1/*
2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
23 * @file libavcodec/h264.h
24 * H.264 / AVC / MPEG4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28#ifndef AVCODEC_H264_H
29#define AVCODEC_H264_H
30
31#include "dsputil.h"
32#include "cabac.h"
33#include "mpegvideo.h"
34#include "h264pred.h"
35
36#define interlaced_dct interlaced_dct_is_a_bad_name
37#define mb_intra mb_intra_is_not_initialized_see_mb_type
38
39#define LUMA_DC_BLOCK_INDEX   25
40#define CHROMA_DC_BLOCK_INDEX 26
41
42#define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
43#define COEFF_TOKEN_VLC_BITS           8
44#define TOTAL_ZEROS_VLC_BITS           9
45#define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
46#define RUN_VLC_BITS                   3
47#define RUN7_VLC_BITS                  6
48
49#define MAX_SPS_COUNT 32
50#define MAX_PPS_COUNT 256
51
52#define MAX_MMCO_COUNT 66
53
54#define MAX_DELAYED_PIC_COUNT 16
55
56/* Compiling in interlaced support reduces the speed
57 * of progressive decoding by about 2%. */
58#define ALLOW_INTERLACE
59
60#define ALLOW_NOCHROMA
61
62/**
63 * The maximum number of slices supported by the decoder.
64 * must be a power of 2
65 */
66#define MAX_SLICES 16
67
68#ifdef ALLOW_INTERLACE
69#define MB_MBAFF h->mb_mbaff
70#define MB_FIELD h->mb_field_decoding_flag
71#define FRAME_MBAFF h->mb_aff_frame
72#define FIELD_PICTURE (s->picture_structure != PICT_FRAME)
73#else
74#define MB_MBAFF 0
75#define MB_FIELD 0
76#define FRAME_MBAFF 0
77#define FIELD_PICTURE 0
78#undef  IS_INTERLACED
79#define IS_INTERLACED(mb_type) 0
80#endif
81#define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
82
83#ifdef ALLOW_NOCHROMA
84#define CHROMA h->sps.chroma_format_idc
85#else
86#define CHROMA 1
87#endif
88
89#define EXTENDED_SAR          255
90
91#define MB_TYPE_REF0       MB_TYPE_ACPRED //dirty but it fits in 16 bit
92#define MB_TYPE_8x8DCT     0x01000000
93#define IS_REF0(a)         ((a) & MB_TYPE_REF0)
94#define IS_8x8DCT(a)       ((a) & MB_TYPE_8x8DCT)
95
96/* NAL unit types */
97enum {
98    NAL_SLICE=1,
99    NAL_DPA,
100    NAL_DPB,
101    NAL_DPC,
102    NAL_IDR_SLICE,
103    NAL_SEI,
104    NAL_SPS,
105    NAL_PPS,
106    NAL_AUD,
107    NAL_END_SEQUENCE,
108    NAL_END_STREAM,
109    NAL_FILLER_DATA,
110    NAL_SPS_EXT,
111    NAL_AUXILIARY_SLICE=19
112};
113
114/**
115 * SEI message types
116 */
117typedef enum {
118    SEI_BUFFERING_PERIOD             =  0, ///< buffering period (H.264, D.1.1)
119    SEI_TYPE_PIC_TIMING              =  1, ///< picture timing
120    SEI_TYPE_USER_DATA_UNREGISTERED  =  5, ///< unregistered user data
121    SEI_TYPE_RECOVERY_POINT          =  6  ///< recovery point (frame # to decoder sync)
122} SEI_Type;
123
124/**
125 * pic_struct in picture timing SEI message
126 */
127typedef enum {
128    SEI_PIC_STRUCT_FRAME             = 0, ///<  0: %frame
129    SEI_PIC_STRUCT_TOP_FIELD         = 1, ///<  1: top field
130    SEI_PIC_STRUCT_BOTTOM_FIELD      = 2, ///<  2: bottom field
131    SEI_PIC_STRUCT_TOP_BOTTOM        = 3, ///<  3: top field, bottom field, in that order
132    SEI_PIC_STRUCT_BOTTOM_TOP        = 4, ///<  4: bottom field, top field, in that order
133    SEI_PIC_STRUCT_TOP_BOTTOM_TOP    = 5, ///<  5: top field, bottom field, top field repeated, in that order
134    SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///<  6: bottom field, top field, bottom field repeated, in that order
135    SEI_PIC_STRUCT_FRAME_DOUBLING    = 7, ///<  7: %frame doubling
136    SEI_PIC_STRUCT_FRAME_TRIPLING    = 8  ///<  8: %frame tripling
137} SEI_PicStructType;
138
139/**
140 * Sequence parameter set
141 */
142typedef struct SPS{
143
144    int profile_idc;
145    int level_idc;
146    int chroma_format_idc;
147    int transform_bypass;              ///< qpprime_y_zero_transform_bypass_flag
148    int log2_max_frame_num;            ///< log2_max_frame_num_minus4 + 4
149    int poc_type;                      ///< pic_order_cnt_type
150    int log2_max_poc_lsb;              ///< log2_max_pic_order_cnt_lsb_minus4
151    int delta_pic_order_always_zero_flag;
152    int offset_for_non_ref_pic;
153    int offset_for_top_to_bottom_field;
154    int poc_cycle_length;              ///< num_ref_frames_in_pic_order_cnt_cycle
155    int ref_frame_count;               ///< num_ref_frames
156    int gaps_in_frame_num_allowed_flag;
157    int mb_width;                      ///< pic_width_in_mbs_minus1 + 1
158    int mb_height;                     ///< pic_height_in_map_units_minus1 + 1
159    int frame_mbs_only_flag;
160    int mb_aff;                        ///<mb_adaptive_frame_field_flag
161    int direct_8x8_inference_flag;
162    int crop;                   ///< frame_cropping_flag
163    unsigned int crop_left;            ///< frame_cropping_rect_left_offset
164    unsigned int crop_right;           ///< frame_cropping_rect_right_offset
165    unsigned int crop_top;             ///< frame_cropping_rect_top_offset
166    unsigned int crop_bottom;          ///< frame_cropping_rect_bottom_offset
167    int vui_parameters_present_flag;
168    AVRational sar;
169    int timing_info_present_flag;
170    uint32_t num_units_in_tick;
171    uint32_t time_scale;
172    int fixed_frame_rate_flag;
173    short offset_for_ref_frame[256]; //FIXME dyn aloc?
174    int bitstream_restriction_flag;
175    int num_reorder_frames;
176    int scaling_matrix_present;
177    uint8_t scaling_matrix4[6][16];
178    uint8_t scaling_matrix8[2][64];
179    int nal_hrd_parameters_present_flag;
180    int vcl_hrd_parameters_present_flag;
181    int pic_struct_present_flag;
182    int time_offset_length;
183    int cpb_cnt;                       ///< See H.264 E.1.2
184    int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 +1
185    int cpb_removal_delay_length;      ///< cpb_removal_delay_length_minus1 + 1
186    int dpb_output_delay_length;       ///< dpb_output_delay_length_minus1 + 1
187    int bit_depth_luma;                ///< bit_depth_luma_minus8 + 8
188    int bit_depth_chroma;              ///< bit_depth_chroma_minus8 + 8
189    int residual_color_transform_flag; ///< residual_colour_transform_flag
190}SPS;
191
192/**
193 * Picture parameter set
194 */
195typedef struct PPS{
196    unsigned int sps_id;
197    int cabac;                  ///< entropy_coding_mode_flag
198    int pic_order_present;      ///< pic_order_present_flag
199    int slice_group_count;      ///< num_slice_groups_minus1 + 1
200    int mb_slice_group_map_type;
201    unsigned int ref_count[2];  ///< num_ref_idx_l0/1_active_minus1 + 1
202    int weighted_pred;          ///< weighted_pred_flag
203    int weighted_bipred_idc;
204    int init_qp;                ///< pic_init_qp_minus26 + 26
205    int init_qs;                ///< pic_init_qs_minus26 + 26
206    int chroma_qp_index_offset[2];
207    int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
208    int constrained_intra_pred; ///< constrained_intra_pred_flag
209    int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
210    int transform_8x8_mode;     ///< transform_8x8_mode_flag
211    uint8_t scaling_matrix4[6][16];
212    uint8_t scaling_matrix8[2][64];
213    uint8_t chroma_qp_table[2][64];  ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
214    int chroma_qp_diff;
215}PPS;
216
217/**
218 * Memory management control operation opcode.
219 */
220typedef enum MMCOOpcode{
221    MMCO_END=0,
222    MMCO_SHORT2UNUSED,
223    MMCO_LONG2UNUSED,
224    MMCO_SHORT2LONG,
225    MMCO_SET_MAX_LONG,
226    MMCO_RESET,
227    MMCO_LONG,
228} MMCOOpcode;
229
230/**
231 * Memory management control operation.
232 */
233typedef struct MMCO{
234    MMCOOpcode opcode;
235    int short_pic_num;  ///< pic_num without wrapping (pic_num & max_pic_num)
236    int long_arg;       ///< index, pic_num, or num long refs depending on opcode
237} MMCO;
238
239/**
240 * H264Context
241 */
242typedef struct H264Context{
243    MpegEncContext s;
244    int nal_ref_idc;
245    int nal_unit_type;
246    uint8_t *rbsp_buffer[2];
247    unsigned int rbsp_buffer_size[2];
248
249    /**
250      * Used to parse AVC variant of h264
251      */
252    int is_avc; ///< this flag is != 0 if codec is avc1
253    int got_avcC; ///< flag used to parse avcC data only once
254    int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
255
256    int chroma_qp[2]; //QPc
257
258    int prev_mb_skipped;
259    int next_mb_skipped;
260
261    //prediction stuff
262    int chroma_pred_mode;
263    int intra16x16_pred_mode;
264
265    int top_mb_xy;
266    int left_mb_xy[2];
267
268    int8_t intra4x4_pred_mode_cache[5*8];
269    int8_t (*intra4x4_pred_mode)[8];
270    H264PredContext hpc;
271    unsigned int topleft_samples_available;
272    unsigned int top_samples_available;
273    unsigned int topright_samples_available;
274    unsigned int left_samples_available;
275    uint8_t (*top_borders[2])[16+2*8];
276    uint8_t left_border[2*(17+2*9)];
277
278    /**
279     * non zero coeff count cache.
280     * is 64 if not available.
281     */
282    DECLARE_ALIGNED_8(uint8_t, non_zero_count_cache[6*8]);
283    uint8_t (*non_zero_count)[16];
284
285    /**
286     * Motion vector cache.
287     */
288    DECLARE_ALIGNED_8(int16_t, mv_cache[2][5*8][2]);
289    DECLARE_ALIGNED_8(int8_t, ref_cache[2][5*8]);
290#define LIST_NOT_USED -1 //FIXME rename?
291#define PART_NOT_AVAILABLE -2
292
293    /**
294     * is 1 if the specific list MV&references are set to 0,0,-2.
295     */
296    int mv_cache_clean[2];
297
298    /**
299     * number of neighbors (top and/or left) that used 8x8 dct
300     */
301    int neighbor_transform_size;
302
303    /**
304     * block_offset[ 0..23] for frame macroblocks
305     * block_offset[24..47] for field macroblocks
306     */
307    int block_offset[2*(16+8)];
308
309    uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
310    uint32_t *mb2b8_xy;
311    int b_stride; //FIXME use s->b4_stride
312    int b8_stride;
313
314    int mb_linesize;   ///< may be equal to s->linesize or s->linesize*2, for mbaff
315    int mb_uvlinesize;
316
317    int emu_edge_width;
318    int emu_edge_height;
319
320    int halfpel_flag;
321    int thirdpel_flag;
322
323    int unknown_svq3_flag;
324    int next_slice_index;
325
326    SPS *sps_buffers[MAX_SPS_COUNT];
327    SPS sps; ///< current sps
328
329    PPS *pps_buffers[MAX_PPS_COUNT];
330    /**
331     * current pps
332     */
333    PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
334
335    uint32_t dequant4_buffer[6][52][16];
336    uint32_t dequant8_buffer[2][52][64];
337    uint32_t (*dequant4_coeff[6])[16];
338    uint32_t (*dequant8_coeff[2])[64];
339    int dequant_coeff_pps;     ///< reinit tables when pps changes
340
341    int slice_num;
342    uint16_t *slice_table_base;
343    uint16_t *slice_table;     ///< slice_table_base + 2*mb_stride + 1
344    int slice_type;
345    int slice_type_nos;        ///< S free slice type (SI/SP are remapped to I/P)
346    int slice_type_fixed;
347
348    //interlacing specific flags
349    int mb_aff_frame;
350    int mb_field_decoding_flag;
351    int mb_mbaff;              ///< mb_aff_frame && mb_field_decoding_flag
352
353    DECLARE_ALIGNED_8(uint16_t, sub_mb_type[4]);
354
355    //POC stuff
356    int poc_lsb;
357    int poc_msb;
358    int delta_poc_bottom;
359    int delta_poc[2];
360    int frame_num;
361    int prev_poc_msb;             ///< poc_msb of the last reference pic for POC type 0
362    int prev_poc_lsb;             ///< poc_lsb of the last reference pic for POC type 0
363    int frame_num_offset;         ///< for POC type 2
364    int prev_frame_num_offset;    ///< for POC type 2
365    int prev_frame_num;           ///< frame_num of the last pic for POC type 1/2
366
367    /**
368     * frame_num for frames or 2*frame_num+1 for field pics.
369     */
370    int curr_pic_num;
371
372    /**
373     * max_frame_num or 2*max_frame_num for field pics.
374     */
375    int max_pic_num;
376
377    //Weighted pred stuff
378    int use_weight;
379    int use_weight_chroma;
380    int luma_log2_weight_denom;
381    int chroma_log2_weight_denom;
382    int luma_weight[2][48];
383    int luma_offset[2][48];
384    int chroma_weight[2][48][2];
385    int chroma_offset[2][48][2];
386    int implicit_weight[48][48];
387
388    //deblock
389    int deblocking_filter;         ///< disable_deblocking_filter_idc with 1<->0
390    int slice_alpha_c0_offset;
391    int slice_beta_offset;
392
393    int redundant_pic_count;
394
395    int direct_spatial_mv_pred;
396    int dist_scale_factor[16];
397    int dist_scale_factor_field[2][32];
398    int map_col_to_list0[2][16+32];
399    int map_col_to_list0_field[2][2][16+32];
400
401    /**
402     * num_ref_idx_l0/1_active_minus1 + 1
403     */
404    unsigned int ref_count[2];   ///< counts frames or fields, depending on current mb mode
405    unsigned int list_count;
406    Picture *short_ref[32];
407    Picture *long_ref[32];
408    Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
409    Picture ref_list[2][48];         /**< 0..15: frame refs, 16..47: mbaff field refs.
410                                          Reordered version of default_ref_list
411                                          according to picture reordering in slice header */
412    int ref2frm[MAX_SLICES][2][64];  ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
413    Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
414    int outputed_poc;
415
416    /**
417     * memory management control operations buffer.
418     */
419    MMCO mmco[MAX_MMCO_COUNT];
420    int mmco_index;
421
422    int long_ref_count;  ///< number of actual long term references
423    int short_ref_count; ///< number of actual short term references
424
425    //data partitioning
426    GetBitContext intra_gb;
427    GetBitContext inter_gb;
428    GetBitContext *intra_gb_ptr;
429    GetBitContext *inter_gb_ptr;
430
431    DECLARE_ALIGNED_16(DCTELEM, mb[16*24]);
432    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
433
434    /**
435     * Cabac
436     */
437    CABACContext cabac;
438    uint8_t      cabac_state[460];
439    int          cabac_init_idc;
440
441    /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
442    uint16_t     *cbp_table;
443    int cbp;
444    int top_cbp;
445    int left_cbp;
446    /* chroma_pred_mode for i4x4 or i16x16, else 0 */
447    uint8_t     *chroma_pred_mode_table;
448    int         last_qscale_diff;
449    int16_t     (*mvd_table[2])[2];
450    DECLARE_ALIGNED_8(int16_t, mvd_cache[2][5*8][2]);
451    uint8_t     *direct_table;
452    uint8_t     direct_cache[5*8];
453
454    uint8_t zigzag_scan[16];
455    uint8_t zigzag_scan8x8[64];
456    uint8_t zigzag_scan8x8_cavlc[64];
457    uint8_t field_scan[16];
458    uint8_t field_scan8x8[64];
459    uint8_t field_scan8x8_cavlc[64];
460    const uint8_t *zigzag_scan_q0;
461    const uint8_t *zigzag_scan8x8_q0;
462    const uint8_t *zigzag_scan8x8_cavlc_q0;
463    const uint8_t *field_scan_q0;
464    const uint8_t *field_scan8x8_q0;
465    const uint8_t *field_scan8x8_cavlc_q0;
466
467    int x264_build;
468
469    /**
470     * @defgroup multithreading Members for slice based multithreading
471     * @{
472     */
473    struct H264Context *thread_context[MAX_THREADS];
474
475    /**
476     * current slice number, used to initalize slice_num of each thread/context
477     */
478    int current_slice;
479
480    /**
481     * Max number of threads / contexts.
482     * This is equal to AVCodecContext.thread_count unless
483     * multithreaded decoding is impossible, in which case it is
484     * reduced to 1.
485     */
486    int max_contexts;
487
488    /**
489     *  1 if the single thread fallback warning has already been
490     *  displayed, 0 otherwise.
491     */
492    int single_decode_warning;
493
494    int last_slice_type;
495    /** @} */
496
497    int mb_xy;
498
499    uint32_t svq3_watermark_key;
500
501    /**
502     * pic_struct in picture timing SEI message
503     */
504    SEI_PicStructType sei_pic_struct;
505
506    /**
507     * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
508     */
509    int sei_dpb_output_delay;
510
511    /**
512     * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
513     */
514    int sei_cpb_removal_delay;
515
516    /**
517     * recovery_frame_cnt from SEI message
518     *
519     * Set to -1 if no recovery point SEI message found or to number of frames
520     * before playback synchronizes. Frames having recovery point are key
521     * frames.
522     */
523    int sei_recovery_frame_cnt;
524
525    int is_complex;
526
527    int luma_weight_flag[2];   ///< 7.4.3.2 luma_weight_lX_flag
528    int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
529
530    // Timestamp stuff
531    int sei_buffering_period_present;  ///< Buffering period SEI flag
532    int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
533}H264Context;
534
535/**
536 * Decode SEI
537 */
538int ff_h264_decode_sei(H264Context *h);
539
540/**
541 * Decode SPS
542 */
543int ff_h264_decode_seq_parameter_set(H264Context *h);
544
545/**
546 * Decode PPS
547 */
548int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
549
550/**
551 * Decodes a network abstraction layer unit.
552 * @param consumed is the number of bytes used as input
553 * @param length is the length of the array
554 * @param dst_length is the number of decoded bytes FIXME here or a decode rbsp tailing?
555 * @returns decoded bytes, might be src+1 if no escapes
556 */
557const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length);
558
559/**
560 * identifies the exact end of the bitstream
561 * @return the length of the trailing, or 0 if damaged
562 */
563int ff_h264_decode_rbsp_trailing(H264Context *h, const uint8_t *src);
564
565#endif /* AVCODEC_H264_H */
566