1/*
2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4 *
5 * This file is part of Libav.
6 *
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
23 * @file
24 * H.264 / AVC / MPEG4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28#ifndef AVCODEC_H264_H
29#define AVCODEC_H264_H
30
31#include "libavutil/intreadwrite.h"
32#include "dsputil.h"
33#include "cabac.h"
34#include "mpegvideo.h"
35#include "h264dsp.h"
36#include "h264pred.h"
37#include "rectangle.h"
38
39#define interlaced_dct interlaced_dct_is_a_bad_name
40#define mb_intra mb_intra_is_not_initialized_see_mb_type
41
42#define MAX_SPS_COUNT 32
43#define MAX_PPS_COUNT 256
44
45#define MAX_MMCO_COUNT 66
46
47#define MAX_DELAYED_PIC_COUNT 16
48
49/* Compiling in interlaced support reduces the speed
50 * of progressive decoding by about 2%. */
51#define ALLOW_INTERLACE
52
53#define FMO 0
54
55/**
56 * The maximum number of slices supported by the decoder.
57 * must be a power of 2
58 */
59#define MAX_SLICES 16
60
61#ifdef ALLOW_INTERLACE
62#define MB_MBAFF h->mb_mbaff
63#define MB_FIELD h->mb_field_decoding_flag
64#define FRAME_MBAFF h->mb_aff_frame
65#define FIELD_PICTURE (s->picture_structure != PICT_FRAME)
66#define LEFT_MBS 2
67#define LTOP 0
68#define LBOT 1
69#define LEFT(i) (i)
70#else
71#define MB_MBAFF 0
72#define MB_FIELD 0
73#define FRAME_MBAFF 0
74#define FIELD_PICTURE 0
75#undef  IS_INTERLACED
76#define IS_INTERLACED(mb_type) 0
77#define LEFT_MBS 1
78#define LTOP 0
79#define LBOT 0
80#define LEFT(i) 0
81#endif
82#define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
83
84#ifndef CABAC
85#define CABAC h->pps.cabac
86#endif
87
88#define CHROMA422 (h->sps.chroma_format_idc == 2)
89#define CHROMA444 (h->sps.chroma_format_idc == 3)
90
91#define EXTENDED_SAR          255
92
93#define MB_TYPE_REF0       MB_TYPE_ACPRED //dirty but it fits in 16 bit
94#define MB_TYPE_8x8DCT     0x01000000
95#define IS_REF0(a)         ((a) & MB_TYPE_REF0)
96#define IS_8x8DCT(a)       ((a) & MB_TYPE_8x8DCT)
97
98/**
99 * Value of Picture.reference when Picture is not a reference picture, but
100 * is held for delayed output.
101 */
102#define DELAYED_PIC_REF 4
103
104#define QP_MAX_NUM (51 + 2*6)           // The maximum supported qp
105
106/* NAL unit types */
107enum {
108    NAL_SLICE=1,
109    NAL_DPA,
110    NAL_DPB,
111    NAL_DPC,
112    NAL_IDR_SLICE,
113    NAL_SEI,
114    NAL_SPS,
115    NAL_PPS,
116    NAL_AUD,
117    NAL_END_SEQUENCE,
118    NAL_END_STREAM,
119    NAL_FILLER_DATA,
120    NAL_SPS_EXT,
121    NAL_AUXILIARY_SLICE=19
122};
123
124/**
125 * SEI message types
126 */
127typedef enum {
128    SEI_BUFFERING_PERIOD             =  0, ///< buffering period (H.264, D.1.1)
129    SEI_TYPE_PIC_TIMING              =  1, ///< picture timing
130    SEI_TYPE_USER_DATA_UNREGISTERED  =  5, ///< unregistered user data
131    SEI_TYPE_RECOVERY_POINT          =  6  ///< recovery point (frame # to decoder sync)
132} SEI_Type;
133
134/**
135 * pic_struct in picture timing SEI message
136 */
137typedef enum {
138    SEI_PIC_STRUCT_FRAME             = 0, ///<  0: %frame
139    SEI_PIC_STRUCT_TOP_FIELD         = 1, ///<  1: top field
140    SEI_PIC_STRUCT_BOTTOM_FIELD      = 2, ///<  2: bottom field
141    SEI_PIC_STRUCT_TOP_BOTTOM        = 3, ///<  3: top field, bottom field, in that order
142    SEI_PIC_STRUCT_BOTTOM_TOP        = 4, ///<  4: bottom field, top field, in that order
143    SEI_PIC_STRUCT_TOP_BOTTOM_TOP    = 5, ///<  5: top field, bottom field, top field repeated, in that order
144    SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///<  6: bottom field, top field, bottom field repeated, in that order
145    SEI_PIC_STRUCT_FRAME_DOUBLING    = 7, ///<  7: %frame doubling
146    SEI_PIC_STRUCT_FRAME_TRIPLING    = 8  ///<  8: %frame tripling
147} SEI_PicStructType;
148
149/**
150 * Sequence parameter set
151 */
152typedef struct SPS{
153
154    int profile_idc;
155    int level_idc;
156    int chroma_format_idc;
157    int transform_bypass;              ///< qpprime_y_zero_transform_bypass_flag
158    int log2_max_frame_num;            ///< log2_max_frame_num_minus4 + 4
159    int poc_type;                      ///< pic_order_cnt_type
160    int log2_max_poc_lsb;              ///< log2_max_pic_order_cnt_lsb_minus4
161    int delta_pic_order_always_zero_flag;
162    int offset_for_non_ref_pic;
163    int offset_for_top_to_bottom_field;
164    int poc_cycle_length;              ///< num_ref_frames_in_pic_order_cnt_cycle
165    int ref_frame_count;               ///< num_ref_frames
166    int gaps_in_frame_num_allowed_flag;
167    int mb_width;                      ///< pic_width_in_mbs_minus1 + 1
168    int mb_height;                     ///< pic_height_in_map_units_minus1 + 1
169    int frame_mbs_only_flag;
170    int mb_aff;                        ///<mb_adaptive_frame_field_flag
171    int direct_8x8_inference_flag;
172    int crop;                   ///< frame_cropping_flag
173    unsigned int crop_left;            ///< frame_cropping_rect_left_offset
174    unsigned int crop_right;           ///< frame_cropping_rect_right_offset
175    unsigned int crop_top;             ///< frame_cropping_rect_top_offset
176    unsigned int crop_bottom;          ///< frame_cropping_rect_bottom_offset
177    int vui_parameters_present_flag;
178    AVRational sar;
179    int video_signal_type_present_flag;
180    int full_range;
181    int colour_description_present_flag;
182    enum AVColorPrimaries color_primaries;
183    enum AVColorTransferCharacteristic color_trc;
184    enum AVColorSpace colorspace;
185    int timing_info_present_flag;
186    uint32_t num_units_in_tick;
187    uint32_t time_scale;
188    int fixed_frame_rate_flag;
189    short offset_for_ref_frame[256]; //FIXME dyn aloc?
190    int bitstream_restriction_flag;
191    int num_reorder_frames;
192    int scaling_matrix_present;
193    uint8_t scaling_matrix4[6][16];
194    uint8_t scaling_matrix8[6][64];
195    int nal_hrd_parameters_present_flag;
196    int vcl_hrd_parameters_present_flag;
197    int pic_struct_present_flag;
198    int time_offset_length;
199    int cpb_cnt;                       ///< See H.264 E.1.2
200    int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 +1
201    int cpb_removal_delay_length;      ///< cpb_removal_delay_length_minus1 + 1
202    int dpb_output_delay_length;       ///< dpb_output_delay_length_minus1 + 1
203    int bit_depth_luma;                ///< bit_depth_luma_minus8 + 8
204    int bit_depth_chroma;              ///< bit_depth_chroma_minus8 + 8
205    int residual_color_transform_flag; ///< residual_colour_transform_flag
206    int constraint_set_flags;          ///< constraint_set[0-3]_flag
207}SPS;
208
209/**
210 * Picture parameter set
211 */
212typedef struct PPS{
213    unsigned int sps_id;
214    int cabac;                  ///< entropy_coding_mode_flag
215    int pic_order_present;      ///< pic_order_present_flag
216    int slice_group_count;      ///< num_slice_groups_minus1 + 1
217    int mb_slice_group_map_type;
218    unsigned int ref_count[2];  ///< num_ref_idx_l0/1_active_minus1 + 1
219    int weighted_pred;          ///< weighted_pred_flag
220    int weighted_bipred_idc;
221    int init_qp;                ///< pic_init_qp_minus26 + 26
222    int init_qs;                ///< pic_init_qs_minus26 + 26
223    int chroma_qp_index_offset[2];
224    int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
225    int constrained_intra_pred; ///< constrained_intra_pred_flag
226    int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
227    int transform_8x8_mode;     ///< transform_8x8_mode_flag
228    uint8_t scaling_matrix4[6][16];
229    uint8_t scaling_matrix8[6][64];
230    uint8_t chroma_qp_table[2][64];  ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
231    int chroma_qp_diff;
232}PPS;
233
234/**
235 * Memory management control operation opcode.
236 */
237typedef enum MMCOOpcode{
238    MMCO_END=0,
239    MMCO_SHORT2UNUSED,
240    MMCO_LONG2UNUSED,
241    MMCO_SHORT2LONG,
242    MMCO_SET_MAX_LONG,
243    MMCO_RESET,
244    MMCO_LONG,
245} MMCOOpcode;
246
247/**
248 * Memory management control operation.
249 */
250typedef struct MMCO{
251    MMCOOpcode opcode;
252    int short_pic_num;  ///< pic_num without wrapping (pic_num & max_pic_num)
253    int long_arg;       ///< index, pic_num, or num long refs depending on opcode
254} MMCO;
255
256/**
257 * H264Context
258 */
259typedef struct H264Context{
260    MpegEncContext s;
261    H264DSPContext h264dsp;
262    int pixel_shift;    ///< 0 for 8-bit H264, 1 for high-bit-depth H264
263    int chroma_qp[2]; //QPc
264
265    int qp_thresh;      ///< QP threshold to skip loopfilter
266
267    int prev_mb_skipped;
268    int next_mb_skipped;
269
270    //prediction stuff
271    int chroma_pred_mode;
272    int intra16x16_pred_mode;
273
274    int topleft_mb_xy;
275    int top_mb_xy;
276    int topright_mb_xy;
277    int left_mb_xy[LEFT_MBS];
278
279    int topleft_type;
280    int top_type;
281    int topright_type;
282    int left_type[LEFT_MBS];
283
284    const uint8_t * left_block;
285    int topleft_partition;
286
287    int8_t intra4x4_pred_mode_cache[5*8];
288    int8_t (*intra4x4_pred_mode);
289    H264PredContext hpc;
290    unsigned int topleft_samples_available;
291    unsigned int top_samples_available;
292    unsigned int topright_samples_available;
293    unsigned int left_samples_available;
294    uint8_t (*top_borders[2])[(16*3)*2];
295
296    /**
297     * non zero coeff count cache.
298     * is 64 if not available.
299     */
300    DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15*8];
301
302    uint8_t (*non_zero_count)[48];
303
304    /**
305     * Motion vector cache.
306     */
307    DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5*8][2];
308    DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5*8];
309#define LIST_NOT_USED -1 //FIXME rename?
310#define PART_NOT_AVAILABLE -2
311
312    /**
313     * number of neighbors (top and/or left) that used 8x8 dct
314     */
315    int neighbor_transform_size;
316
317    /**
318     * block_offset[ 0..23] for frame macroblocks
319     * block_offset[24..47] for field macroblocks
320     */
321    int block_offset[2*(16*3)];
322
323    uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
324    uint32_t *mb2br_xy;
325    int b_stride; //FIXME use s->b4_stride
326
327    int mb_linesize;   ///< may be equal to s->linesize or s->linesize*2, for mbaff
328    int mb_uvlinesize;
329
330    int emu_edge_width;
331    int emu_edge_height;
332
333    SPS sps; ///< current sps
334
335    /**
336     * current pps
337     */
338    PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
339
340    uint32_t dequant4_buffer[6][QP_MAX_NUM+1][16]; //FIXME should these be moved down?
341    uint32_t dequant8_buffer[6][QP_MAX_NUM+1][64];
342    uint32_t (*dequant4_coeff[6])[16];
343    uint32_t (*dequant8_coeff[6])[64];
344
345    int slice_num;
346    uint16_t *slice_table;     ///< slice_table_base + 2*mb_stride + 1
347    int slice_type;
348    int slice_type_nos;        ///< S free slice type (SI/SP are remapped to I/P)
349    int slice_type_fixed;
350
351    //interlacing specific flags
352    int mb_aff_frame;
353    int mb_field_decoding_flag;
354    int mb_mbaff;              ///< mb_aff_frame && mb_field_decoding_flag
355
356    DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
357
358    //Weighted pred stuff
359    int use_weight;
360    int use_weight_chroma;
361    int luma_log2_weight_denom;
362    int chroma_log2_weight_denom;
363    //The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
364    int luma_weight[48][2][2];
365    int chroma_weight[48][2][2][2];
366    int implicit_weight[48][48][2];
367
368    int direct_spatial_mv_pred;
369    int col_parity;
370    int col_fieldoff;
371    int dist_scale_factor[16];
372    int dist_scale_factor_field[2][32];
373    int map_col_to_list0[2][16+32];
374    int map_col_to_list0_field[2][2][16+32];
375
376    /**
377     * num_ref_idx_l0/1_active_minus1 + 1
378     */
379    unsigned int ref_count[2];   ///< counts frames or fields, depending on current mb mode
380    unsigned int list_count;
381    uint8_t *list_counts;            ///< Array of list_count per MB specifying the slice type
382    Picture ref_list[2][48];         /**< 0..15: frame refs, 16..47: mbaff field refs.
383                                          Reordered version of default_ref_list
384                                          according to picture reordering in slice header */
385    int ref2frm[MAX_SLICES][2][64];  ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
386
387    //data partitioning
388    GetBitContext intra_gb;
389    GetBitContext inter_gb;
390    GetBitContext *intra_gb_ptr;
391    GetBitContext *inter_gb_ptr;
392
393    DECLARE_ALIGNED(16, DCTELEM, mb)[16*48*2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
394    DECLARE_ALIGNED(16, DCTELEM, mb_luma_dc)[3][16*2];
395    DCTELEM mb_padding[256*2];        ///< 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
396
397    /**
398     * Cabac
399     */
400    CABACContext cabac;
401    uint8_t      cabac_state[1024];
402
403    /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
404    uint16_t     *cbp_table;
405    int cbp;
406    int top_cbp;
407    int left_cbp;
408    /* chroma_pred_mode for i4x4 or i16x16, else 0 */
409    uint8_t     *chroma_pred_mode_table;
410    int         last_qscale_diff;
411    uint8_t     (*mvd_table[2])[2];
412    DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5*8][2];
413    uint8_t     *direct_table;
414    uint8_t     direct_cache[5*8];
415
416    uint8_t zigzag_scan[16];
417    uint8_t zigzag_scan8x8[64];
418    uint8_t zigzag_scan8x8_cavlc[64];
419    uint8_t field_scan[16];
420    uint8_t field_scan8x8[64];
421    uint8_t field_scan8x8_cavlc[64];
422    const uint8_t *zigzag_scan_q0;
423    const uint8_t *zigzag_scan8x8_q0;
424    const uint8_t *zigzag_scan8x8_cavlc_q0;
425    const uint8_t *field_scan_q0;
426    const uint8_t *field_scan8x8_q0;
427    const uint8_t *field_scan8x8_cavlc_q0;
428
429    int x264_build;
430
431    int mb_xy;
432
433    int is_complex;
434
435    //deblock
436    int deblocking_filter;         ///< disable_deblocking_filter_idc with 1<->0
437    int slice_alpha_c0_offset;
438    int slice_beta_offset;
439
440//=============================================================
441    //Things below are not used in the MB or more inner code
442
443    int nal_ref_idc;
444    int nal_unit_type;
445    uint8_t *rbsp_buffer[2];
446    unsigned int rbsp_buffer_size[2];
447
448    /**
449     * Used to parse AVC variant of h264
450     */
451    int is_avc; ///< this flag is != 0 if codec is avc1
452    int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
453    int got_first; ///< this flag is != 0 if we've parsed a frame
454
455    SPS *sps_buffers[MAX_SPS_COUNT];
456    PPS *pps_buffers[MAX_PPS_COUNT];
457
458    int dequant_coeff_pps;     ///< reinit tables when pps changes
459
460    uint16_t *slice_table_base;
461
462
463    //POC stuff
464    int poc_lsb;
465    int poc_msb;
466    int delta_poc_bottom;
467    int delta_poc[2];
468    int frame_num;
469    int prev_poc_msb;             ///< poc_msb of the last reference pic for POC type 0
470    int prev_poc_lsb;             ///< poc_lsb of the last reference pic for POC type 0
471    int frame_num_offset;         ///< for POC type 2
472    int prev_frame_num_offset;    ///< for POC type 2
473    int prev_frame_num;           ///< frame_num of the last pic for POC type 1/2
474
475    /**
476     * frame_num for frames or 2*frame_num+1 for field pics.
477     */
478    int curr_pic_num;
479
480    /**
481     * max_frame_num or 2*max_frame_num for field pics.
482     */
483    int max_pic_num;
484
485    int redundant_pic_count;
486
487    Picture *short_ref[32];
488    Picture *long_ref[32];
489    Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
490    Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
491    int last_pocs[MAX_DELAYED_PIC_COUNT];
492    Picture *next_output_pic;
493    int outputed_poc;
494    int next_outputed_poc;
495
496    /**
497     * memory management control operations buffer.
498     */
499    MMCO mmco[MAX_MMCO_COUNT];
500    int mmco_index;
501    int mmco_reset;
502
503    int long_ref_count;  ///< number of actual long term references
504    int short_ref_count; ///< number of actual short term references
505
506    int          cabac_init_idc;
507
508    /**
509     * @name Members for slice based multithreading
510     * @{
511     */
512    struct H264Context *thread_context[MAX_THREADS];
513
514    /**
515     * current slice number, used to initalize slice_num of each thread/context
516     */
517    int current_slice;
518
519    /**
520     * Max number of threads / contexts.
521     * This is equal to AVCodecContext.thread_count unless
522     * multithreaded decoding is impossible, in which case it is
523     * reduced to 1.
524     */
525    int max_contexts;
526
527    /**
528     *  1 if the single thread fallback warning has already been
529     *  displayed, 0 otherwise.
530     */
531    int single_decode_warning;
532
533    int last_slice_type;
534    /** @} */
535
536    /**
537     * pic_struct in picture timing SEI message
538     */
539    SEI_PicStructType sei_pic_struct;
540
541    /**
542     * Complement sei_pic_struct
543     * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
544     * However, soft telecined frames may have these values.
545     * This is used in an attempt to flag soft telecine progressive.
546     */
547    int prev_interlaced_frame;
548
549    /**
550     * Bit set of clock types for fields/frames in picture timing SEI message.
551     * For each found ct_type, appropriate bit is set (e.g., bit 1 for
552     * interlaced).
553     */
554    int sei_ct_type;
555
556    /**
557     * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
558     */
559    int sei_dpb_output_delay;
560
561    /**
562     * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
563     */
564    int sei_cpb_removal_delay;
565
566    /**
567     * recovery_frame_cnt from SEI message
568     *
569     * Set to -1 if no recovery point SEI message found or to number of frames
570     * before playback synchronizes. Frames having recovery point are key
571     * frames.
572     */
573    int sei_recovery_frame_cnt;
574
575    int luma_weight_flag[2];   ///< 7.4.3.2 luma_weight_lX_flag
576    int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
577
578    // Timestamp stuff
579    int sei_buffering_period_present;  ///< Buffering period SEI flag
580    int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
581
582    int cur_chroma_format_idc;
583}H264Context;
584
585
586extern const uint8_t ff_h264_chroma_qp[3][QP_MAX_NUM+1]; ///< One chroma qp table for each supported bit depth (8, 9, 10).
587
588/**
589 * Decode SEI
590 */
591int ff_h264_decode_sei(H264Context *h);
592
593/**
594 * Decode SPS
595 */
596int ff_h264_decode_seq_parameter_set(H264Context *h);
597
598/**
599 * compute profile from sps
600 */
601int ff_h264_get_profile(SPS *sps);
602
603/**
604 * Decode PPS
605 */
606int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
607
608/**
609 * Decode a network abstraction layer unit.
610 * @param consumed is the number of bytes used as input
611 * @param length is the length of the array
612 * @param dst_length is the number of decoded bytes FIXME here or a decode rbsp tailing?
613 * @return decoded bytes, might be src+1 if no escapes
614 */
615const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length);
616
617/**
618 * Free any data that may have been allocated in the H264 context like SPS, PPS etc.
619 */
620av_cold void ff_h264_free_context(H264Context *h);
621
622/**
623 * Reconstruct bitstream slice_type.
624 */
625int ff_h264_get_slice_type(const H264Context *h);
626
627/**
628 * Allocate tables.
629 * needs width/height
630 */
631int ff_h264_alloc_tables(H264Context *h);
632
633/**
634 * Fill the default_ref_list.
635 */
636int ff_h264_fill_default_ref_list(H264Context *h);
637
638int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
639void ff_h264_fill_mbaff_ref_list(H264Context *h);
640void ff_h264_remove_all_refs(H264Context *h);
641
642/**
643 * Execute the reference picture marking (memory management control operations).
644 */
645int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
646
647int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb);
648
649void ff_generate_sliding_window_mmcos(H264Context *h);
650
651
652/**
653 * Check if the top & left blocks are available if needed & change the dc mode so it only uses the available blocks.
654 */
655int ff_h264_check_intra4x4_pred_mode(H264Context *h);
656
657/**
658 * Check if the top & left blocks are available if needed & change the dc mode so it only uses the available blocks.
659 */
660int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
661
662void ff_h264_hl_decode_mb(H264Context *h);
663int ff_h264_frame_start(H264Context *h);
664int ff_h264_decode_extradata(H264Context *h);
665av_cold int ff_h264_decode_init(AVCodecContext *avctx);
666av_cold int ff_h264_decode_end(AVCodecContext *avctx);
667av_cold void ff_h264_decode_init_vlc(void);
668
669/**
670 * Decode a macroblock
671 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR if an error is noticed
672 */
673int ff_h264_decode_mb_cavlc(H264Context *h);
674
675/**
676 * Decode a CABAC coded macroblock
677 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR if an error is noticed
678 */
679int ff_h264_decode_mb_cabac(H264Context *h);
680
681void ff_h264_init_cabac_states(H264Context *h);
682
683void ff_h264_direct_dist_scale_factor(H264Context * const h);
684void ff_h264_direct_ref_list_init(H264Context * const h);
685void ff_h264_pred_direct_motion(H264Context * const h, int *mb_type);
686
687void 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);
688void 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);
689
690/**
691 * Reset SEI values at the beginning of the frame.
692 *
693 * @param h H.264 context.
694 */
695void ff_h264_reset_sei(H264Context *h);
696
697
698/*
699o-o o-o
700 / / /
701o-o o-o
702 ,---'
703o-o o-o
704 / / /
705o-o o-o
706*/
707
708/* Scan8 organization:
709 *    0 1 2 3 4 5 6 7
710 * 0  DY    y y y y y
711 * 1        y Y Y Y Y
712 * 2        y Y Y Y Y
713 * 3        y Y Y Y Y
714 * 4        y Y Y Y Y
715 * 5  DU    u u u u u
716 * 6        u U U U U
717 * 7        u U U U U
718 * 8        u U U U U
719 * 9        u U U U U
720 * 10 DV    v v v v v
721 * 11       v V V V V
722 * 12       v V V V V
723 * 13       v V V V V
724 * 14       v V V V V
725 * DY/DU/DV are for luma/chroma DC.
726 */
727
728#define LUMA_DC_BLOCK_INDEX   48
729#define CHROMA_DC_BLOCK_INDEX 49
730
731//This table must be here because scan8[constant] must be known at compiletime
732static const uint8_t scan8[16*3 + 3]={
733 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8,
734 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8,
735 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8,
736 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8,
737 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8,
738 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8,
739 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8,
740 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8,
741 4+11*8, 5+11*8, 4+12*8, 5+12*8,
742 6+11*8, 7+11*8, 6+12*8, 7+12*8,
743 4+13*8, 5+13*8, 4+14*8, 5+14*8,
744 6+13*8, 7+13*8, 6+14*8, 7+14*8,
745 0+ 0*8, 0+ 5*8, 0+10*8
746};
747
748static av_always_inline uint32_t pack16to32(int a, int b){
749#if HAVE_BIGENDIAN
750   return (b&0xFFFF) + (a<<16);
751#else
752   return (a&0xFFFF) + (b<<16);
753#endif
754}
755
756static av_always_inline uint16_t pack8to16(int a, int b){
757#if HAVE_BIGENDIAN
758   return (b&0xFF) + (a<<8);
759#else
760   return (a&0xFF) + (b<<8);
761#endif
762}
763
764/**
765 * Get the chroma qp.
766 */
767static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale){
768    return h->pps.chroma_qp_table[t][qscale];
769}
770
771/**
772 * Get the predicted intra4x4 prediction mode.
773 */
774static av_always_inline int pred_intra_mode(H264Context *h, int n){
775    const int index8= scan8[n];
776    const int left= h->intra4x4_pred_mode_cache[index8 - 1];
777    const int top = h->intra4x4_pred_mode_cache[index8 - 8];
778    const int min= FFMIN(left, top);
779
780    tprintf(h->s.avctx, "mode:%d %d min:%d\n", left ,top, min);
781
782    if(min<0) return DC_PRED;
783    else      return min;
784}
785
786static av_always_inline void write_back_intra_pred_mode(H264Context *h){
787    int8_t *i4x4= h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
788    int8_t *i4x4_cache= h->intra4x4_pred_mode_cache;
789
790    AV_COPY32(i4x4, i4x4_cache + 4 + 8*4);
791    i4x4[4]= i4x4_cache[7+8*3];
792    i4x4[5]= i4x4_cache[7+8*2];
793    i4x4[6]= i4x4_cache[7+8*1];
794}
795
796static av_always_inline void write_back_non_zero_count(H264Context *h){
797    const int mb_xy= h->mb_xy;
798    uint8_t *nnz = h->non_zero_count[mb_xy];
799    uint8_t *nnz_cache = h->non_zero_count_cache;
800
801    AV_COPY32(&nnz[ 0], &nnz_cache[4+8* 1]);
802    AV_COPY32(&nnz[ 4], &nnz_cache[4+8* 2]);
803    AV_COPY32(&nnz[ 8], &nnz_cache[4+8* 3]);
804    AV_COPY32(&nnz[12], &nnz_cache[4+8* 4]);
805    AV_COPY32(&nnz[16], &nnz_cache[4+8* 6]);
806    AV_COPY32(&nnz[20], &nnz_cache[4+8* 7]);
807    AV_COPY32(&nnz[32], &nnz_cache[4+8*11]);
808    AV_COPY32(&nnz[36], &nnz_cache[4+8*12]);
809
810    if(!h->s.chroma_y_shift){
811        AV_COPY32(&nnz[24], &nnz_cache[4+8* 8]);
812        AV_COPY32(&nnz[28], &nnz_cache[4+8* 9]);
813        AV_COPY32(&nnz[40], &nnz_cache[4+8*13]);
814        AV_COPY32(&nnz[44], &nnz_cache[4+8*14]);
815    }
816}
817
818static av_always_inline void write_back_motion_list(H264Context *h, MpegEncContext * const s, int b_stride,
819                                                    int b_xy, int b8_xy, int mb_type, int list )
820{
821    int16_t (*mv_dst)[2] = &s->current_picture.f.motion_val[list][b_xy];
822    int16_t (*mv_src)[2] = &h->mv_cache[list][scan8[0]];
823    AV_COPY128(mv_dst + 0*b_stride, mv_src + 8*0);
824    AV_COPY128(mv_dst + 1*b_stride, mv_src + 8*1);
825    AV_COPY128(mv_dst + 2*b_stride, mv_src + 8*2);
826    AV_COPY128(mv_dst + 3*b_stride, mv_src + 8*3);
827    if( CABAC ) {
828        uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8*h->mb_xy : h->mb2br_xy[h->mb_xy]];
829        uint8_t (*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
830        if(IS_SKIP(mb_type))
831            AV_ZERO128(mvd_dst);
832        else{
833            AV_COPY64(mvd_dst, mvd_src + 8*3);
834            AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8*0);
835            AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8*1);
836            AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8*2);
837        }
838    }
839
840    {
841        int8_t *ref_index = &s->current_picture.f.ref_index[list][b8_xy];
842        int8_t *ref_cache = h->ref_cache[list];
843        ref_index[0+0*2]= ref_cache[scan8[0]];
844        ref_index[1+0*2]= ref_cache[scan8[4]];
845        ref_index[0+1*2]= ref_cache[scan8[8]];
846        ref_index[1+1*2]= ref_cache[scan8[12]];
847    }
848}
849
850static av_always_inline void write_back_motion(H264Context *h, int mb_type){
851    MpegEncContext * const s = &h->s;
852    const int b_stride = h->b_stride;
853    const int b_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride; //try mb2b(8)_xy
854    const int b8_xy= 4*h->mb_xy;
855
856    if(USES_LIST(mb_type, 0)){
857        write_back_motion_list(h, s, b_stride, b_xy, b8_xy, mb_type, 0);
858    }else{
859        fill_rectangle(&s->current_picture.f.ref_index[0][b8_xy],
860                       2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
861    }
862    if(USES_LIST(mb_type, 1)){
863        write_back_motion_list(h, s, b_stride, b_xy, b8_xy, mb_type, 1);
864    }
865
866    if(h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC){
867        if(IS_8X8(mb_type)){
868            uint8_t *direct_table = &h->direct_table[4*h->mb_xy];
869            direct_table[1] = h->sub_mb_type[1]>>1;
870            direct_table[2] = h->sub_mb_type[2]>>1;
871            direct_table[3] = h->sub_mb_type[3]>>1;
872        }
873    }
874}
875
876static av_always_inline int get_dct8x8_allowed(H264Context *h){
877    if(h->sps.direct_8x8_inference_flag)
878        return !(AV_RN64A(h->sub_mb_type) & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8                )*0x0001000100010001ULL));
879    else
880        return !(AV_RN64A(h->sub_mb_type) & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8|MB_TYPE_DIRECT2)*0x0001000100010001ULL));
881}
882
883#endif /* AVCODEC_H264_H */
884