead of %d\n", consumed, nalsize);

5580 5581
        buf_index += consumed;

5582
        if( s->hurry_up == 1 && h->nal_ref_idc  == 0 )
5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594
            continue;
        
        switch(h->nal_unit_type){
        case NAL_IDR_SLICE:
            idr(h); //FIXME ensure we dont loose some frames if there is reordering
        case NAL_SLICE:
            init_get_bits(&s->gb, ptr, bit_length);
            h->intra_gb_ptr=
            h->inter_gb_ptr= &s->gb;
            s->data_partitioning = 0;
            
            if(decode_slice_header(h) < 0) return -1;
5595
            if(h->redundant_pic_count==0 && s->hurry_up < 5 )
5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612
                decode_slice(h);
            break;
        case NAL_DPA:
            init_get_bits(&s->gb, ptr, bit_length);
            h->intra_gb_ptr=
            h->inter_gb_ptr= NULL;
            s->data_partitioning = 1;
            
            if(decode_slice_header(h) < 0) return -1;
            break;
        case NAL_DPB:
            init_get_bits(&h->intra_gb, ptr, bit_length);
            h->intra_gb_ptr= &h->intra_gb;
            break;
        case NAL_DPC:
            init_get_bits(&h->inter_gb, ptr, bit_length);
            h->inter_gb_ptr= &h->inter_gb;
5613

5614
            if(h->redundant_pic_count==0 && h->intra_gb_ptr && s->data_partitioning && s->hurry_up < 5 )
5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637
                decode_slice(h);
            break;
        case NAL_SEI:
            break;
        case NAL_SPS:
            init_get_bits(&s->gb, ptr, bit_length);
            decode_seq_parameter_set(h);
            
            if(s->flags& CODEC_FLAG_LOW_DELAY)
                s->low_delay=1;
      
            avctx->has_b_frames= !s->low_delay;
            break;
        case NAL_PPS:
            init_get_bits(&s->gb, ptr, bit_length);
            
            decode_picture_parameter_set(h);

            break;
        case NAL_PICTURE_DELIMITER:
            break;
        case NAL_FILTER_DATA:
            break;
Alex Beregszaszi's avatar
Alex Beregszaszi committed
5638 5639
	default:
	    av_log(avctx, AV_LOG_ERROR, "Unknown NAL code: %d\n", h->nal_unit_type);
5640 5641 5642 5643 5644 5645 5646
        }        

        //FIXME move after where irt is set
        s->current_picture.pict_type= s->pict_type;
        s->current_picture.key_frame= s->pict_type == I_TYPE;
    }
    
5647 5648
    if(!s->current_picture_ptr) return buf_index; //no frame
    
5649 5650 5651 5652 5653 5654 5655 5656 5657
    h->prev_frame_num_offset= h->frame_num_offset;
    h->prev_frame_num= h->frame_num;
    if(s->current_picture_ptr->reference){
        h->prev_poc_msb= h->poc_msb;
        h->prev_poc_lsb= h->poc_lsb;
    }
    if(s->current_picture_ptr->reference)
        execute_ref_pic_marking(h, h->mmco, h->mmco_index);
    else
5658
        assert(h->mmco_index==0);
5659 5660

    ff_er_frame_end(s);
5661

5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693
    MPV_frame_end(s);

    return buf_index;
}

/**
 * retunrs the number of bytes consumed for building the current frame
 */
static int get_consumed_bytes(MpegEncContext *s, int pos, int buf_size){
    if(s->flags&CODEC_FLAG_TRUNCATED){
        pos -= s->parse_context.last_index;
        if(pos<0) pos=0; // FIXME remove (uneeded?)
        
        return pos;
    }else{
        if(pos==0) pos=1; //avoid infinite loops (i doubt thats needed but ...)
        if(pos+10>buf_size) pos=buf_size; // oops ;)

        return pos;
    }
}

static int decode_frame(AVCodecContext *avctx, 
                             void *data, int *data_size,
                             uint8_t *buf, int buf_size)
{
    H264Context *h = avctx->priv_data;
    MpegEncContext *s = &h->s;
    AVFrame *pict = data; 
    int buf_index;
    
    s->flags= avctx->flags;
5694
    s->flags2= avctx->flags2;
5695 5696 5697 5698 5699 5700 5701

   /* no supplementary picture */
    if (buf_size == 0) {
        return 0;
    }
    
    if(s->flags&CODEC_FLAG_TRUNCATED){
5702
        int next= find_frame_end(&s->parse_context, buf, buf_size);
5703
        
5704
        if( ff_combine_frame(&s->parse_context, next, &buf, &buf_size) < 0 )
5705 5706 5707 5708
            return buf_size;
//printf("next:%d buf_size:%d last_index:%d\n", next, buf_size, s->parse_context.last_index);
    }

5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750
    if(h->is_avc && !h->got_avcC) {
        int i, cnt, nalsize;
        unsigned char *p = avctx->extradata;
        if(avctx->extradata_size < 7) {
            av_log(avctx, AV_LOG_ERROR, "avcC too short\n");
            return -1;
        }
        if(*p != 1) {
            av_log(avctx, AV_LOG_ERROR, "Unknown avcC version %d\n", *p);
            return -1;
        }
        /* sps and pps in the avcC always have length coded with 2 bytes,
           so put a fake nal_length_size = 2 while parsing them */
        h->nal_length_size = 2;
        // Decode sps from avcC
        cnt = *(p+5) & 0x1f; // Number of sps
        p += 6;
        for (i = 0; i < cnt; i++) {
            nalsize = BE_16(p) + 2;
            if(decode_nal_units(h, p, nalsize) != nalsize) {
                av_log(avctx, AV_LOG_ERROR, "Decoding sps %d from avcC failed\n", i);
                return -1;
            }
            p += nalsize;
        }        
        // Decode pps from avcC
        cnt = *(p++); // Number of pps
        for (i = 0; i < cnt; i++) {
            nalsize = BE_16(p) + 2;
            if(decode_nal_units(h, p, nalsize)  != nalsize) {
                av_log(avctx, AV_LOG_ERROR, "Decoding pps %d from avcC failed\n", i);
                return -1;
            }
            p += nalsize;
        }        
        // Now store right nal length size, that will be use to parse all other nals
        h->nal_length_size = ((*(((char*)(avctx->extradata))+4))&0x03)+1;
        // Do not reparse avcC
        h->got_avcC = 1;
    }

    if(!h->is_avc && s->avctx->extradata_size && s->picture_number==0){
5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768
        if(0 < decode_nal_units(h, s->avctx->extradata, s->avctx->extradata_size) ) 
            return -1;
    }

    buf_index=decode_nal_units(h, buf, buf_size);
    if(buf_index < 0) 
        return -1;

    //FIXME do something with unavailable reference frames    
 
//    if(ret==FRAME_SKIPED) return get_consumed_bytes(s, buf_index, buf_size);
#if 0
    if(s->pict_type==B_TYPE || s->low_delay){
        *pict= *(AVFrame*)&s->current_picture;
    } else {
        *pict= *(AVFrame*)&s->last_picture;
    }
#endif
5769
    if(!s->current_picture_ptr){
5770
        av_log(h->s.avctx, AV_LOG_DEBUG, "error, NO frame\n");
5771 5772 5773
        return -1;
    }

5774
    *pict= *(AVFrame*)&s->current_picture; //FIXME 
5775
    ff_print_debug_info(s, pict);
5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794
    assert(pict->data[0]);
//printf("out %d\n", (int)pict->data[0]);
#if 0 //?

    /* Return the Picture timestamp as the frame number */
    /* we substract 1 because it is added on utils.c    */
    avctx->frame_number = s->picture_number - 1;
#endif
#if 0
    /* dont output the last pic after seeking */
    if(s->last_picture_ptr || s->low_delay)
    //Note this isnt a issue as a IDR pic should flush teh buffers
#endif
        *data_size = sizeof(AVFrame);
    return get_consumed_bytes(s, buf_index, buf_size);
}
#if 0
static inline void fill_mb_avail(H264Context *h){
    MpegEncContext * const s = &h->s;
5795
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
5796 5797

    if(s->mb_y){
5798 5799 5800
        h->mb_avail[0]= s->mb_x                 && h->slice_table[mb_xy - s->mb_stride - 1] == h->slice_num;
        h->mb_avail[1]=                            h->slice_table[mb_xy - s->mb_stride    ] == h->slice_num;
        h->mb_avail[2]= s->mb_x+1 < s->mb_width && h->slice_table[mb_xy - s->mb_stride + 1] == h->slice_num;
5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825
    }else{
        h->mb_avail[0]=
        h->mb_avail[1]=
        h->mb_avail[2]= 0;
    }
    h->mb_avail[3]= s->mb_x && h->slice_table[mb_xy - 1] == h->slice_num;
    h->mb_avail[4]= 1; //FIXME move out
    h->mb_avail[5]= 0; //FIXME move out
}
#endif

#if 0 //selftest
#define COUNT 8000
#define SIZE (COUNT*40)
int main(){
    int i;
    uint8_t temp[SIZE];
    PutBitContext pb;
    GetBitContext gb;
//    int int_temp[10000];
    DSPContext dsp;
    AVCodecContext avctx;
    
    dsputil_init(&dsp, &avctx);

5826
    init_put_bits(&pb, temp, SIZE);
5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850
    printf("testing unsigned exp golomb\n");
    for(i=0; i<COUNT; i++){
        START_TIMER
        set_ue_golomb(&pb, i);
        STOP_TIMER("set_ue_golomb");
    }
    flush_put_bits(&pb);
    
    init_get_bits(&gb, temp, 8*SIZE);
    for(i=0; i<COUNT; i++){
        int j, s;
        
        s= show_bits(&gb, 24);
        
        START_TIMER
        j= get_ue_golomb(&gb);
        if(j != i){
            printf("missmatch! at %d (%d should be %d) bits:%6X\n", i, j, i, s);
//            return -1;
        }
        STOP_TIMER("get_ue_golomb");
    }
    
    
Alex Beregszaszi's avatar
Alex Beregszaszi committed
5851
    init_put_bits(&pb, temp, SIZE);
5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899
    printf("testing signed exp golomb\n");
    for(i=0; i<COUNT; i++){
        START_TIMER
        set_se_golomb(&pb, i - COUNT/2);
        STOP_TIMER("set_se_golomb");
    }
    flush_put_bits(&pb);
    
    init_get_bits(&gb, temp, 8*SIZE);
    for(i=0; i<COUNT; i++){
        int j, s;
        
        s= show_bits(&gb, 24);
        
        START_TIMER
        j= get_se_golomb(&gb);
        if(j != i - COUNT/2){
            printf("missmatch! at %d (%d should be %d) bits:%6X\n", i, j, i, s);
//            return -1;
        }
        STOP_TIMER("get_se_golomb");
    }

    printf("testing 4x4 (I)DCT\n");
    
    DCTELEM block[16];
    uint8_t src[16], ref[16];
    uint64_t error= 0, max_error=0;

    for(i=0; i<COUNT; i++){
        int j;
//        printf("%d %d %d\n", r1, r2, (r2-r1)*16);
        for(j=0; j<16; j++){
            ref[j]= random()%255;
            src[j]= random()%255;
        }

        h264_diff_dct_c(block, src, ref, 4);
        
        //normalize
        for(j=0; j<16; j++){
//            printf("%d ", block[j]);
            block[j]= block[j]*4;
            if(j&1) block[j]= (block[j]*4 + 2)/5;
            if(j&4) block[j]= (block[j]*4 + 2)/5;
        }
//        printf("\n");
        
5900
        s->dsp.h264_idct_add(ref, block, 4);
5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940 5941 5942 5943 5944 5945 5946 5947 5948 5949 5950 5951 5952 5953 5954 5955 5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008
/*        for(j=0; j<16; j++){
            printf("%d ", ref[j]);
        }
        printf("\n");*/
            
        for(j=0; j<16; j++){
            int diff= ABS(src[j] - ref[j]);
            
            error+= diff*diff;
            max_error= FFMAX(max_error, diff);
        }
    }
    printf("error=%f max_error=%d\n", ((float)error)/COUNT/16, (int)max_error );
#if 0
    printf("testing quantizer\n");
    for(qp=0; qp<52; qp++){
        for(i=0; i<16; i++)
            src1_block[i]= src2_block[i]= random()%255;
        
    }
#endif
    printf("Testing NAL layer\n");
    
    uint8_t bitstream[COUNT];
    uint8_t nal[COUNT*2];
    H264Context h;
    memset(&h, 0, sizeof(H264Context));
    
    for(i=0; i<COUNT; i++){
        int zeros= i;
        int nal_length;
        int consumed;
        int out_length;
        uint8_t *out;
        int j;
        
        for(j=0; j<COUNT; j++){
            bitstream[j]= (random() % 255) + 1;
        }
        
        for(j=0; j<zeros; j++){
            int pos= random() % COUNT;
            while(bitstream[pos] == 0){
                pos++;
                pos %= COUNT;
            }
            bitstream[pos]=0;
        }
        
        START_TIMER
        
        nal_length= encode_nal(&h, nal, bitstream, COUNT, COUNT*2);
        if(nal_length<0){
            printf("encoding failed\n");
            return -1;
        }
        
        out= decode_nal(&h, nal, &out_length, &consumed, nal_length);

        STOP_TIMER("NAL")
        
        if(out_length != COUNT){
            printf("incorrect length %d %d\n", out_length, COUNT);
            return -1;
        }
        
        if(consumed != nal_length){
            printf("incorrect consumed length %d %d\n", nal_length, consumed);
            return -1;
        }
        
        if(memcmp(bitstream, out, COUNT)){
            printf("missmatch\n");
            return -1;
        }
    }
    
    printf("Testing RBSP\n");
    
    
    return 0;
}
#endif


static int decode_end(AVCodecContext *avctx)
{
    H264Context *h = avctx->priv_data;
    MpegEncContext *s = &h->s;
    
    free_tables(h); //FIXME cleanup init stuff perhaps
    MPV_common_end(s);

//    memset(h, 0, sizeof(H264Context));
        
    return 0;
}


AVCodec h264_decoder = {
    "h264",
    CODEC_TYPE_VIDEO,
    CODEC_ID_H264,
    sizeof(H264Context),
    decode_init,
    NULL,
    decode_end,
    decode_frame,
6009
    /*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_TRUNCATED,
6010 6011
};

6012 6013 6014 6015 6016 6017 6018 6019
AVCodecParser h264_parser = {
    { CODEC_ID_H264 },
    sizeof(ParseContext),
    NULL,
    h264_parse,
    ff_parse_close,
};

6020
#include "svq3.c"