h264.h 55.9 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
/*
 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg 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 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg 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 FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
23
 * @file libavcodec/h264.h
24 25 26 27
 * H.264 / AVC / MPEG4 part10 codec.
 * @author Michael Niedermayer <michaelni@gmx.at>
 */

28 29
#ifndef AVCODEC_H264_H
#define AVCODEC_H264_H
30 31 32 33

#include "dsputil.h"
#include "cabac.h"
#include "mpegvideo.h"
34
#include "h264pred.h"
35
#include "rectangle.h"
36 37

#define interlaced_dct interlaced_dct_is_a_bad_name
Diego Biurrun's avatar
Diego Biurrun committed
38
#define mb_intra mb_intra_is_not_initialized_see_mb_type
39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54

#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

55 56
#define MAX_DELAYED_PIC_COUNT 16

57 58 59 60
/* Compiling in interlaced support reduces the speed
 * of progressive decoding by about 2%. */
#define ALLOW_INTERLACE

61 62
#define ALLOW_NOCHROMA

63 64 65 66 67 68
/**
 * The maximum number of slices supported by the decoder.
 * must be a power of 2
 */
#define MAX_SLICES 16

69 70 71 72
#ifdef ALLOW_INTERLACE
#define MB_MBAFF h->mb_mbaff
#define MB_FIELD h->mb_field_decoding_flag
#define FRAME_MBAFF h->mb_aff_frame
Carl Eugen Hoyos's avatar
Carl Eugen Hoyos committed
73
#define FIELD_PICTURE (s->picture_structure != PICT_FRAME)
74 75 76 77
#else
#define MB_MBAFF 0
#define MB_FIELD 0
#define FRAME_MBAFF 0
78
#define FIELD_PICTURE 0
79 80 81
#undef  IS_INTERLACED
#define IS_INTERLACED(mb_type) 0
#endif
82
#define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
83

84 85 86 87 88 89
#ifdef ALLOW_NOCHROMA
#define CHROMA h->sps.chroma_format_idc
#else
#define CHROMA 1
#endif

90 91 92 93
#ifndef CABAC
#define CABAC h->pps.cabac
#endif

94 95
#define EXTENDED_SAR          255

96 97 98 99 100
#define MB_TYPE_REF0       MB_TYPE_ACPRED //dirty but it fits in 16 bit
#define MB_TYPE_8x8DCT     0x01000000
#define IS_REF0(a)         ((a) & MB_TYPE_REF0)
#define IS_8x8DCT(a)       ((a) & MB_TYPE_8x8DCT)

101 102 103 104 105 106 107
/**
 * Value of Picture.reference when Picture is not a reference picture, but
 * is held for delayed output.
 */
#define DELAYED_PIC_REF 4


108 109
/* NAL unit types */
enum {
110 111 112 113 114 115 116 117 118 119 120 121 122 123
    NAL_SLICE=1,
    NAL_DPA,
    NAL_DPB,
    NAL_DPC,
    NAL_IDR_SLICE,
    NAL_SEI,
    NAL_SPS,
    NAL_PPS,
    NAL_AUD,
    NAL_END_SEQUENCE,
    NAL_END_STREAM,
    NAL_FILLER_DATA,
    NAL_SPS_EXT,
    NAL_AUXILIARY_SLICE=19
124 125
};

126 127 128 129
/**
 * SEI message types
 */
typedef enum {
130
    SEI_BUFFERING_PERIOD             =  0, ///< buffering period (H.264, D.1.1)
131 132 133 134 135
    SEI_TYPE_PIC_TIMING              =  1, ///< picture timing
    SEI_TYPE_USER_DATA_UNREGISTERED  =  5, ///< unregistered user data
    SEI_TYPE_RECOVERY_POINT          =  6  ///< recovery point (frame # to decoder sync)
} SEI_Type;

136 137 138 139 140 141 142 143 144 145 146 147 148 149 150
/**
 * pic_struct in picture timing SEI message
 */
typedef enum {
    SEI_PIC_STRUCT_FRAME             = 0, ///<  0: %frame
    SEI_PIC_STRUCT_TOP_FIELD         = 1, ///<  1: top field
    SEI_PIC_STRUCT_BOTTOM_FIELD      = 2, ///<  2: bottom field
    SEI_PIC_STRUCT_TOP_BOTTOM        = 3, ///<  3: top field, bottom field, in that order
    SEI_PIC_STRUCT_BOTTOM_TOP        = 4, ///<  4: bottom field, top field, in that order
    SEI_PIC_STRUCT_TOP_BOTTOM_TOP    = 5, ///<  5: top field, bottom field, top field repeated, in that order
    SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///<  6: bottom field, top field, bottom field repeated, in that order
    SEI_PIC_STRUCT_FRAME_DOUBLING    = 7, ///<  7: %frame doubling
    SEI_PIC_STRUCT_FRAME_TRIPLING    = 8  ///<  8: %frame tripling
} SEI_PicStructType;

151 152 153 154 155 156 157
/**
 * Sequence parameter set
 */
typedef struct SPS{

    int profile_idc;
    int level_idc;
158
    int chroma_format_idc;
159 160 161 162 163 164 165 166 167 168
    int transform_bypass;              ///< qpprime_y_zero_transform_bypass_flag
    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
    int gaps_in_frame_num_allowed_flag;
169 170
    int mb_width;                      ///< pic_width_in_mbs_minus1 + 1
    int mb_height;                     ///< pic_height_in_map_units_minus1 + 1
171 172 173 174
    int frame_mbs_only_flag;
    int mb_aff;                        ///<mb_adaptive_frame_field_flag
    int direct_8x8_inference_flag;
    int crop;                   ///< frame_cropping_flag
175 176 177 178
    unsigned int crop_left;            ///< frame_cropping_rect_left_offset
    unsigned int crop_right;           ///< frame_cropping_rect_right_offset
    unsigned int crop_top;             ///< frame_cropping_rect_top_offset
    unsigned int crop_bottom;          ///< frame_cropping_rect_bottom_offset
179 180
    int vui_parameters_present_flag;
    AVRational sar;
181 182 183 184 185 186
    int video_signal_type_present_flag;
    int full_range;
    int colour_description_present_flag;
    enum AVColorPrimaries color_primaries;
    enum AVColorTransferCharacteristic color_trc;
    enum AVColorSpace colorspace;
187 188 189 190 191 192 193 194 195 196
    int timing_info_present_flag;
    uint32_t num_units_in_tick;
    uint32_t time_scale;
    int fixed_frame_rate_flag;
    short offset_for_ref_frame[256]; //FIXME dyn aloc?
    int bitstream_restriction_flag;
    int num_reorder_frames;
    int scaling_matrix_present;
    uint8_t scaling_matrix4[6][16];
    uint8_t scaling_matrix8[2][64];
197 198 199 200
    int nal_hrd_parameters_present_flag;
    int vcl_hrd_parameters_present_flag;
    int pic_struct_present_flag;
    int time_offset_length;
201
    int cpb_cnt;                       ///< See H.264 E.1.2
202
    int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 +1
203 204
    int cpb_removal_delay_length;      ///< cpb_removal_delay_length_minus1 + 1
    int dpb_output_delay_length;       ///< dpb_output_delay_length_minus1 + 1
205 206 207
    int bit_depth_luma;                ///< bit_depth_luma_minus8 + 8
    int bit_depth_chroma;              ///< bit_depth_chroma_minus8 + 8
    int residual_color_transform_flag; ///< residual_colour_transform_flag
208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
}SPS;

/**
 * Picture parameter set
 */
typedef struct PPS{
    unsigned 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;
    unsigned 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
224
    int chroma_qp_index_offset[2];
225 226 227 228 229 230
    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
    int transform_8x8_mode;     ///< transform_8x8_mode_flag
    uint8_t scaling_matrix4[6][16];
    uint8_t scaling_matrix8[2][64];
231
    uint8_t chroma_qp_table[2][64];  ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
232
    int chroma_qp_diff;
233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
}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;
253 254
    int short_pic_num;  ///< pic_num without wrapping (pic_num & max_pic_num)
    int long_arg;       ///< index, pic_num, or num long refs depending on opcode
255 256 257 258 259 260 261 262 263
} MMCO;

/**
 * H264Context
 */
typedef struct H264Context{
    MpegEncContext s;
    int nal_ref_idc;
    int nal_unit_type;
264 265
    uint8_t *rbsp_buffer[2];
    unsigned int rbsp_buffer_size[2];
266 267 268 269 270 271 272 273

    /**
      * 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)

274
    int chroma_qp[2]; //QPc
275

276 277
    int qp_thresh;      ///< QP threshold to skip loopfilter

278 279 280 281 282 283 284 285 286 287
    int prev_mb_skipped;
    int next_mb_skipped;

    //prediction stuff
    int chroma_pred_mode;
    int intra16x16_pred_mode;

    int top_mb_xy;
    int left_mb_xy[2];

288 289 290
    int top_type;
    int left_type[2];

291 292
    int8_t intra4x4_pred_mode_cache[5*8];
    int8_t (*intra4x4_pred_mode)[8];
293
    H264PredContext hpc;
294 295 296 297 298 299 300 301 302 303 304
    unsigned int topleft_samples_available;
    unsigned int top_samples_available;
    unsigned int topright_samples_available;
    unsigned int left_samples_available;
    uint8_t (*top_borders[2])[16+2*8];
    uint8_t left_border[2*(17+2*9)];

    /**
     * non zero coeff count cache.
     * is 64 if not available.
     */
305
    DECLARE_ALIGNED_8(uint8_t, non_zero_count_cache)[6*8];
306 307 308 309 310 311 312

    /*
    .UU.YYYY
    .UU.YYYY
    .vv.YYYY
    .VV.YYYY
    */
313
    uint8_t (*non_zero_count)[32];
314 315 316 317

    /**
     * Motion vector cache.
     */
318
    DECLARE_ALIGNED_16(int16_t, mv_cache)[2][5*8][2];
319
    DECLARE_ALIGNED_8(int8_t, ref_cache)[2][5*8];
320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
#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];

    /**
     * number of neighbors (top and/or left) that used 8x8 dct
     */
    int neighbor_transform_size;

    /**
     * block_offset[ 0..23] for frame macroblocks
     * block_offset[24..47] for field macroblocks
     */
    int block_offset[2*(16+8)];

    uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
    uint32_t *mb2b8_xy;
    int b_stride; //FIXME use s->b4_stride
    int b8_stride;

    int mb_linesize;   ///< may be equal to s->linesize or s->linesize*2, for mbaff
    int mb_uvlinesize;

    int emu_edge_width;
    int emu_edge_height;

    int halfpel_flag;
    int thirdpel_flag;

    int unknown_svq3_flag;
    int next_slice_index;

356
    SPS *sps_buffers[MAX_SPS_COUNT];
357 358
    SPS sps; ///< current sps

359
    PPS *pps_buffers[MAX_PPS_COUNT];
360 361 362 363 364 365 366 367 368 369 370 371
    /**
     * current pps
     */
    PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?

    uint32_t dequant4_buffer[6][52][16];
    uint32_t dequant8_buffer[2][52][64];
    uint32_t (*dequant4_coeff[6])[16];
    uint32_t (*dequant8_coeff[2])[64];
    int dequant_coeff_pps;     ///< reinit tables when pps changes

    int slice_num;
372 373
    uint16_t *slice_table_base;
    uint16_t *slice_table;     ///< slice_table_base + 2*mb_stride + 1
374
    int slice_type;
375
    int slice_type_nos;        ///< S free slice type (SI/SP are remapped to I/P)
376 377 378 379 380 381 382
    int slice_type_fixed;

    //interlacing specific flags
    int mb_aff_frame;
    int mb_field_decoding_flag;
    int mb_mbaff;              ///< mb_aff_frame && mb_field_decoding_flag

383
    DECLARE_ALIGNED_8(uint16_t, sub_mb_type)[4];
384 385 386 387 388 389 390 391 392 393 394 395 396 397

    //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

    /**
398
     * frame_num for frames or 2*frame_num+1 for field pics.
399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425
     */
    int curr_pic_num;

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

    //Weighted pred stuff
    int use_weight;
    int use_weight_chroma;
    int luma_log2_weight_denom;
    int chroma_log2_weight_denom;
    int luma_weight[2][48];
    int luma_offset[2][48];
    int chroma_weight[2][48][2];
    int chroma_offset[2][48][2];
    int implicit_weight[48][48];

    //deblock
    int deblocking_filter;         ///< disable_deblocking_filter_idc with 1<->0
    int slice_alpha_c0_offset;
    int slice_beta_offset;

    int redundant_pic_count;

    int direct_spatial_mv_pred;
426 427
    int col_parity;
    int col_fieldoff;
428
    int dist_scale_factor[16];
429
    int dist_scale_factor_field[2][32];
430 431
    int map_col_to_list0[2][16+32];
    int map_col_to_list0_field[2][2][16+32];
432 433 434 435 436 437

    /**
     * num_ref_idx_l0/1_active_minus1 + 1
     */
    unsigned int ref_count[2];   ///< counts frames or fields, depending on current mb mode
    unsigned int list_count;
438
    uint8_t *list_counts;            ///< Array of list_count per MB specifying the slice type
439 440
    Picture *short_ref[32];
    Picture *long_ref[32];
441 442 443 444
    Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
    Picture ref_list[2][48];         /**< 0..15: frame refs, 16..47: mbaff field refs.
                                          Reordered version of default_ref_list
                                          according to picture reordering in slice header */
445
    int ref2frm[MAX_SLICES][2][64];  ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
446
    Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
447
    int outputed_poc;
448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463

    /**
     * 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;

464
    DECLARE_ALIGNED_16(DCTELEM, mb)[16*24];
Diego Biurrun's avatar
Diego Biurrun committed
465
    DCTELEM mb_padding[256];        ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either check that i is not too large or ensure that there is some unused stuff after mb
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482

    /**
     * Cabac
     */
    CABACContext cabac;
    uint8_t      cabac_state[460];
    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;
    int cbp;
    int top_cbp;
    int left_cbp;
    /* chroma_pred_mode for i4x4 or i16x16, else 0 */
    uint8_t     *chroma_pred_mode_table;
    int         last_qscale_diff;
    int16_t     (*mvd_table[2])[2];
483
    DECLARE_ALIGNED_16(int16_t, mvd_cache)[2][5*8][2];
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
    uint8_t     *direct_table;
    uint8_t     direct_cache[5*8];

    uint8_t zigzag_scan[16];
    uint8_t zigzag_scan8x8[64];
    uint8_t zigzag_scan8x8_cavlc[64];
    uint8_t field_scan[16];
    uint8_t field_scan8x8[64];
    uint8_t field_scan8x8_cavlc[64];
    const uint8_t *zigzag_scan_q0;
    const uint8_t *zigzag_scan8x8_q0;
    const uint8_t *zigzag_scan8x8_cavlc_q0;
    const uint8_t *field_scan_q0;
    const uint8_t *field_scan8x8_q0;
    const uint8_t *field_scan8x8_cavlc_q0;

    int x264_build;
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529

    /**
     * @defgroup multithreading Members for slice based multithreading
     * @{
     */
    struct H264Context *thread_context[MAX_THREADS];

    /**
     * current slice number, used to initalize slice_num of each thread/context
     */
    int current_slice;

    /**
     * Max number of threads / contexts.
     * This is equal to AVCodecContext.thread_count unless
     * multithreaded decoding is impossible, in which case it is
     * reduced to 1.
     */
    int max_contexts;

    /**
     *  1 if the single thread fallback warning has already been
     *  displayed, 0 otherwise.
     */
    int single_decode_warning;

    int last_slice_type;
    /** @} */

530 531
    int mb_xy;

532
    uint32_t svq3_watermark_key;
533 534 535 536 537

    /**
     * pic_struct in picture timing SEI message
     */
    SEI_PicStructType sei_pic_struct;
538

539
    /**
540 541 542 543 544 545 546 547
     * Complement sei_pic_struct
     * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
     * However, soft telecined frames may have these values.
     * This is used in an attempt to flag soft telecine progressive.
     */
    int prev_interlaced_frame;

    /**
548 549 550 551 552 553
     * Bit set of clock types for fields/frames in picture timing SEI message.
     * For each found ct_type, appropriate bit is set (e.g., bit 1 for
     * interlaced).
     */
    int sei_ct_type;

554 555 556 557 558
    /**
     * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
     */
    int sei_dpb_output_delay;

559 560 561 562 563
    /**
     * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
     */
    int sei_cpb_removal_delay;

564 565 566 567 568 569 570 571 572
    /**
     * recovery_frame_cnt from SEI message
     *
     * Set to -1 if no recovery point SEI message found or to number of frames
     * before playback synchronizes. Frames having recovery point are key
     * frames.
     */
    int sei_recovery_frame_cnt;

573
    int is_complex;
574 575 576

    int luma_weight_flag[2];   ///< 7.4.3.2 luma_weight_lX_flag
    int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
577 578 579 580

    // Timestamp stuff
    int sei_buffering_period_present;  ///< Buffering period SEI flag
    int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
581 582
}H264Context;

583 584 585

extern const uint8_t ff_h264_chroma_qp[52];

586 587 588
void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);

void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
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
/**
 * Decode SEI
 */
int ff_h264_decode_sei(H264Context *h);

/**
 * Decode SPS
 */
int ff_h264_decode_seq_parameter_set(H264Context *h);

/**
 * Decode PPS
 */
int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);

/**
 * 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 tailing?
 * @returns decoded bytes, might be src+1 if no escapes
 */
const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length);

/**
 * identifies the exact end of the bitstream
 * @return the length of the trailing, or 0 if damaged
 */
int ff_h264_decode_rbsp_trailing(H264Context *h, const uint8_t *src);

620 621 622
/**
 * frees any data that may have been allocated in the H264 context like SPS, PPS etc.
 */
623
av_cold void ff_h264_free_context(H264Context *h);
624

625 626 627
/**
 * reconstructs bitstream slice_type.
 */
628
int ff_h264_get_slice_type(const H264Context *h);
629

630 631 632 633 634 635
/**
 * allocates tables.
 * needs width/height
 */
int ff_h264_alloc_tables(H264Context *h);

636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652
/**
 * fills the default_ref_list.
 */
int ff_h264_fill_default_ref_list(H264Context *h);

int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
void ff_h264_fill_mbaff_ref_list(H264Context *h);
void ff_h264_remove_all_refs(H264Context *h);

/**
 * Executes the reference picture marking (memory management control operations).
 */
int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);

int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb);


653 654 655 656 657
/**
 * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
 */
int ff_h264_check_intra4x4_pred_mode(H264Context *h);

658 659 660 661 662 663 664 665 666 667
/**
 * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
 */
int ff_h264_check_intra_pred_mode(H264Context *h, int mode);

void ff_h264_write_back_intra_pred_mode(H264Context *h);
void ff_h264_hl_decode_mb(H264Context *h);
int ff_h264_frame_start(H264Context *h);
av_cold int ff_h264_decode_init(AVCodecContext *avctx);
av_cold int ff_h264_decode_end(AVCodecContext *avctx);
668 669 670 671 672 673 674
av_cold void ff_h264_decode_init_vlc(void);

/**
 * decodes a macroblock
 * @returns 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
 */
int ff_h264_decode_mb_cavlc(H264Context *h);
675

676 677 678 679 680 681 682 683
/**
 * decodes a CABAC coded macroblock
 * @returns 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
 */
int ff_h264_decode_mb_cabac(H264Context *h);

void ff_h264_init_cabac_states(H264Context *h);

684 685 686 687
void ff_h264_direct_dist_scale_factor(H264Context * const h);
void ff_h264_direct_ref_list_init(H264Context * const h);
void ff_h264_pred_direct_motion(H264Context * const h, int *mb_type);

688 689 690
void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);
void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);

691 692 693 694 695 696 697 698
/**
 * Reset SEI values at the beginning of the frame.
 *
 * @param h H.264 context.
 */
void ff_h264_reset_sei(H264Context *h);


699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727
/*
o-o o-o
 / / /
o-o o-o
 ,---'
o-o o-o
 / / /
o-o o-o
*/
//This table must be here because scan8[constant] must be known at compiletime
static const uint8_t scan8[16 + 2*4]={
 4+1*8, 5+1*8, 4+2*8, 5+2*8,
 6+1*8, 7+1*8, 6+2*8, 7+2*8,
 4+3*8, 5+3*8, 4+4*8, 5+4*8,
 6+3*8, 7+3*8, 6+4*8, 7+4*8,
 1+1*8, 2+1*8,
 1+2*8, 2+2*8,
 1+4*8, 2+4*8,
 1+5*8, 2+5*8,
};

static av_always_inline uint32_t pack16to32(int a, int b){
#if HAVE_BIGENDIAN
   return (b&0xFFFF) + (a<<16);
#else
   return (a&0xFFFF) + (b<<16);
#endif
}

728 729 730 731 732 733 734
/**
 * gets the chroma qp.
 */
static inline int get_chroma_qp(H264Context *h, int t, int qscale){
    return h->pps.chroma_qp_table[t][qscale];
}

735 736
static inline void pred_pskip_motion(H264Context * const h, int * const mx, int * const my);

737
static void fill_decode_caches(H264Context *h, int mb_type){
738 739 740 741 742 743 744
    MpegEncContext * const s = &h->s;
    const int mb_xy= h->mb_xy;
    int topleft_xy, top_xy, topright_xy, left_xy[2];
    int topleft_type, top_type, topright_type, left_type[2];
    const uint8_t * left_block;
    int topleft_partition= -1;
    int i;
745 746 747 748 749
    static const uint8_t left_block_options[4][16]={
        {0,1,2,3,7,10,8,11,7+0*8, 7+1*8, 7+2*8, 7+3*8, 2+0*8, 2+3*8, 2+1*8, 2+2*8},
        {2,2,3,3,8,11,8,11,7+2*8, 7+2*8, 7+3*8, 7+3*8, 2+1*8, 2+2*8, 2+1*8, 2+2*8},
        {0,0,1,1,7,10,7,10,7+0*8, 7+0*8, 7+1*8, 7+1*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8},
        {0,2,0,2,7,10,7,10,7+0*8, 7+2*8, 7+0*8, 7+2*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8}
750 751
    };

752
    top_xy     = mb_xy  - (s->mb_stride << MB_FIELD);
753 754 755 756 757 758 759 760 761

    /* Wow, what a mess, why didn't they simplify the interlacing & intra
     * stuff, I can't imagine that these complex rules are worth it. */

    topleft_xy = top_xy - 1;
    topright_xy= top_xy + 1;
    left_xy[1] = left_xy[0] = mb_xy-1;
    left_block = left_block_options[0];
    if(FRAME_MBAFF){
762
        const int left_mb_field_flag     = IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]);
763
        const int curr_mb_field_flag     = IS_INTERLACED(mb_type);
764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790
        if(s->mb_y&1){
            if (left_mb_field_flag != curr_mb_field_flag) {
                left_xy[1] = left_xy[0] = mb_xy - s->mb_stride - 1;
                if (curr_mb_field_flag) {
                    left_xy[1] += s->mb_stride;
                    left_block = left_block_options[3];
                } else {
                    topleft_xy += s->mb_stride;
                    // take top left mv from the middle of the mb, as opposed to all other modes which use the bottom right partition
                    topleft_partition = 0;
                    left_block = left_block_options[1];
                }
            }
        }else{
            if(curr_mb_field_flag){
                topleft_xy  += s->mb_stride & (((s->current_picture.mb_type[top_xy - 1]>>7)&1)-1);
                topright_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy + 1]>>7)&1)-1);
                top_xy      += s->mb_stride & (((s->current_picture.mb_type[top_xy    ]>>7)&1)-1);
            }
            if (left_mb_field_flag != curr_mb_field_flag) {
                left_xy[1] = left_xy[0] = mb_xy - 1;
                if (curr_mb_field_flag) {
                    left_xy[1] += s->mb_stride;
                    left_block = left_block_options[3];
                } else {
                    left_block = left_block_options[2];
                }
791 792 793 794 795 796 797 798 799 800 801 802 803
            }
        }
    }

    h->top_mb_xy = top_xy;
    h->left_mb_xy[0] = left_xy[0];
    h->left_mb_xy[1] = left_xy[1];
        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;

804
    if(!IS_SKIP(mb_type)){
805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834
        if(IS_INTRA(mb_type)){
            int type_mask= h->pps.constrained_intra_pred ? IS_INTRA(-1) : -1;
            h->topleft_samples_available=
            h->top_samples_available=
            h->left_samples_available= 0xFFFF;
            h->topright_samples_available= 0xEEEA;

            if(!(top_type & type_mask)){
                h->topleft_samples_available= 0xB3FF;
                h->top_samples_available= 0x33FF;
                h->topright_samples_available= 0x26EA;
            }
            if(IS_INTERLACED(mb_type) != IS_INTERLACED(left_type[0])){
                if(IS_INTERLACED(mb_type)){
                    if(!(left_type[0] & type_mask)){
                        h->topleft_samples_available&= 0xDFFF;
                        h->left_samples_available&= 0x5FFF;
                    }
                    if(!(left_type[1] & type_mask)){
                        h->topleft_samples_available&= 0xFF5F;
                        h->left_samples_available&= 0xFF5F;
                    }
                }else{
                    int left_typei = h->slice_table[left_xy[0] + s->mb_stride ] == h->slice_num
                                    ? s->current_picture.mb_type[left_xy[0] + s->mb_stride] : 0;
                    assert(left_xy[0] == left_xy[1]);
                    if(!((left_typei & type_mask) && (left_type[0] & type_mask))){
                        h->topleft_samples_available&= 0xDF5F;
                        h->left_samples_available&= 0x5F5F;
                    }
835 836
                }
            }else{
837
                if(!(left_type[0] & type_mask)){
838 839 840 841 842
                    h->topleft_samples_available&= 0xDF5F;
                    h->left_samples_available&= 0x5F5F;
                }
            }

843 844
            if(!(topleft_type & type_mask))
                h->topleft_samples_available&= 0x7FFF;
845

846 847
            if(!(topright_type & type_mask))
                h->topright_samples_available&= 0xFBFF;
848

849 850 851 852 853 854
            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];
855 856
                }else{
                    int pred;
857
                    if(!(top_type & type_mask))
858 859 860 861
                        pred= -1;
                    else{
                        pred= 2;
                    }
862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880
                    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(!(left_type[i] & type_mask))
                            pred= -1;
                        else{
                            pred= 2;
                        }
                        h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]=
                        h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= pred;
                    }
881 882 883 884 885 886 887 888 889 890 891 892 893 894 895
                }
            }
        }


/*
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 (let us hope this is just a typo in the spec)
    if(top_type){
896
        *(uint32_t*)&h->non_zero_count_cache[4+8*0]= *(uint32_t*)&h->non_zero_count[top_xy][4+3*8];
897 898
            h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][1+1*8];
            h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][2+1*8];
899

900 901
            h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][1+2*8];
            h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][2+2*8];
902
    }else {
903 904
            h->non_zero_count_cache[1+8*0]=
            h->non_zero_count_cache[2+8*0]=
905

906 907 908
            h->non_zero_count_cache[1+8*3]=
            h->non_zero_count_cache[2+8*3]=
            *(uint32_t*)&h->non_zero_count_cache[4+8*0]= CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040;
909 910 911 912
    }

    for (i=0; i<2; i++) {
        if(left_type[i]){
913 914
            h->non_zero_count_cache[3+8*1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]];
            h->non_zero_count_cache[3+8*2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]];
915 916
                h->non_zero_count_cache[0+8*1 +   8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]];
                h->non_zero_count_cache[0+8*4 +   8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]];
917
        }else{
918 919 920 921
                h->non_zero_count_cache[3+8*1 + 2*8*i]=
                h->non_zero_count_cache[3+8*2 + 2*8*i]=
                h->non_zero_count_cache[0+8*1 +   8*i]=
                h->non_zero_count_cache[0+8*4 +   8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64;
922 923 924
        }
    }

925
    if( CABAC ) {
926 927 928 929
        // top_cbp
        if(top_type) {
            h->top_cbp = h->cbp_table[top_xy];
        } else if(IS_INTRA(mb_type)) {
930
            h->top_cbp = 0x1CF;
931
        } else {
932
            h->top_cbp = 0x00F;
933 934 935 936 937
        }
        // left_cbp
        if (left_type[0]) {
            h->left_cbp = h->cbp_table[left_xy[0]] & 0x1f0;
        } else if(IS_INTRA(mb_type)) {
938
            h->left_cbp = 0x1CF;
939
        } else {
940
            h->left_cbp = 0x00F;
941 942 943 944 945 946 947 948
        }
        if (left_type[0]) {
            h->left_cbp |= ((h->cbp_table[left_xy[0]]>>((left_block[0]&(~1))+1))&0x1) << 1;
        }
        if (left_type[1]) {
            h->left_cbp |= ((h->cbp_table[left_xy[1]]>>((left_block[2]&(~1))+1))&0x1) << 3;
        }
    }
949
    }
950 951

#if 1
952
    if(IS_INTER(mb_type) || (IS_DIRECT(mb_type) && h->direct_spatial_mv_pred)){
953 954
        int list;
        for(list=0; list<h->list_count; list++){
955
            if(!USES_LIST(mb_type, list)){
956 957 958 959 960 961 962
                /*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;
            }
963 964
            assert(!(IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred));

965 966 967 968 969
            h->mv_cache_clean[list]= 0;

            if(USES_LIST(top_type, list)){
                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;
970
                AV_COPY128(h->mv_cache[list][scan8[0] + 0 - 1*8], s->current_picture.motion_val[list][b_xy + 0]);
971 972 973 974
                    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];
975
            }else{
976
                AV_ZERO128(h->mv_cache[list][scan8[0] + 0 - 1*8]);
977
                *(uint32_t*)&h->ref_cache[list][scan8[0] + 0 - 1*8]= ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101;
978 979 980 981 982 983 984 985 986
            }

            for(i=0; i<2; i++){
                int cache_idx = scan8[0] - 1 + i*2*8;
                if(USES_LIST(left_type[i], list)){
                    const int b_xy= h->mb2b_xy[left_xy[i]] + 3;
                    const int b8_xy= h->mb2b8_xy[left_xy[i]] + 1;
                    *(uint32_t*)h->mv_cache[list][cache_idx  ]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0+i*2]];
                    *(uint32_t*)h->mv_cache[list][cache_idx+8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[1+i*2]];
987 988
                        h->ref_cache[list][cache_idx  ]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[0+i*2]>>1)];
                        h->ref_cache[list][cache_idx+8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[1+i*2]>>1)];
989 990 991 992
                }else{
                    *(uint32_t*)h->mv_cache [list][cache_idx  ]=
                    *(uint32_t*)h->mv_cache [list][cache_idx+8]= 0;
                    h->ref_cache[list][cache_idx  ]=
993
                    h->ref_cache[list][cache_idx+8]= (left_type[i]) ? LIST_NOT_USED : PART_NOT_AVAILABLE;
994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016
                }
            }

            if(USES_LIST(topleft_type, list)){
                const int b_xy = h->mb2b_xy[topleft_xy] + 3 + h->b_stride + (topleft_partition & 2*h->b_stride);
                const int b8_xy= h->mb2b8_xy[topleft_xy] + 1 + (topleft_partition & 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(USES_LIST(topright_type, list)){
                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;
            }

1017
            if((mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2)) && !FRAME_MBAFF)
1018 1019
                continue;

1020
            if(!(mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2))) {
1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
            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 somewhere 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 somewhere else)
            *(uint32_t*)h->mv_cache [list][scan8[4 ]]=
            *(uint32_t*)h->mv_cache [list][scan8[12]]= 0;

1032
            if( CABAC ) {
1033 1034 1035
                /* XXX beurk, Load mvd */
                if(USES_LIST(top_type, list)){
                    const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
1036
                    AV_COPY128(h->mvd_cache[list][scan8[0] + 0 - 1*8], h->mvd_table[list][b_xy + 0]);
1037
                }else{
1038
                    AV_ZERO128(h->mvd_cache[list][scan8[0] + 0 - 1*8]);
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
                }
                if(USES_LIST(left_type[0], list)){
                    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(USES_LIST(left_type[1], list)){
                    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 somewhere else)
                *(uint32_t*)h->mvd_cache [list][scan8[4 ]]=
                *(uint32_t*)h->mvd_cache [list][scan8[12]]= 0;

                if(h->slice_type_nos == FF_B_TYPE){
1063
                    fill_rectangle(&h->direct_cache[scan8[0]], 4, 4, 8, MB_TYPE_16x16>>1, 1);
1064 1065

                    if(IS_DIRECT(top_type)){
1066
                        *(uint32_t*)&h->direct_cache[scan8[0] - 1*8]= 0x01010101*(MB_TYPE_DIRECT2>>1);
1067 1068 1069 1070 1071
                    }else if(IS_8X8(top_type)){
                        int b8_xy = h->mb2b8_xy[top_xy] + h->b8_stride;
                        h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy];
                        h->direct_cache[scan8[0] + 2 - 1*8]= h->direct_table[b8_xy + 1];
                    }else{
1072
                        *(uint32_t*)&h->direct_cache[scan8[0] - 1*8]= 0x01010101*(MB_TYPE_16x16>>1);
1073 1074 1075
                    }

                    if(IS_DIRECT(left_type[0]))
1076
                        h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_DIRECT2>>1;
1077 1078 1079
                    else if(IS_8X8(left_type[0]))
                        h->direct_cache[scan8[0] - 1 + 0*8]= h->direct_table[h->mb2b8_xy[left_xy[0]] + 1 + h->b8_stride*(left_block[0]>>1)];
                    else
1080
                        h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_16x16>>1;
1081 1082

                    if(IS_DIRECT(left_type[1]))
1083
                        h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_DIRECT2>>1;
1084 1085 1086
                    else if(IS_8X8(left_type[1]))
                        h->direct_cache[scan8[0] - 1 + 2*8]= h->direct_table[h->mb2b8_xy[left_xy[1]] + 1 + h->b8_stride*(left_block[2]>>1)];
                    else
1087
                        h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_16x16>>1;
1088 1089
                }
            }
1090
            }
1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
            if(FRAME_MBAFF){
#define MAP_MVS\
                    MAP_F2F(scan8[0] - 1 - 1*8, topleft_type)\
                    MAP_F2F(scan8[0] + 0 - 1*8, top_type)\
                    MAP_F2F(scan8[0] + 1 - 1*8, top_type)\
                    MAP_F2F(scan8[0] + 2 - 1*8, top_type)\
                    MAP_F2F(scan8[0] + 3 - 1*8, top_type)\
                    MAP_F2F(scan8[0] + 4 - 1*8, topright_type)\
                    MAP_F2F(scan8[0] - 1 + 0*8, left_type[0])\
                    MAP_F2F(scan8[0] - 1 + 1*8, left_type[0])\
                    MAP_F2F(scan8[0] - 1 + 2*8, left_type[1])\
                    MAP_F2F(scan8[0] - 1 + 3*8, left_type[1])
                if(MB_FIELD){
#define MAP_F2F(idx, mb_type)\
                    if(!IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
                        h->ref_cache[list][idx] <<= 1;\
                        h->mv_cache[list][idx][1] /= 2;\
                        h->mvd_cache[list][idx][1] /= 2;\
                    }
                    MAP_MVS
#undef MAP_F2F
                }else{
#define MAP_F2F(idx, mb_type)\
                    if(IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
                        h->ref_cache[list][idx] >>= 1;\
                        h->mv_cache[list][idx][1] <<= 1;\
                        h->mvd_cache[list][idx][1] <<= 1;\
                    }
                    MAP_MVS
#undef MAP_F2F
                }
            }
        }
    }
#endif

1127
        h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
1128 1129
}

1130 1131 1132 1133 1134
/**
 *
 * @returns non zero if the loop filter can be skiped
 */
static int fill_filter_caches(H264Context *h, int mb_type){
1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
    MpegEncContext * const s = &h->s;
    const int mb_xy= h->mb_xy;
    int top_xy, left_xy[2];
    int top_type, left_type[2];

    top_xy     = mb_xy  - (s->mb_stride << MB_FIELD);

    //FIXME deblocking could skip the intra and nnz parts.

    /* Wow, what a mess, why didn't they simplify the interlacing & intra
     * stuff, I can't imagine that these complex rules are worth it. */

    left_xy[1] = left_xy[0] = mb_xy-1;
    if(FRAME_MBAFF){
        const int left_mb_field_flag     = IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]);
        const int curr_mb_field_flag     = IS_INTERLACED(mb_type);
        if(s->mb_y&1){
            if (left_mb_field_flag != curr_mb_field_flag) {
Michael Niedermayer's avatar
Michael Niedermayer committed
1153
                left_xy[0] -= s->mb_stride;
1154 1155 1156 1157 1158 1159
            }
        }else{
            if(curr_mb_field_flag){
                top_xy      += s->mb_stride & (((s->current_picture.mb_type[top_xy    ]>>7)&1)-1);
            }
            if (left_mb_field_flag != curr_mb_field_flag) {
Michael Niedermayer's avatar
Michael Niedermayer committed
1160
                left_xy[1] += s->mb_stride;
1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 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
            }
        }
    }

    h->top_mb_xy = top_xy;
    h->left_mb_xy[0] = left_xy[0];
    h->left_mb_xy[1] = left_xy[1];
    {
        //for sufficiently low qp, filtering wouldn't do anything
        //this is a conservative estimate: could also check beta_offset and more accurate chroma_qp
        int qp_thresh = h->qp_thresh; //FIXME strictly we should store qp_thresh for each mb of a slice
        int qp = s->current_picture.qscale_table[mb_xy];
        if(qp <= qp_thresh
           && (left_xy[0]<0 || ((qp + s->current_picture.qscale_table[left_xy[0]] + 1)>>1) <= qp_thresh)
           && (top_xy   < 0 || ((qp + s->current_picture.qscale_table[top_xy    ] + 1)>>1) <= qp_thresh)){
            if(!FRAME_MBAFF)
                return 1;
            if(   (left_xy[0]< 0            || ((qp + s->current_picture.qscale_table[left_xy[1]             ] + 1)>>1) <= qp_thresh)
               && (top_xy    < s->mb_stride || ((qp + s->current_picture.qscale_table[top_xy    -s->mb_stride] + 1)>>1) <= qp_thresh))
                return 1;
        }
    }

    if(h->deblocking_filter == 2){
        h->top_type    = top_type     = h->slice_table[top_xy     ] == h->slice_num ? s->current_picture.mb_type[top_xy]     : 0;
        h->left_type[0]= left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
        h->left_type[1]= left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
    }else{
        h->top_type    = top_type     = h->slice_table[top_xy     ] < 0xFFFF ? s->current_picture.mb_type[top_xy]     : 0;
        h->left_type[0]= left_type[0] = h->slice_table[left_xy[0] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[0]] : 0;
        h->left_type[1]= left_type[1] = h->slice_table[left_xy[1] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[1]] : 0;
    }
    if(IS_INTRA(mb_type))
        return 0;

    AV_COPY64(&h->non_zero_count_cache[0+8*1], &h->non_zero_count[mb_xy][ 0]);
    AV_COPY64(&h->non_zero_count_cache[0+8*2], &h->non_zero_count[mb_xy][ 8]);
    *((uint32_t*)&h->non_zero_count_cache[0+8*5])= *((uint32_t*)&h->non_zero_count[mb_xy][16]);
    *((uint32_t*)&h->non_zero_count_cache[4+8*3])= *((uint32_t*)&h->non_zero_count[mb_xy][20]);
    AV_COPY64(&h->non_zero_count_cache[0+8*4], &h->non_zero_count[mb_xy][24]);

    h->cbp= h->cbp_table[mb_xy];

    {
        int list;
        for(list=0; list<h->list_count; list++){
            int8_t *ref;
            int y, b_stride;
            int16_t (*mv_dst)[2];
            int16_t (*mv_src)[2];

            if(!USES_LIST(mb_type, list)){
                fill_rectangle(  h->mv_cache[list][scan8[0]], 4, 4, 8, pack16to32(0,0), 4);
                *(uint32_t*)&h->ref_cache[list][scan8[ 0]] =
                *(uint32_t*)&h->ref_cache[</