Imported Debian version 2.5.3~trusty1
[deb_ffmpeg.git] / ffmpeg / libavcodec / h264.h
CommitLineData
2ba45a60
DM
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 "me_cmp.h"
41#include "mpegutils.h"
42#include "parser.h"
43#include "qpeldsp.h"
44#include "rectangle.h"
45#include "videodsp.h"
46
47#define H264_MAX_PICTURE_COUNT 36
48#define H264_MAX_THREADS 32
49
50#define MAX_SPS_COUNT 32
51#define MAX_PPS_COUNT 256
52
53#define MAX_MMCO_COUNT 66
54
55#define MAX_DELAYED_PIC_COUNT 16
56
57#define MAX_MBPAIR_SIZE (256*1024) // a tighter bound could be calculated if someone cares about a few bytes
58
59/* Compiling in interlaced support reduces the speed
60 * of progressive decoding by about 2%. */
61#define ALLOW_INTERLACE
62
63#define FMO 0
64
65/**
66 * The maximum number of slices supported by the decoder.
67 * must be a power of 2
68 */
69#define MAX_SLICES 16
70
71#ifdef ALLOW_INTERLACE
72#define MB_MBAFF(h) (h)->mb_mbaff
73#define MB_FIELD(h) (h)->mb_field_decoding_flag
74#define FRAME_MBAFF(h) (h)->mb_aff_frame
75#define FIELD_PICTURE(h) ((h)->picture_structure != PICT_FRAME)
76#define LEFT_MBS 2
77#define LTOP 0
78#define LBOT 1
79#define LEFT(i) (i)
80#else
81#define MB_MBAFF(h) 0
82#define MB_FIELD(h) 0
83#define FRAME_MBAFF(h) 0
84#define FIELD_PICTURE(h) 0
85#undef IS_INTERLACED
86#define IS_INTERLACED(mb_type) 0
87#define LEFT_MBS 1
88#define LTOP 0
89#define LBOT 0
90#define LEFT(i) 0
91#endif
92#define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
93
94#ifndef CABAC
95#define CABAC(h) (h)->pps.cabac
96#endif
97
98#define CHROMA(h) ((h)->sps.chroma_format_idc)
99#define CHROMA422(h) ((h)->sps.chroma_format_idc == 2)
100#define CHROMA444(h) ((h)->sps.chroma_format_idc == 3)
101
102#define EXTENDED_SAR 255
103
104#define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
105#define MB_TYPE_8x8DCT 0x01000000
106#define IS_REF0(a) ((a) & MB_TYPE_REF0)
107#define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
108
109#define QP_MAX_NUM (51 + 6*6) // The maximum supported qp
110
111/* NAL unit types */
112enum {
113 NAL_SLICE = 1,
114 NAL_DPA = 2,
115 NAL_DPB = 3,
116 NAL_DPC = 4,
117 NAL_IDR_SLICE = 5,
118 NAL_SEI = 6,
119 NAL_SPS = 7,
120 NAL_PPS = 8,
121 NAL_AUD = 9,
122 NAL_END_SEQUENCE = 10,
123 NAL_END_STREAM = 11,
124 NAL_FILLER_DATA = 12,
125 NAL_SPS_EXT = 13,
126 NAL_AUXILIARY_SLICE = 19,
127 NAL_FF_IGNORE = 0xff0f001,
128};
129
130/**
131 * SEI message types
132 */
133typedef enum {
134 SEI_TYPE_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
135 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
136 SEI_TYPE_USER_DATA_ITU_T_T35 = 4, ///< user data registered by ITU-T Recommendation T.35
137 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
138 SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
139 SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
140 SEI_TYPE_DISPLAY_ORIENTATION = 47, ///< display orientation
141} SEI_Type;
142
143/**
144 * pic_struct in picture timing SEI message
145 */
146typedef enum {
147 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
148 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
149 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
150 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
151 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
152 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
153 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
154 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
155 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
156} SEI_PicStructType;
157
158/**
159 * frame_packing_arrangement types
160 */
161typedef enum {
162 SEI_FPA_TYPE_CHECKERBOARD = 0,
163 SEI_FPA_TYPE_INTERLEAVE_COLUMN = 1,
164 SEI_FPA_TYPE_INTERLEAVE_ROW = 2,
165 SEI_FPA_TYPE_SIDE_BY_SIDE = 3,
166 SEI_FPA_TYPE_TOP_BOTTOM = 4,
167 SEI_FPA_TYPE_INTERLEAVE_TEMPORAL = 5,
168 SEI_FPA_TYPE_2D = 6,
169} SEI_FpaType;
170
171/**
172 * Sequence parameter set
173 */
174typedef struct SPS {
175 unsigned int sps_id;
176 int profile_idc;
177 int level_idc;
178 int chroma_format_idc;
179 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
180 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
181 int poc_type; ///< pic_order_cnt_type
182 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
183 int delta_pic_order_always_zero_flag;
184 int offset_for_non_ref_pic;
185 int offset_for_top_to_bottom_field;
186 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
187 int ref_frame_count; ///< num_ref_frames
188 int gaps_in_frame_num_allowed_flag;
189 int mb_width; ///< pic_width_in_mbs_minus1 + 1
190 int mb_height; ///< pic_height_in_map_units_minus1 + 1
191 int frame_mbs_only_flag;
192 int mb_aff; ///< mb_adaptive_frame_field_flag
193 int direct_8x8_inference_flag;
194 int crop; ///< frame_cropping_flag
195
196 /* those 4 are already in luma samples */
197 unsigned int crop_left; ///< frame_cropping_rect_left_offset
198 unsigned int crop_right; ///< frame_cropping_rect_right_offset
199 unsigned int crop_top; ///< frame_cropping_rect_top_offset
200 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
201 int vui_parameters_present_flag;
202 AVRational sar;
203 int video_signal_type_present_flag;
204 int full_range;
205 int colour_description_present_flag;
206 enum AVColorPrimaries color_primaries;
207 enum AVColorTransferCharacteristic color_trc;
208 enum AVColorSpace colorspace;
209 int timing_info_present_flag;
210 uint32_t num_units_in_tick;
211 uint32_t time_scale;
212 int fixed_frame_rate_flag;
213 short offset_for_ref_frame[256]; // FIXME dyn aloc?
214 int bitstream_restriction_flag;
215 int num_reorder_frames;
216 int scaling_matrix_present;
217 uint8_t scaling_matrix4[6][16];
218 uint8_t scaling_matrix8[6][64];
219 int nal_hrd_parameters_present_flag;
220 int vcl_hrd_parameters_present_flag;
221 int pic_struct_present_flag;
222 int time_offset_length;
223 int cpb_cnt; ///< See H.264 E.1.2
224 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
225 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
226 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
227 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
228 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
229 int residual_color_transform_flag; ///< residual_colour_transform_flag
230 int constraint_set_flags; ///< constraint_set[0-3]_flag
231 int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
232} SPS;
233
234/**
235 * Picture parameter set
236 */
237typedef struct PPS {
238 unsigned int sps_id;
239 int cabac; ///< entropy_coding_mode_flag
240 int pic_order_present; ///< pic_order_present_flag
241 int slice_group_count; ///< num_slice_groups_minus1 + 1
242 int mb_slice_group_map_type;
243 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
244 int weighted_pred; ///< weighted_pred_flag
245 int weighted_bipred_idc;
246 int init_qp; ///< pic_init_qp_minus26 + 26
247 int init_qs; ///< pic_init_qs_minus26 + 26
248 int chroma_qp_index_offset[2];
249 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
250 int constrained_intra_pred; ///< constrained_intra_pred_flag
251 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
252 int transform_8x8_mode; ///< transform_8x8_mode_flag
253 uint8_t scaling_matrix4[6][16];
254 uint8_t scaling_matrix8[6][64];
255 uint8_t chroma_qp_table[2][QP_MAX_NUM+1]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
256 int chroma_qp_diff;
257} PPS;
258
259/**
260 * Frame Packing Arrangement Type
261 */
262typedef struct FPA {
263 int frame_packing_arrangement_id;
264 int frame_packing_arrangement_cancel_flag; ///< is previous arrangement canceled, -1 if never received
265 SEI_FpaType frame_packing_arrangement_type;
266 int frame_packing_arrangement_repetition_period;
267 int content_interpretation_type;
268 int quincunx_sampling_flag;
269} FPA;
270
271/**
272 * Memory management control operation opcode.
273 */
274typedef enum MMCOOpcode {
275 MMCO_END = 0,
276 MMCO_SHORT2UNUSED,
277 MMCO_LONG2UNUSED,
278 MMCO_SHORT2LONG,
279 MMCO_SET_MAX_LONG,
280 MMCO_RESET,
281 MMCO_LONG,
282} MMCOOpcode;
283
284/**
285 * Memory management control operation.
286 */
287typedef struct MMCO {
288 MMCOOpcode opcode;
289 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
290 int long_arg; ///< index, pic_num, or num long refs depending on opcode
291} MMCO;
292
293typedef struct H264Picture {
294 struct AVFrame f;
295 uint8_t avframe_padding[1024]; // hack to allow linking to a avutil with larger AVFrame
296 ThreadFrame tf;
297
298 AVBufferRef *qscale_table_buf;
299 int8_t *qscale_table;
300
301 AVBufferRef *motion_val_buf[2];
302 int16_t (*motion_val[2])[2];
303
304 AVBufferRef *mb_type_buf;
305 uint32_t *mb_type;
306
307 AVBufferRef *hwaccel_priv_buf;
308 void *hwaccel_picture_private; ///< hardware accelerator private data
309
310 AVBufferRef *ref_index_buf[2];
311 int8_t *ref_index[2];
312
313 int field_poc[2]; ///< top/bottom POC
314 int poc; ///< frame POC
315 int frame_num; ///< frame_num (raw frame_num from slice header)
316 int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
317 not mix pictures before and after MMCO_RESET. */
318 int pic_id; /**< pic_num (short -> no wrap version of pic_num,
319 pic_num & max_pic_num; long -> long_pic_num) */
320 int long_ref; ///< 1->long term reference 0->short term reference
321 int ref_poc[2][2][32]; ///< POCs of the frames/fields used as reference (FIXME need per slice)
322 int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
323 int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
324 int field_picture; ///< whether or not picture was encoded in separate fields
325
326 int needs_realloc; ///< picture needs to be reallocated (eg due to a frame size change)
327 int reference;
328 int recovered; ///< picture at IDR or recovery point + recovery count
329 int invalid_gap;
330 int sei_recovery_frame_cnt;
331
332 int crop;
333 int crop_left;
334 int crop_top;
335} H264Picture;
336
337/**
338 * H264Context
339 */
340typedef struct H264Context {
092a9121 341 AVClass *av_class;
2ba45a60
DM
342 AVCodecContext *avctx;
343 MECmpContext mecc;
344 VideoDSPContext vdsp;
345 H264DSPContext h264dsp;
346 H264ChromaContext h264chroma;
347 H264QpelContext h264qpel;
348 ParseContext parse_context;
349 GetBitContext gb;
350 ERContext er;
351
352 H264Picture *DPB;
353 H264Picture *cur_pic_ptr;
354 H264Picture cur_pic;
355
356 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
357 int chroma_qp[2]; // QPc
358
359 int qp_thresh; ///< QP threshold to skip loopfilter
360
361 /* coded dimensions -- 16 * mb w/h */
362 int width, height;
363 ptrdiff_t linesize, uvlinesize;
364 int chroma_x_shift, chroma_y_shift;
365
366 int qscale;
367 int droppable;
368 int data_partitioning;
369 int coded_picture_number;
370 int low_delay;
371
372 int context_initialized;
373 int flags;
374 int workaround_bugs;
375
376 int prev_mb_skipped;
377 int next_mb_skipped;
378
379 // prediction stuff
380 int chroma_pred_mode;
381 int intra16x16_pred_mode;
382
383 int topleft_mb_xy;
384 int top_mb_xy;
385 int topright_mb_xy;
386 int left_mb_xy[LEFT_MBS];
387
388 int topleft_type;
389 int top_type;
390 int topright_type;
391 int left_type[LEFT_MBS];
392
393 const uint8_t *left_block;
394 int topleft_partition;
395
396 int8_t intra4x4_pred_mode_cache[5 * 8];
397 int8_t(*intra4x4_pred_mode);
398 H264PredContext hpc;
399 unsigned int topleft_samples_available;
400 unsigned int top_samples_available;
401 unsigned int topright_samples_available;
402 unsigned int left_samples_available;
403 uint8_t (*top_borders[2])[(16 * 3) * 2];
404
405 /**
406 * non zero coeff count cache.
407 * is 64 if not available.
408 */
409 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
410
411 uint8_t (*non_zero_count)[48];
412
413 /**
414 * Motion vector cache.
415 */
416 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
417 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
418#define LIST_NOT_USED -1 // FIXME rename?
419#define PART_NOT_AVAILABLE -2
420
421 /**
422 * number of neighbors (top and/or left) that used 8x8 dct
423 */
424 int neighbor_transform_size;
425
426 /**
427 * block_offset[ 0..23] for frame macroblocks
428 * block_offset[24..47] for field macroblocks
429 */
430 int block_offset[2 * (16 * 3)];
431
432 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
433 uint32_t *mb2br_xy;
434 int b_stride; // FIXME use s->b4_stride
435
436 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
437 ptrdiff_t mb_uvlinesize;
438
439 unsigned current_sps_id; ///< id of the current SPS
440 SPS sps; ///< current sps
441 PPS pps; ///< current pps
442
443 int au_pps_id; ///< pps_id of current access unit
444
445 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
446 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
447 uint32_t(*dequant4_coeff[6])[16];
448 uint32_t(*dequant8_coeff[6])[64];
449
450 int slice_num;
451 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
452 int slice_type;
453 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
454 int slice_type_fixed;
455
456 // interlacing specific flags
457 int mb_aff_frame;
458 int mb_field_decoding_flag;
459 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
460 int picture_structure;
461 int first_field;
462
463 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
464
465 // Weighted pred stuff
466 int use_weight;
467 int use_weight_chroma;
468 int luma_log2_weight_denom;
469 int chroma_log2_weight_denom;
470 // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
471 int luma_weight[48][2][2];
472 int chroma_weight[48][2][2][2];
473 int implicit_weight[48][48][2];
474
475 int direct_spatial_mv_pred;
476 int col_parity;
477 int col_fieldoff;
478 int dist_scale_factor[32];
479 int dist_scale_factor_field[2][32];
480 int map_col_to_list0[2][16 + 32];
481 int map_col_to_list0_field[2][2][16 + 32];
482
483 /**
484 * num_ref_idx_l0/1_active_minus1 + 1
485 */
486 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
487 unsigned int list_count;
488 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
489 H264Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
490 * Reordered version of default_ref_list
491 * according to picture reordering in slice header */
492 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
493
494 // data partitioning
495 GetBitContext intra_gb;
496 GetBitContext inter_gb;
497 GetBitContext *intra_gb_ptr;
498 GetBitContext *inter_gb_ptr;
499
500 const uint8_t *intra_pcm_ptr;
501 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.
502 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
503 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
504
505 /**
506 * Cabac
507 */
508 CABACContext cabac;
509 uint8_t cabac_state[1024];
510
511 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
512 uint16_t *cbp_table;
513 int cbp;
514 int top_cbp;
515 int left_cbp;
516 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
517 uint8_t *chroma_pred_mode_table;
518 int last_qscale_diff;
519 uint8_t (*mvd_table[2])[2];
520 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
521 uint8_t *direct_table;
522 uint8_t direct_cache[5 * 8];
523
524 uint8_t zigzag_scan[16];
525 uint8_t zigzag_scan8x8[64];
526 uint8_t zigzag_scan8x8_cavlc[64];
527 uint8_t field_scan[16];
528 uint8_t field_scan8x8[64];
529 uint8_t field_scan8x8_cavlc[64];
530 uint8_t zigzag_scan_q0[16];
531 uint8_t zigzag_scan8x8_q0[64];
532 uint8_t zigzag_scan8x8_cavlc_q0[64];
533 uint8_t field_scan_q0[16];
534 uint8_t field_scan8x8_q0[64];
535 uint8_t field_scan8x8_cavlc_q0[64];
536
537 int x264_build;
538
539 int mb_x, mb_y;
540 int resync_mb_x;
541 int resync_mb_y;
542 int mb_skip_run;
543 int mb_height, mb_width;
544 int mb_stride;
545 int mb_num;
546 int mb_xy;
547
548 int is_complex;
549
550 // deblock
551 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
552 int slice_alpha_c0_offset;
553 int slice_beta_offset;
554
555 // =============================================================
556 // Things below are not used in the MB or more inner code
557
558 int nal_ref_idc;
559 int nal_unit_type;
560 uint8_t *rbsp_buffer[2];
561 unsigned int rbsp_buffer_size[2];
562
563 /**
564 * Used to parse AVC variant of h264
565 */
566 int is_avc; ///< this flag is != 0 if codec is avc1
567 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
568 int got_first; ///< this flag is != 0 if we've parsed a frame
569
570 int bit_depth_luma; ///< luma bit depth from sps to detect changes
571 int chroma_format_idc; ///< chroma format from sps to detect changes
572
573 SPS *sps_buffers[MAX_SPS_COUNT];
574 PPS *pps_buffers[MAX_PPS_COUNT];
575
576 int dequant_coeff_pps; ///< reinit tables when pps changes
577
578 uint16_t *slice_table_base;
579
580 // POC stuff
581 int poc_lsb;
582 int poc_msb;
583 int delta_poc_bottom;
584 int delta_poc[2];
585 int frame_num;
586 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
587 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
588 int frame_num_offset; ///< for POC type 2
589 int prev_frame_num_offset; ///< for POC type 2
590 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
591
592 /**
593 * frame_num for frames or 2 * frame_num + 1 for field pics.
594 */
595 int curr_pic_num;
596
597 /**
598 * max_frame_num or 2 * max_frame_num for field pics.
599 */
600 int max_pic_num;
601
602 int redundant_pic_count;
603
604 H264Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
605 H264Picture *short_ref[32];
606 H264Picture *long_ref[32];
607 H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
608 int last_pocs[MAX_DELAYED_PIC_COUNT];
609 H264Picture *next_output_pic;
610 int outputed_poc;
611 int next_outputed_poc;
612
613 /**
614 * memory management control operations buffer.
615 */
616 MMCO mmco[MAX_MMCO_COUNT];
617 int mmco_index;
618 int mmco_reset;
619
620 int long_ref_count; ///< number of actual long term references
621 int short_ref_count; ///< number of actual short term references
622
623 int cabac_init_idc;
624
625 /**
626 * @name Members for slice based multithreading
627 * @{
628 */
629 struct H264Context *thread_context[H264_MAX_THREADS];
630
631 /**
632 * current slice number, used to initialize slice_num of each thread/context
633 */
634 int current_slice;
635
636 /**
637 * Max number of threads / contexts.
638 * This is equal to AVCodecContext.thread_count unless
639 * multithreaded decoding is impossible, in which case it is
640 * reduced to 1.
641 */
642 int max_contexts;
643
644 int slice_context_count;
645
646 /**
647 * 1 if the single thread fallback warning has already been
648 * displayed, 0 otherwise.
649 */
650 int single_decode_warning;
651
652 enum AVPictureType pict_type;
653
654 int last_slice_type;
655 unsigned int last_ref_count[2];
656 /** @} */
657
658 /**
659 * pic_struct in picture timing SEI message
660 */
661 SEI_PicStructType sei_pic_struct;
662
663 /**
664 * Complement sei_pic_struct
665 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
666 * However, soft telecined frames may have these values.
667 * This is used in an attempt to flag soft telecine progressive.
668 */
669 int prev_interlaced_frame;
670
671 /**
672 * frame_packing_arrangment SEI message
673 */
674 int sei_frame_packing_present;
675 int frame_packing_arrangement_type;
676 int content_interpretation_type;
677 int quincunx_subsampling;
678
679 /**
680 * display orientation SEI message
681 */
682 int sei_display_orientation_present;
683 int sei_anticlockwise_rotation;
684 int sei_hflip, sei_vflip;
685
686 /**
687 * Bit set of clock types for fields/frames in picture timing SEI message.
688 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
689 * interlaced).
690 */
691 int sei_ct_type;
692
693 /**
694 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
695 */
696 int sei_dpb_output_delay;
697
698 /**
699 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
700 */
701 int sei_cpb_removal_delay;
702
703 /**
704 * recovery_frame_cnt from SEI message
705 *
706 * Set to -1 if no recovery point SEI message found or to number of frames
707 * before playback synchronizes. Frames having recovery point are key
708 * frames.
709 */
710 int sei_recovery_frame_cnt;
711
712 /**
713 * Are the SEI recovery points looking valid.
714 */
715 int valid_recovery_point;
716
717 FPA sei_fpa;
718
719 /**
720 * recovery_frame is the frame_num at which the next frame should
721 * be fully constructed.
722 *
723 * Set to -1 when not expecting a recovery point.
724 */
725 int recovery_frame;
726
727/**
728 * We have seen an IDR, so all the following frames in coded order are correctly
729 * decodable.
730 */
731#define FRAME_RECOVERED_IDR (1 << 0)
732/**
733 * Sufficient number of frames have been decoded since a SEI recovery point,
734 * so all the following frames in presentation order are correct.
735 */
736#define FRAME_RECOVERED_SEI (1 << 1)
737
738 int frame_recovered; ///< Initial frame has been completely recovered
739
740 int has_recovery_point;
741
742 int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
743 int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
744
745 // Timestamp stuff
746 int sei_buffering_period_present; ///< Buffering period SEI flag
747 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
748
749 int cur_chroma_format_idc;
750 uint8_t *bipred_scratchpad;
751
752 int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
753
754 uint8_t parse_history[6];
755 int parse_history_count;
756 int parse_last_mb;
757 uint8_t *edge_emu_buffer;
758 int16_t *dc_val_base;
759
760 AVBufferPool *qscale_table_pool;
761 AVBufferPool *mb_type_pool;
762 AVBufferPool *motion_val_pool;
763 AVBufferPool *ref_index_pool;
764
765 /* Motion Estimation */
766 qpel_mc_func (*qpel_put)[16];
767 qpel_mc_func (*qpel_avg)[16];
768} H264Context;
769
770extern const uint8_t ff_h264_chroma_qp[7][QP_MAX_NUM + 1]; ///< One chroma qp table for each possible bit depth (8-14).
771extern const uint16_t ff_h264_mb_sizes[4];
772
773/**
774 * Decode SEI
775 */
776int ff_h264_decode_sei(H264Context *h);
777
778/**
779 * Decode SPS
780 */
781int ff_h264_decode_seq_parameter_set(H264Context *h);
782
783/**
784 * compute profile from sps
785 */
786int ff_h264_get_profile(SPS *sps);
787
788/**
789 * Decode PPS
790 */
791int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
792
793/**
794 * Decode a network abstraction layer unit.
795 * @param consumed is the number of bytes used as input
796 * @param length is the length of the array
797 * @param dst_length is the number of decoded bytes FIXME here
798 * or a decode rbsp tailing?
799 * @return decoded bytes, might be src+1 if no escapes
800 */
801const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
802 int *dst_length, int *consumed, int length);
803
804/**
805 * Free any data that may have been allocated in the H264 context
806 * like SPS, PPS etc.
807 */
808void ff_h264_free_context(H264Context *h);
809
810/**
811 * Reconstruct bitstream slice_type.
812 */
813int ff_h264_get_slice_type(const H264Context *h);
814
815/**
816 * Allocate tables.
817 * needs width/height
818 */
819int ff_h264_alloc_tables(H264Context *h);
820
821/**
822 * Fill the default_ref_list.
823 */
824int ff_h264_fill_default_ref_list(H264Context *h);
825
826int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
827void ff_h264_fill_mbaff_ref_list(H264Context *h);
828void ff_h264_remove_all_refs(H264Context *h);
829
830/**
831 * Execute the reference picture marking (memory management control operations).
832 */
833int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
834
835int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
836 int first_slice);
837
838int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
839
840/**
841 * Check if the top & left blocks are available if needed & change the
842 * dc mode so it only uses the available blocks.
843 */
844int ff_h264_check_intra4x4_pred_mode(H264Context *h);
845
846/**
847 * Check if the top & left blocks are available if needed & change the
848 * dc mode so it only uses the available blocks.
849 */
850int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
851
852void ff_h264_hl_decode_mb(H264Context *h);
853int ff_h264_decode_extradata(H264Context *h, const uint8_t *buf, int size);
854int ff_h264_decode_init(AVCodecContext *avctx);
855void ff_h264_decode_init_vlc(void);
856
857/**
858 * Decode a macroblock
859 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
860 */
861int ff_h264_decode_mb_cavlc(H264Context *h);
862
863/**
864 * Decode a CABAC coded macroblock
865 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
866 */
867int ff_h264_decode_mb_cabac(H264Context *h);
868
869void ff_h264_init_cabac_states(H264Context *h);
870
871void h264_init_dequant_tables(H264Context *h);
872
873void ff_h264_direct_dist_scale_factor(H264Context *const h);
874void ff_h264_direct_ref_list_init(H264Context *const h);
875void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type);
876
877void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y,
878 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
879 unsigned int linesize, unsigned int uvlinesize);
880void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
881 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
882 unsigned int linesize, unsigned int uvlinesize);
883
884/**
885 * Reset SEI values at the beginning of the frame.
886 *
887 * @param h H.264 context.
888 */
889void ff_h264_reset_sei(H264Context *h);
890
891/**
892 * Get stereo_mode string from the h264 frame_packing_arrangement
893 * @param h H.264 context.
894 */
895const char* ff_h264_sei_stereo_mode(H264Context *h);
896
897/*
898 * o-o o-o
899 * / / /
900 * o-o o-o
901 * ,---'
902 * o-o o-o
903 * / / /
904 * o-o o-o
905 */
906
907/* Scan8 organization:
908 * 0 1 2 3 4 5 6 7
909 * 0 DY y y y y y
910 * 1 y Y Y Y Y
911 * 2 y Y Y Y Y
912 * 3 y Y Y Y Y
913 * 4 y Y Y Y Y
914 * 5 DU u u u u u
915 * 6 u U U U U
916 * 7 u U U U U
917 * 8 u U U U U
918 * 9 u U U U U
919 * 10 DV v v v v v
920 * 11 v V V V V
921 * 12 v V V V V
922 * 13 v V V V V
923 * 14 v V V V V
924 * DY/DU/DV are for luma/chroma DC.
925 */
926
927#define LUMA_DC_BLOCK_INDEX 48
928#define CHROMA_DC_BLOCK_INDEX 49
929
930// This table must be here because scan8[constant] must be known at compiletime
931static const uint8_t scan8[16 * 3 + 3] = {
932 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
933 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
934 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
935 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
936 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
937 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
938 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
939 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
940 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
941 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
942 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
943 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
944 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
945};
946
947static av_always_inline uint32_t pack16to32(int a, int b)
948{
949#if HAVE_BIGENDIAN
950 return (b & 0xFFFF) + (a << 16);
951#else
952 return (a & 0xFFFF) + (b << 16);
953#endif
954}
955
956static av_always_inline uint16_t pack8to16(int a, int b)
957{
958#if HAVE_BIGENDIAN
959 return (b & 0xFF) + (a << 8);
960#else
961 return (a & 0xFF) + (b << 8);
962#endif
963}
964
965/**
966 * Get the chroma qp.
967 */
968static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
969{
970 return h->pps.chroma_qp_table[t][qscale];
971}
972
973/**
974 * Get the predicted intra4x4 prediction mode.
975 */
976static av_always_inline int pred_intra_mode(H264Context *h, int n)
977{
978 const int index8 = scan8[n];
979 const int left = h->intra4x4_pred_mode_cache[index8 - 1];
980 const int top = h->intra4x4_pred_mode_cache[index8 - 8];
981 const int min = FFMIN(left, top);
982
983 tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min);
984
985 if (min < 0)
986 return DC_PRED;
987 else
988 return min;
989}
990
991static av_always_inline void write_back_intra_pred_mode(H264Context *h)
992{
993 int8_t *i4x4 = h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
994 int8_t *i4x4_cache = h->intra4x4_pred_mode_cache;
995
996 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
997 i4x4[4] = i4x4_cache[7 + 8 * 3];
998 i4x4[5] = i4x4_cache[7 + 8 * 2];
999 i4x4[6] = i4x4_cache[7 + 8 * 1];
1000}
1001
1002static av_always_inline void write_back_non_zero_count(H264Context *h)
1003{
1004 const int mb_xy = h->mb_xy;
1005 uint8_t *nnz = h->non_zero_count[mb_xy];
1006 uint8_t *nnz_cache = h->non_zero_count_cache;
1007
1008 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
1009 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
1010 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
1011 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
1012 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
1013 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
1014 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
1015 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
1016
1017 if (!h->chroma_y_shift) {
1018 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
1019 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
1020 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
1021 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
1022 }
1023}
1024
1025static av_always_inline void write_back_motion_list(H264Context *h,
1026 int b_stride,
1027 int b_xy, int b8_xy,
1028 int mb_type, int list)
1029{
1030 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
1031 int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
1032 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
1033 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
1034 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
1035 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
1036 if (CABAC(h)) {
1037 uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8 * h->mb_xy
1038 : h->mb2br_xy[h->mb_xy]];
1039 uint8_t(*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
1040 if (IS_SKIP(mb_type)) {
1041 AV_ZERO128(mvd_dst);
1042 } else {
1043 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
1044 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
1045 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
1046 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
1047 }
1048 }
1049
1050 {
1051 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
1052 int8_t *ref_cache = h->ref_cache[list];
1053 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
1054 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
1055 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
1056 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
1057 }
1058}
1059
1060static av_always_inline void write_back_motion(H264Context *h, int mb_type)
1061{
1062 const int b_stride = h->b_stride;
1063 const int b_xy = 4 * h->mb_x + 4 * h->mb_y * h->b_stride; // try mb2b(8)_xy
1064 const int b8_xy = 4 * h->mb_xy;
1065
1066 if (USES_LIST(mb_type, 0)) {
1067 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 0);
1068 } else {
1069 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
1070 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
1071 }
1072 if (USES_LIST(mb_type, 1))
1073 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 1);
1074
1075 if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
1076 if (IS_8X8(mb_type)) {
1077 uint8_t *direct_table = &h->direct_table[4 * h->mb_xy];
1078 direct_table[1] = h->sub_mb_type[1] >> 1;
1079 direct_table[2] = h->sub_mb_type[2] >> 1;
1080 direct_table[3] = h->sub_mb_type[3] >> 1;
1081 }
1082 }
1083}
1084
1085static av_always_inline int get_dct8x8_allowed(H264Context *h)
1086{
1087 if (h->sps.direct_8x8_inference_flag)
1088 return !(AV_RN64A(h->sub_mb_type) &
1089 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
1090 0x0001000100010001ULL));
1091 else
1092 return !(AV_RN64A(h->sub_mb_type) &
1093 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
1094 0x0001000100010001ULL));
1095}
1096
1097static inline int find_start_code(const uint8_t *buf, int buf_size,
1098 int buf_index, int next_avc)
1099{
1100 uint32_t state = -1;
1101
1102 buf_index = avpriv_find_start_code(buf + buf_index, buf + next_avc + 1, &state) - buf - 1;
1103
1104 return FFMIN(buf_index, buf_size);
1105}
1106
1107static inline int get_avc_nalsize(H264Context *h, const uint8_t *buf,
1108 int buf_size, int *buf_index)
1109{
1110 int i, nalsize = 0;
1111
1112 if (*buf_index >= buf_size - h->nal_length_size)
1113 return -1;
1114
1115 for (i = 0; i < h->nal_length_size; i++)
1116 nalsize = ((unsigned)nalsize << 8) | buf[(*buf_index)++];
1117 if (nalsize <= 0 || nalsize > buf_size - *buf_index) {
1118 av_log(h->avctx, AV_LOG_ERROR,
1119 "AVC: nal size %d\n", nalsize);
1120 return -1;
1121 }
1122 return nalsize;
1123}
1124
1125int ff_h264_field_end(H264Context *h, int in_setup);
1126
1127int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
1128void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
1129
1130int ff_h264_context_init(H264Context *h);
1131int ff_h264_set_parameter_from_sps(H264Context *h);
1132
1133void ff_h264_draw_horiz_band(H264Context *h, int y, int height);
1134int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
1135int ff_pred_weight_table(H264Context *h);
1136int ff_set_ref_count(H264Context *h);
1137
1138int ff_h264_decode_slice_header(H264Context *h, H264Context *h0);
1139#define SLICE_SINGLETHREAD 1
1140#define SLICE_SKIPED 2
1141
1142int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count);
1143int ff_h264_update_thread_context(AVCodecContext *dst,
1144 const AVCodecContext *src);
1145
1146void ff_h264_flush_change(H264Context *h);
1147
1148void ff_h264_free_tables(H264Context *h, int free_rbsp);
1149
1150void ff_h264_set_erpic(ERPicture *dst, H264Picture *src);
1151
1152#endif /* AVCODEC_H264_H */