FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
h264.h
Go to the documentation of this file.
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
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 "cabac.h"
33 #include "error_resilience.h"
34 #include "get_bits.h"
35 #include "h264chroma.h"
36 #include "h264dsp.h"
37 #include "h264pred.h"
38 #include "h264qpel.h"
39 #include "internal.h" // for avpriv_find_start_code()
40 #include "mpegutils.h"
41 #include "parser.h"
42 #include "qpeldsp.h"
43 #include "rectangle.h"
44 #include "videodsp.h"
45 
46 #define H264_MAX_PICTURE_COUNT 36
47 #define H264_MAX_THREADS 32
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 #define MAX_MBPAIR_SIZE (256*1024) // a tighter bound could be calculated if someone cares about a few bytes
57 
58 /* Compiling in interlaced support reduces the speed
59  * of progressive decoding by about 2%. */
60 #define ALLOW_INTERLACE
61 
62 #define FMO 0
63 
64 /**
65  * The maximum number of slices supported by the decoder.
66  * must be a power of 2
67  */
68 #define MAX_SLICES 32
69 
70 #ifdef ALLOW_INTERLACE
71 #define MB_MBAFF(h) (h)->mb_mbaff
72 #define MB_FIELD(h) (h)->mb_field_decoding_flag
73 #define FRAME_MBAFF(h) (h)->mb_aff_frame
74 #define FIELD_PICTURE(h) ((h)->picture_structure != PICT_FRAME)
75 #define LEFT_MBS 2
76 #define LTOP 0
77 #define LBOT 1
78 #define LEFT(i) (i)
79 #else
80 #define MB_MBAFF(h) 0
81 #define MB_FIELD(h) 0
82 #define FRAME_MBAFF(h) 0
83 #define FIELD_PICTURE(h) 0
84 #undef IS_INTERLACED
85 #define IS_INTERLACED(mb_type) 0
86 #define LEFT_MBS 1
87 #define LTOP 0
88 #define LBOT 0
89 #define LEFT(i) 0
90 #endif
91 #define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
92 
93 #ifndef CABAC
94 #define CABAC(h) (h)->pps.cabac
95 #endif
96 
97 #define CHROMA(h) ((h)->sps.chroma_format_idc)
98 #define CHROMA422(h) ((h)->sps.chroma_format_idc == 2)
99 #define CHROMA444(h) ((h)->sps.chroma_format_idc == 3)
100 
101 #define EXTENDED_SAR 255
102 
103 #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
104 #define MB_TYPE_8x8DCT 0x01000000
105 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
106 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
107 
108 #define QP_MAX_NUM (51 + 6*6) // The maximum supported qp
109 
110 /* NAL unit types */
111 enum {
112  NAL_SLICE = 1,
113  NAL_DPA = 2,
114  NAL_DPB = 3,
115  NAL_DPC = 4,
116  NAL_IDR_SLICE = 5,
117  NAL_SEI = 6,
118  NAL_SPS = 7,
119  NAL_PPS = 8,
120  NAL_AUD = 9,
121  NAL_END_SEQUENCE = 10,
122  NAL_END_STREAM = 11,
123  NAL_FILLER_DATA = 12,
124  NAL_SPS_EXT = 13,
125  NAL_AUXILIARY_SLICE = 19,
126  NAL_FF_IGNORE = 0xff0f001,
127 };
128 
129 /**
130  * SEI message types
131  */
132 typedef enum {
133  SEI_TYPE_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
134  SEI_TYPE_PIC_TIMING = 1, ///< picture timing
135  SEI_TYPE_USER_DATA_ITU_T_T35 = 4, ///< user data registered by ITU-T Recommendation T.35
136  SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
137  SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
138  SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
139  SEI_TYPE_DISPLAY_ORIENTATION = 47, ///< display orientation
140 } SEI_Type;
141 
142 /**
143  * pic_struct in picture timing SEI message
144  */
145 typedef enum {
146  SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
147  SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
148  SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
149  SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
150  SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
151  SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
152  SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
153  SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
154  SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
155 } SEI_PicStructType;
156 
157 /**
158  * frame_packing_arrangement types
159  */
160 typedef enum {
161  SEI_FPA_TYPE_CHECKERBOARD = 0,
162  SEI_FPA_TYPE_INTERLEAVE_COLUMN = 1,
163  SEI_FPA_TYPE_INTERLEAVE_ROW = 2,
164  SEI_FPA_TYPE_SIDE_BY_SIDE = 3,
165  SEI_FPA_TYPE_TOP_BOTTOM = 4,
166  SEI_FPA_TYPE_INTERLEAVE_TEMPORAL = 5,
167  SEI_FPA_TYPE_2D = 6,
168 } SEI_FpaType;
169 
170 /**
171  * Sequence parameter set
172  */
173 typedef struct SPS {
174  unsigned int sps_id;
175  int profile_idc;
176  int level_idc;
177  int chroma_format_idc;
178  int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
179  int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
180  int poc_type; ///< pic_order_cnt_type
181  int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
182  int delta_pic_order_always_zero_flag;
183  int offset_for_non_ref_pic;
184  int offset_for_top_to_bottom_field;
185  int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
186  int ref_frame_count; ///< num_ref_frames
187  int gaps_in_frame_num_allowed_flag;
188  int mb_width; ///< pic_width_in_mbs_minus1 + 1
189  int mb_height; ///< pic_height_in_map_units_minus1 + 1
190  int frame_mbs_only_flag;
191  int mb_aff; ///< mb_adaptive_frame_field_flag
192  int direct_8x8_inference_flag;
193  int crop; ///< frame_cropping_flag
194 
195  /* those 4 are already in luma samples */
196  unsigned int crop_left; ///< frame_cropping_rect_left_offset
197  unsigned int crop_right; ///< frame_cropping_rect_right_offset
198  unsigned int crop_top; ///< frame_cropping_rect_top_offset
199  unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
200  int vui_parameters_present_flag;
201  AVRational sar;
202  int video_signal_type_present_flag;
203  int full_range;
204  int colour_description_present_flag;
205  enum AVColorPrimaries color_primaries;
206  enum AVColorTransferCharacteristic color_trc;
207  enum AVColorSpace colorspace;
208  int timing_info_present_flag;
209  uint32_t num_units_in_tick;
210  uint32_t time_scale;
211  int fixed_frame_rate_flag;
212  short offset_for_ref_frame[256]; // FIXME dyn aloc?
213  int bitstream_restriction_flag;
214  int num_reorder_frames;
215  int scaling_matrix_present;
216  uint8_t scaling_matrix4[6][16];
217  uint8_t scaling_matrix8[6][64];
218  int nal_hrd_parameters_present_flag;
219  int vcl_hrd_parameters_present_flag;
220  int pic_struct_present_flag;
221  int time_offset_length;
222  int cpb_cnt; ///< See H.264 E.1.2
223  int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
224  int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
225  int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
226  int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
227  int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
228  int residual_color_transform_flag; ///< residual_colour_transform_flag
229  int constraint_set_flags; ///< constraint_set[0-3]_flag
230  int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
231 } SPS;
232 
233 /**
234  * Picture parameter set
235  */
236 typedef struct PPS {
237  unsigned int sps_id;
238  int cabac; ///< entropy_coding_mode_flag
239  int pic_order_present; ///< pic_order_present_flag
240  int slice_group_count; ///< num_slice_groups_minus1 + 1
241  int mb_slice_group_map_type;
242  unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
243  int weighted_pred; ///< weighted_pred_flag
244  int weighted_bipred_idc;
245  int init_qp; ///< pic_init_qp_minus26 + 26
246  int init_qs; ///< pic_init_qs_minus26 + 26
247  int chroma_qp_index_offset[2];
248  int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
249  int constrained_intra_pred; ///< constrained_intra_pred_flag
250  int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
251  int transform_8x8_mode; ///< transform_8x8_mode_flag
252  uint8_t scaling_matrix4[6][16];
253  uint8_t scaling_matrix8[6][64];
254  uint8_t chroma_qp_table[2][QP_MAX_NUM+1]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
255  int chroma_qp_diff;
256 } PPS;
257 
258 /**
259  * Frame Packing Arrangement Type
260  */
261 typedef struct FPA {
262  int frame_packing_arrangement_id;
263  int frame_packing_arrangement_cancel_flag; ///< is previous arrangement canceled, -1 if never received
264  SEI_FpaType frame_packing_arrangement_type;
265  int frame_packing_arrangement_repetition_period;
266  int content_interpretation_type;
267  int quincunx_sampling_flag;
268 } FPA;
269 
270 /**
271  * Memory management control operation opcode.
272  */
273 typedef enum MMCOOpcode {
274  MMCO_END = 0,
275  MMCO_SHORT2UNUSED,
276  MMCO_LONG2UNUSED,
277  MMCO_SHORT2LONG,
278  MMCO_SET_MAX_LONG,
279  MMCO_RESET,
280  MMCO_LONG,
281 } MMCOOpcode;
282 
283 /**
284  * Memory management control operation.
285  */
286 typedef struct MMCO {
287  MMCOOpcode opcode;
288  int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
289  int long_arg; ///< index, pic_num, or num long refs depending on opcode
290 } MMCO;
291 
292 typedef struct H264Picture {
293  struct AVFrame f;
294  uint8_t avframe_padding[1024]; // hack to allow linking to a avutil with larger AVFrame
295  ThreadFrame tf;
296 
297  AVBufferRef *qscale_table_buf;
298  int8_t *qscale_table;
299 
300  AVBufferRef *motion_val_buf[2];
301  int16_t (*motion_val[2])[2];
302 
303  AVBufferRef *mb_type_buf;
304  uint32_t *mb_type;
305 
306  AVBufferRef *hwaccel_priv_buf;
307  void *hwaccel_picture_private; ///< hardware accelerator private data
308 
309  AVBufferRef *ref_index_buf[2];
310  int8_t *ref_index[2];
311 
312  int field_poc[2]; ///< top/bottom POC
313  int poc; ///< frame POC
314  int frame_num; ///< frame_num (raw frame_num from slice header)
315  int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
316  not mix pictures before and after MMCO_RESET. */
317  int pic_id; /**< pic_num (short -> no wrap version of pic_num,
318  pic_num & max_pic_num; long -> long_pic_num) */
319  int long_ref; ///< 1->long term reference 0->short term reference
320  int ref_poc[2][2][32]; ///< POCs of the frames/fields used as reference (FIXME need per slice)
321  int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
322  int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
323  int field_picture; ///< whether or not picture was encoded in separate fields
324 
325  int needs_realloc; ///< picture needs to be reallocated (eg due to a frame size change)
326  int reference;
327  int recovered; ///< picture at IDR or recovery point + recovery count
328  int invalid_gap;
329  int sei_recovery_frame_cnt;
330 
331  int crop;
332  int crop_left;
333  int crop_top;
334 } H264Picture;
335 
336 /**
337  * H264Context
338  */
339 typedef struct H264Context {
340  AVClass *av_class;
341  AVCodecContext *avctx;
342  VideoDSPContext vdsp;
343  H264DSPContext h264dsp;
344  H264ChromaContext h264chroma;
345  H264QpelContext h264qpel;
346  GetBitContext gb;
347  ERContext er;
348 
349  H264Picture *DPB;
350  H264Picture *cur_pic_ptr;
351  H264Picture cur_pic;
352  H264Picture last_pic_for_ec;
353 
354  int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
355  int chroma_qp[2]; // QPc
356 
357  int qp_thresh; ///< QP threshold to skip loopfilter
358 
359  /* coded dimensions -- 16 * mb w/h */
360  int width, height;
361  ptrdiff_t linesize, uvlinesize;
362  int chroma_x_shift, chroma_y_shift;
363 
364  int qscale;
365  int droppable;
366  int coded_picture_number;
367  int low_delay;
368 
369  int context_initialized;
370  int flags;
371  int workaround_bugs;
372 
373  int prev_mb_skipped;
374  int next_mb_skipped;
375 
376  // prediction stuff
377  int chroma_pred_mode;
378  int intra16x16_pred_mode;
379 
380  int topleft_mb_xy;
381  int top_mb_xy;
382  int topright_mb_xy;
383  int left_mb_xy[LEFT_MBS];
384 
385  int topleft_type;
386  int top_type;
387  int topright_type;
388  int left_type[LEFT_MBS];
389 
390  const uint8_t *left_block;
391  int topleft_partition;
392 
393  int8_t intra4x4_pred_mode_cache[5 * 8];
394  int8_t(*intra4x4_pred_mode);
395  H264PredContext hpc;
396  unsigned int topleft_samples_available;
397  unsigned int top_samples_available;
398  unsigned int topright_samples_available;
399  unsigned int left_samples_available;
400  uint8_t (*top_borders[2])[(16 * 3) * 2];
401 
402  /**
403  * non zero coeff count cache.
404  * is 64 if not available.
405  */
406  DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
407 
408  uint8_t (*non_zero_count)[48];
409 
410  /**
411  * Motion vector cache.
412  */
413  DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
414  DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
415 #define LIST_NOT_USED -1 // FIXME rename?
416 #define PART_NOT_AVAILABLE -2
417 
418  /**
419  * number of neighbors (top and/or left) that used 8x8 dct
420  */
421  int neighbor_transform_size;
422 
423  /**
424  * block_offset[ 0..23] for frame macroblocks
425  * block_offset[24..47] for field macroblocks
426  */
427  int block_offset[2 * (16 * 3)];
428 
429  uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
430  uint32_t *mb2br_xy;
431  int b_stride; // FIXME use s->b4_stride
432 
433  ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
434  ptrdiff_t mb_uvlinesize;
435 
436  unsigned current_sps_id; ///< id of the current SPS
437  SPS sps; ///< current sps
438  PPS pps; ///< current pps
439 
440  int au_pps_id; ///< pps_id of current access unit
441 
442  uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
443  uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
444  uint32_t(*dequant4_coeff[6])[16];
445  uint32_t(*dequant8_coeff[6])[64];
446 
447  int slice_num;
448  uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
449  int slice_type;
450  int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
451  int slice_type_fixed;
452 
453  // interlacing specific flags
454  int mb_aff_frame;
455  int mb_field_decoding_flag;
456  int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
457  int picture_structure;
458  int first_field;
459 
460  DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
461 
462  // Weighted pred stuff
463  int use_weight;
464  int use_weight_chroma;
465  int luma_log2_weight_denom;
466  int chroma_log2_weight_denom;
467  // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
468  int luma_weight[48][2][2];
469  int chroma_weight[48][2][2][2];
470  int implicit_weight[48][48][2];
471 
472  int direct_spatial_mv_pred;
473  int col_parity;
474  int col_fieldoff;
475  int dist_scale_factor[32];
476  int dist_scale_factor_field[2][32];
477  int map_col_to_list0[2][16 + 32];
478  int map_col_to_list0_field[2][2][16 + 32];
479 
480  /**
481  * num_ref_idx_l0/1_active_minus1 + 1
482  */
483  unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
484  unsigned int list_count;
485  uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
486  H264Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
487  * Reordered version of default_ref_list
488  * according to picture reordering in slice header */
489  int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
490 
491  // data partitioning
492  GetBitContext intra_gb;
493  GetBitContext inter_gb;
494  GetBitContext *intra_gb_ptr;
495  GetBitContext *inter_gb_ptr;
496 
497  const uint8_t *intra_pcm_ptr;
498  DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2]; ///< as a dct coefficient is int32_t in high depth, we need to reserve twice the space.
499  DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
500  int16_t 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
501 
502  /**
503  * Cabac
504  */
505  CABACContext cabac;
506  uint8_t cabac_state[1024];
507 
508  /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
509  uint16_t *cbp_table;
510  int cbp;
511  int top_cbp;
512  int left_cbp;
513  /* chroma_pred_mode for i4x4 or i16x16, else 0 */
514  uint8_t *chroma_pred_mode_table;
515  int last_qscale_diff;
516  uint8_t (*mvd_table[2])[2];
517  DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
518  uint8_t *direct_table;
519  uint8_t direct_cache[5 * 8];
520 
521  uint8_t zigzag_scan[16];
522  uint8_t zigzag_scan8x8[64];
523  uint8_t zigzag_scan8x8_cavlc[64];
524  uint8_t field_scan[16];
525  uint8_t field_scan8x8[64];
526  uint8_t field_scan8x8_cavlc[64];
527  uint8_t zigzag_scan_q0[16];
528  uint8_t zigzag_scan8x8_q0[64];
529  uint8_t zigzag_scan8x8_cavlc_q0[64];
530  uint8_t field_scan_q0[16];
531  uint8_t field_scan8x8_q0[64];
532  uint8_t field_scan8x8_cavlc_q0[64];
533 
534  int x264_build;
535 
536  int mb_x, mb_y;
537  int resync_mb_x;
538  int resync_mb_y;
539  int mb_skip_run;
540  int mb_height, mb_width;
541  int mb_stride;
542  int mb_num;
543  int mb_xy;
544 
545  int is_complex;
546 
547  // deblock
548  int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
549  int slice_alpha_c0_offset;
550  int slice_beta_offset;
551 
552  // =============================================================
553  // Things below are not used in the MB or more inner code
554 
555  int nal_ref_idc;
556  int nal_unit_type;
557  uint8_t *rbsp_buffer[2];
558  unsigned int rbsp_buffer_size[2];
559 
560  /**
561  * Used to parse AVC variant of h264
562  */
563  int is_avc; ///< this flag is != 0 if codec is avc1
564  int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
565 
566  int bit_depth_luma; ///< luma bit depth from sps to detect changes
567  int chroma_format_idc; ///< chroma format from sps to detect changes
568 
569  SPS *sps_buffers[MAX_SPS_COUNT];
570  PPS *pps_buffers[MAX_PPS_COUNT];
571 
572  int dequant_coeff_pps; ///< reinit tables when pps changes
573 
574  uint16_t *slice_table_base;
575 
576  // POC stuff
577  int poc_lsb;
578  int poc_msb;
579  int delta_poc_bottom;
580  int delta_poc[2];
581  int frame_num;
582  int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
583  int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
584  int frame_num_offset; ///< for POC type 2
585  int prev_frame_num_offset; ///< for POC type 2
586  int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
587 
588  /**
589  * frame_num for frames or 2 * frame_num + 1 for field pics.
590  */
591  int curr_pic_num;
592 
593  /**
594  * max_frame_num or 2 * max_frame_num for field pics.
595  */
596  int max_pic_num;
597 
598  int redundant_pic_count;
599 
600  H264Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
601  H264Picture *short_ref[32];
602  H264Picture *long_ref[32];
603  H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
604  int last_pocs[MAX_DELAYED_PIC_COUNT];
605  H264Picture *next_output_pic;
606  int outputed_poc;
607  int next_outputed_poc;
608 
609  /**
610  * memory management control operations buffer.
611  */
612  MMCO mmco[MAX_MMCO_COUNT];
613  int mmco_index;
614  int mmco_reset;
615 
616  int long_ref_count; ///< number of actual long term references
617  int short_ref_count; ///< number of actual short term references
618 
619  int cabac_init_idc;
620 
621  /**
622  * @name Members for slice based multithreading
623  * @{
624  */
625  struct H264Context *thread_context[H264_MAX_THREADS];
626 
627  /**
628  * current slice number, used to initialize slice_num of each thread/context
629  */
630  int current_slice;
631 
632  /**
633  * Max number of threads / contexts.
634  * This is equal to AVCodecContext.thread_count unless
635  * multithreaded decoding is impossible, in which case it is
636  * reduced to 1.
637  */
638  int max_contexts;
639 
640  int slice_context_count;
641 
642  /**
643  * 1 if the single thread fallback warning has already been
644  * displayed, 0 otherwise.
645  */
646  int single_decode_warning;
647 
648  enum AVPictureType pict_type;
649 
650  int last_slice_type;
651  unsigned int last_ref_count[2];
652  /** @} */
653 
654  /**
655  * pic_struct in picture timing SEI message
656  */
657  SEI_PicStructType sei_pic_struct;
658 
659  /**
660  * Complement sei_pic_struct
661  * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
662  * However, soft telecined frames may have these values.
663  * This is used in an attempt to flag soft telecine progressive.
664  */
665  int prev_interlaced_frame;
666 
667  /**
668  * frame_packing_arrangment SEI message
669  */
670  int sei_frame_packing_present;
671  int frame_packing_arrangement_type;
672  int content_interpretation_type;
673  int quincunx_subsampling;
674 
675  /**
676  * display orientation SEI message
677  */
678  int sei_display_orientation_present;
679  int sei_anticlockwise_rotation;
680  int sei_hflip, sei_vflip;
681 
682  /**
683  * Bit set of clock types for fields/frames in picture timing SEI message.
684  * For each found ct_type, appropriate bit is set (e.g., bit 1 for
685  * interlaced).
686  */
687  int sei_ct_type;
688 
689  /**
690  * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
691  */
692  int sei_dpb_output_delay;
693 
694  /**
695  * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
696  */
697  int sei_cpb_removal_delay;
698 
699  /**
700  * recovery_frame_cnt from SEI message
701  *
702  * Set to -1 if no recovery point SEI message found or to number of frames
703  * before playback synchronizes. Frames having recovery point are key
704  * frames.
705  */
706  int sei_recovery_frame_cnt;
707 
708  /**
709  * Are the SEI recovery points looking valid.
710  */
711  int valid_recovery_point;
712 
713  FPA sei_fpa;
714 
715  /**
716  * recovery_frame is the frame_num at which the next frame should
717  * be fully constructed.
718  *
719  * Set to -1 when not expecting a recovery point.
720  */
721  int recovery_frame;
722 
723 /**
724  * We have seen an IDR, so all the following frames in coded order are correctly
725  * decodable.
726  */
727 #define FRAME_RECOVERED_IDR (1 << 0)
728 /**
729  * Sufficient number of frames have been decoded since a SEI recovery point,
730  * so all the following frames in presentation order are correct.
731  */
732 #define FRAME_RECOVERED_SEI (1 << 1)
733 
734  int frame_recovered; ///< Initial frame has been completely recovered
735 
736  int has_recovery_point;
737 
738  int missing_fields;
739 
740  int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
741  int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
742 
743  // Timestamp stuff
744  int sei_buffering_period_present; ///< Buffering period SEI flag
745  int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
746 
747  int cur_chroma_format_idc;
748  uint8_t *bipred_scratchpad;
749 
750  int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
751 
752  uint8_t parse_history[6];
753  int parse_history_count;
754  int parse_last_mb;
755  uint8_t *edge_emu_buffer;
756  int16_t *dc_val_base;
757 
758  AVBufferPool *qscale_table_pool;
759  AVBufferPool *mb_type_pool;
760  AVBufferPool *motion_val_pool;
761  AVBufferPool *ref_index_pool;
762 
763  /* Motion Estimation */
764  qpel_mc_func (*qpel_put)[16];
765  qpel_mc_func (*qpel_avg)[16];
766 } H264Context;
767 
768 extern const uint8_t ff_h264_chroma_qp[7][QP_MAX_NUM + 1]; ///< One chroma qp table for each possible bit depth (8-14).
769 extern const uint16_t ff_h264_mb_sizes[4];
770 
771 /**
772  * Decode SEI
773  */
774 int ff_h264_decode_sei(H264Context *h);
775 
776 /**
777  * Decode SPS
778  */
779 int ff_h264_decode_seq_parameter_set(H264Context *h);
780 
781 /**
782  * compute profile from sps
783  */
784 int ff_h264_get_profile(SPS *sps);
785 
786 /**
787  * Decode PPS
788  */
789 int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
790 
791 /**
792  * Decode a network abstraction layer unit.
793  * @param consumed is the number of bytes used as input
794  * @param length is the length of the array
795  * @param dst_length is the number of decoded bytes FIXME here
796  * or a decode rbsp tailing?
797  * @return decoded bytes, might be src+1 if no escapes
798  */
799 const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
800  int *dst_length, int *consumed, int length);
801 
802 /**
803  * Free any data that may have been allocated in the H264 context
804  * like SPS, PPS etc.
805  */
806 void ff_h264_free_context(H264Context *h);
807 
808 /**
809  * Reconstruct bitstream slice_type.
810  */
811 int ff_h264_get_slice_type(const H264Context *h);
812 
813 /**
814  * Allocate tables.
815  * needs width/height
816  */
817 int ff_h264_alloc_tables(H264Context *h);
818 
819 /**
820  * Fill the default_ref_list.
821  */
822 int ff_h264_fill_default_ref_list(H264Context *h);
823 
824 int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
825 void ff_h264_fill_mbaff_ref_list(H264Context *h);
826 void ff_h264_remove_all_refs(H264Context *h);
827 
828 /**
829  * Execute the reference picture marking (memory management control operations).
830  */
831 int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
832 
833 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
834  int first_slice);
835 
836 int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
837 
838 /**
839  * Check if the top & left blocks are available if needed & change the
840  * dc mode so it only uses the available blocks.
841  */
842 int ff_h264_check_intra4x4_pred_mode(H264Context *h);
843 
844 /**
845  * Check if the top & left blocks are available if needed & change the
846  * dc mode so it only uses the available blocks.
847  */
848 int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
849 
850 void ff_h264_hl_decode_mb(H264Context *h);
851 int ff_h264_decode_extradata(H264Context *h, const uint8_t *buf, int size);
852 int ff_h264_decode_init(AVCodecContext *avctx);
853 void ff_h264_decode_init_vlc(void);
854 
855 /**
856  * Decode a macroblock
857  * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
858  */
859 int ff_h264_decode_mb_cavlc(H264Context *h);
860 
861 /**
862  * Decode a CABAC coded macroblock
863  * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
864  */
865 int ff_h264_decode_mb_cabac(H264Context *h);
866 
867 void ff_h264_init_cabac_states(H264Context *h);
868 
869 void ff_h264_init_dequant_tables(H264Context *h);
870 
871 void ff_h264_direct_dist_scale_factor(H264Context *const h);
872 void ff_h264_direct_ref_list_init(H264Context *const h);
873 void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type);
874 
875 void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y,
876  uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
877  unsigned int linesize, unsigned int uvlinesize);
878 void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
879  uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
880  unsigned int linesize, unsigned int uvlinesize);
881 
882 /**
883  * Reset SEI values at the beginning of the frame.
884  *
885  * @param h H.264 context.
886  */
887 void ff_h264_reset_sei(H264Context *h);
888 
889 /**
890  * Get stereo_mode string from the h264 frame_packing_arrangement
891  * @param h H.264 context.
892  */
893 const char* ff_h264_sei_stereo_mode(H264Context *h);
894 
895 #define COPY_PICTURE(dst, src) \
896 do {\
897  *(dst) = *(src);\
898  (dst)->f.extended_data = (dst)->f.data;\
899  (dst)->tf.f = &(dst)->f;\
900 } while (0)
901 
902 /*
903  * o-o o-o
904  * / / /
905  * o-o o-o
906  * ,---'
907  * o-o o-o
908  * / / /
909  * o-o o-o
910  */
911 
912 /* Scan8 organization:
913  * 0 1 2 3 4 5 6 7
914  * 0 DY y y y y y
915  * 1 y Y Y Y Y
916  * 2 y Y Y Y Y
917  * 3 y Y Y Y Y
918  * 4 y Y Y Y Y
919  * 5 DU u u u u u
920  * 6 u U U U U
921  * 7 u U U U U
922  * 8 u U U U U
923  * 9 u U U U U
924  * 10 DV v v v v v
925  * 11 v V V V V
926  * 12 v V V V V
927  * 13 v V V V V
928  * 14 v V V V V
929  * DY/DU/DV are for luma/chroma DC.
930  */
931 
932 #define LUMA_DC_BLOCK_INDEX 48
933 #define CHROMA_DC_BLOCK_INDEX 49
934 
935 // This table must be here because scan8[constant] must be known at compiletime
936 static const uint8_t scan8[16 * 3 + 3] = {
937  4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
938  6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
939  4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
940  6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
941  4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
942  6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
943  4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
944  6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
945  4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
946  6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
947  4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
948  6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
949  0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
950 };
951 
952 static av_always_inline uint32_t pack16to32(int a, int b)
953 {
954 #if HAVE_BIGENDIAN
955  return (b & 0xFFFF) + (a << 16);
956 #else
957  return (a & 0xFFFF) + (b << 16);
958 #endif
959 }
960 
961 static av_always_inline uint16_t pack8to16(int a, int b)
962 {
963 #if HAVE_BIGENDIAN
964  return (b & 0xFF) + (a << 8);
965 #else
966  return (a & 0xFF) + (b << 8);
967 #endif
968 }
969 
970 /**
971  * Get the chroma qp.
972  */
973 static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
974 {
975  return h->pps.chroma_qp_table[t][qscale];
976 }
977 
978 /**
979  * Get the predicted intra4x4 prediction mode.
980  */
981 static av_always_inline int pred_intra_mode(H264Context *h, int n)
982 {
983  const int index8 = scan8[n];
984  const int left = h->intra4x4_pred_mode_cache[index8 - 1];
985  const int top = h->intra4x4_pred_mode_cache[index8 - 8];
986  const int min = FFMIN(left, top);
987 
988  tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min);
989 
990  if (min < 0)
991  return DC_PRED;
992  else
993  return min;
994 }
995 
996 static av_always_inline void write_back_intra_pred_mode(H264Context *h)
997 {
998  int8_t *i4x4 = h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
999  int8_t *i4x4_cache = h->intra4x4_pred_mode_cache;
1000 
1001  AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
1002  i4x4[4] = i4x4_cache[7 + 8 * 3];
1003  i4x4[5] = i4x4_cache[7 + 8 * 2];
1004  i4x4[6] = i4x4_cache[7 + 8 * 1];
1005 }
1006 
1007 static av_always_inline void write_back_non_zero_count(H264Context *h)
1008 {
1009  const int mb_xy = h->mb_xy;
1010  uint8_t *nnz = h->non_zero_count[mb_xy];
1011  uint8_t *nnz_cache = h->non_zero_count_cache;
1012 
1013  AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
1014  AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
1015  AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
1016  AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
1017  AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
1018  AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
1019  AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
1020  AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
1021 
1022  if (!h->chroma_y_shift) {
1023  AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
1024  AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
1025  AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
1026  AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
1027  }
1028 }
1029 
1030 static av_always_inline void write_back_motion_list(H264Context *h,
1031  int b_stride,
1032  int b_xy, int b8_xy,
1033  int mb_type, int list)
1034 {
1035  int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
1036  int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
1037  AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
1038  AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
1039  AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
1040  AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
1041  if (CABAC(h)) {
1042  uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8 * h->mb_xy
1043  : h->mb2br_xy[h->mb_xy]];
1044  uint8_t(*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
1045  if (IS_SKIP(mb_type)) {
1046  AV_ZERO128(mvd_dst);
1047  } else {
1048  AV_COPY64(mvd_dst, mvd_src + 8 * 3);
1049  AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
1050  AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
1051  AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
1052  }
1053  }
1054 
1055  {
1056  int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
1057  int8_t *ref_cache = h->ref_cache[list];
1058  ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
1059  ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
1060  ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
1061  ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
1062  }
1063 }
1064 
1065 static av_always_inline void write_back_motion(H264Context *h, int mb_type)
1066 {
1067  const int b_stride = h->b_stride;
1068  const int b_xy = 4 * h->mb_x + 4 * h->mb_y * h->b_stride; // try mb2b(8)_xy
1069  const int b8_xy = 4 * h->mb_xy;
1070 
1071  if (USES_LIST(mb_type, 0)) {
1072  write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 0);
1073  } else {
1074  fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
1075  2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
1076  }
1077  if (USES_LIST(mb_type, 1))
1078  write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 1);
1079 
1080  if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
1081  if (IS_8X8(mb_type)) {
1082  uint8_t *direct_table = &h->direct_table[4 * h->mb_xy];
1083  direct_table[1] = h->sub_mb_type[1] >> 1;
1084  direct_table[2] = h->sub_mb_type[2] >> 1;
1085  direct_table[3] = h->sub_mb_type[3] >> 1;
1086  }
1087  }
1088 }
1089 
1090 static av_always_inline int get_dct8x8_allowed(H264Context *h)
1091 {
1092  if (h->sps.direct_8x8_inference_flag)
1093  return !(AV_RN64A(h->sub_mb_type) &
1094  ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
1095  0x0001000100010001ULL));
1096  else
1097  return !(AV_RN64A(h->sub_mb_type) &
1098  ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
1099  0x0001000100010001ULL));
1100 }
1101 
1102 static inline int find_start_code(const uint8_t *buf, int buf_size,
1103  int buf_index, int next_avc)
1104 {
1105  uint32_t state = -1;
1106 
1107  buf_index = avpriv_find_start_code(buf + buf_index, buf + next_avc + 1, &state) - buf - 1;
1108 
1109  return FFMIN(buf_index, buf_size);
1110 }
1111 
1112 static inline int get_avc_nalsize(H264Context *h, const uint8_t *buf,
1113  int buf_size, int *buf_index)
1114 {
1115  int i, nalsize = 0;
1116 
1117  if (*buf_index >= buf_size - h->nal_length_size)
1118  return -1;
1119 
1120  for (i = 0; i < h->nal_length_size; i++)
1121  nalsize = ((unsigned)nalsize << 8) | buf[(*buf_index)++];
1122  if (nalsize <= 0 || nalsize > buf_size - *buf_index) {
1123  av_log(h->avctx, AV_LOG_ERROR,
1124  "AVC: nal size %d\n", nalsize);
1125  return -1;
1126  }
1127  return nalsize;
1128 }
1129 
1130 int ff_h264_field_end(H264Context *h, int in_setup);
1131 
1132 int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
1133 void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
1134 
1135 int ff_h264_context_init(H264Context *h);
1136 int ff_h264_set_parameter_from_sps(H264Context *h);
1137 
1138 void ff_h264_draw_horiz_band(H264Context *h, int y, int height);
1139 int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
1140 int ff_pred_weight_table(H264Context *h);
1141 int ff_set_ref_count(H264Context *h);
1142 
1143 int ff_h264_decode_slice_header(H264Context *h, H264Context *h0);
1144 #define SLICE_SINGLETHREAD 1
1145 #define SLICE_SKIPED 2
1146 
1147 int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count);
1148 int ff_h264_update_thread_context(AVCodecContext *dst,
1149  const AVCodecContext *src);
1150 
1151 void ff_h264_flush_change(H264Context *h);
1152 
1153 void ff_h264_free_tables(H264Context *h, int free_rbsp);
1154 
1155 void ff_h264_set_erpic(ERPicture *dst, H264Picture *src);
1156 
1157 #endif /* AVCODEC_H264_H */
Generated on Sun Mar 8 2015 02:34:52 for FFmpeg by   doxygen 1.8.2