summaryrefslogtreecommitdiff
path: root/libavcodec/h264.h (plain)
blob: eaffeec6921235924d9fa00a7bff22efdbe655b9
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 "mpegvideo.h"
36#include "h264chroma.h"
37#include "h264dsp.h"
38#include "h264pred.h"
39#include "h264qpel.h"
40#include "rectangle.h"
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#define MAX_MBPAIR_SIZE (256*1024) // a tighter bound could be calculated if someone cares about a few bytes
50
51/* Compiling in interlaced support reduces the speed
52 * of progressive decoding by about 2%. */
53#define ALLOW_INTERLACE
54
55#define FMO 0
56
57/**
58 * The maximum number of slices supported by the decoder.
59 * must be a power of 2
60 */
61#define MAX_SLICES 16
62
63#ifdef ALLOW_INTERLACE
64#define MB_MBAFF(h) h->mb_mbaff
65#define MB_FIELD(h) h->mb_field_decoding_flag
66#define FRAME_MBAFF(h) h->mb_aff_frame
67#define FIELD_PICTURE(h) (h->picture_structure != PICT_FRAME)
68#define LEFT_MBS 2
69#define LTOP 0
70#define LBOT 1
71#define LEFT(i) (i)
72#else
73#define MB_MBAFF(h) 0
74#define MB_FIELD(h) 0
75#define FRAME_MBAFF(h) 0
76#define FIELD_PICTURE(h) 0
77#undef IS_INTERLACED
78#define IS_INTERLACED(mb_type) 0
79#define LEFT_MBS 1
80#define LTOP 0
81#define LBOT 0
82#define LEFT(i) 0
83#endif
84#define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
85
86#ifndef CABAC
87#define CABAC(h) h->pps.cabac
88#endif
89
90#define CHROMA(h) (h->sps.chroma_format_idc)
91#define CHROMA422(h) (h->sps.chroma_format_idc == 2)
92#define CHROMA444(h) (h->sps.chroma_format_idc == 3)
93
94#define EXTENDED_SAR 255
95
96#define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
97#define MB_TYPE_8x8DCT 0x01000000
98#define IS_REF0(a) ((a) & MB_TYPE_REF0)
99#define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
100
101#define QP_MAX_NUM (51 + 6*6) // The maximum supported qp
102
103/* NAL unit types */
104enum {
105 NAL_SLICE = 1,
106 NAL_DPA,
107 NAL_DPB,
108 NAL_DPC,
109 NAL_IDR_SLICE,
110 NAL_SEI,
111 NAL_SPS,
112 NAL_PPS,
113 NAL_AUD,
114 NAL_END_SEQUENCE,
115 NAL_END_STREAM,
116 NAL_FILLER_DATA,
117 NAL_SPS_EXT,
118 NAL_PREFIX = 14,
119 NAL_SUB_SPS,
120 NAL_AUXILIARY_SLICE = 19,
121 NAL_SLC_EXT,
122 NAL_FF_IGNORE = 0xff0f001,
123};
124
125/**
126 * SEI message types
127 */
128typedef enum {
129 SEI_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
130 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
131 SEI_TYPE_USER_DATA_ITU_T_T35 = 4, ///< user data registered by ITU-T Recommendation T.35
132 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
133 SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
134 SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
135} SEI_Type;
136
137/**
138 * pic_struct in picture timing SEI message
139 */
140typedef enum {
141 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
142 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
143 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
144 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
145 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
146 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
147 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
148 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
149 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
150} SEI_PicStructType;
151
152/**
153 * frame_packing_arrangement types
154 */
155typedef enum {
156 SEI_FPA_TYPE_CHECKERBOARD = 0,
157 SEI_FPA_TYPE_INTERLEAVE_COLUMN = 1,
158 SEI_FPA_TYPE_INTERLEAVE_ROW = 2,
159 SEI_FPA_TYPE_SIDE_BY_SIDE = 3,
160 SEI_FPA_TYPE_TOP_BOTTOM = 4,
161 SEI_FPA_TYPE_INTERLEAVE_TEMPORAL = 5,
162 SEI_FPA_TYPE_2D = 6,
163} SEI_FpaType;
164
165/**
166 * Sequence parameter set
167 */
168typedef struct SPS {
169 int profile_idc;
170 int level_idc;
171 int chroma_format_idc;
172 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
173 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
174 int poc_type; ///< pic_order_cnt_type
175 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
176 int delta_pic_order_always_zero_flag;
177 int offset_for_non_ref_pic;
178 int offset_for_top_to_bottom_field;
179 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
180 int ref_frame_count; ///< num_ref_frames
181 int gaps_in_frame_num_allowed_flag;
182 int mb_width; ///< pic_width_in_mbs_minus1 + 1
183 int mb_height; ///< pic_height_in_map_units_minus1 + 1
184 int frame_mbs_only_flag;
185 int mb_aff; ///< mb_adaptive_frame_field_flag
186 int direct_8x8_inference_flag;
187 int crop; ///< frame_cropping_flag
188
189 /* those 4 are already in luma samples */
190 unsigned int crop_left; ///< frame_cropping_rect_left_offset
191 unsigned int crop_right; ///< frame_cropping_rect_right_offset
192 unsigned int crop_top; ///< frame_cropping_rect_top_offset
193 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
194 int vui_parameters_present_flag;
195 AVRational sar;
196 int video_signal_type_present_flag;
197 int full_range;
198 int colour_description_present_flag;
199 enum AVColorPrimaries color_primaries;
200 enum AVColorTransferCharacteristic color_trc;
201 enum AVColorSpace colorspace;
202 int timing_info_present_flag;
203 uint32_t num_units_in_tick;
204 uint32_t time_scale;
205 int fixed_frame_rate_flag;
206 short offset_for_ref_frame[256]; // FIXME dyn aloc?
207 int bitstream_restriction_flag;
208 int num_reorder_frames;
209 int scaling_matrix_present;
210 uint8_t scaling_matrix4[6][16];
211 uint8_t scaling_matrix8[6][64];
212 int nal_hrd_parameters_present_flag;
213 int vcl_hrd_parameters_present_flag;
214 int pic_struct_present_flag;
215 int time_offset_length;
216 int cpb_cnt; ///< See H.264 E.1.2
217 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
218 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
219 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
220 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
221 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
222 int residual_color_transform_flag; ///< residual_colour_transform_flag
223 int constraint_set_flags; ///< constraint_set[0-3]_flag
224 int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
225} SPS;
226
227/**
228 * Picture parameter set
229 */
230typedef struct PPS {
231 unsigned int sps_id;
232 int cabac; ///< entropy_coding_mode_flag
233 int pic_order_present; ///< pic_order_present_flag
234 int slice_group_count; ///< num_slice_groups_minus1 + 1
235 int mb_slice_group_map_type;
236 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
237 int weighted_pred; ///< weighted_pred_flag
238 int weighted_bipred_idc;
239 int init_qp; ///< pic_init_qp_minus26 + 26
240 int init_qs; ///< pic_init_qs_minus26 + 26
241 int chroma_qp_index_offset[2];
242 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
243 int constrained_intra_pred; ///< constrained_intra_pred_flag
244 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
245 int transform_8x8_mode; ///< transform_8x8_mode_flag
246 uint8_t scaling_matrix4[6][16];
247 uint8_t scaling_matrix8[6][64];
248 uint8_t chroma_qp_table[2][QP_MAX_NUM+1]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
249 int chroma_qp_diff;
250} PPS;
251
252/**
253 * Frame Packing Arrangement Type
254 */
255typedef struct FPA {
256 int frame_packing_arrangement_id;
257 int frame_packing_arrangement_cancel_flag; ///< is previous arrangement canceled, -1 if never received
258 SEI_FpaType frame_packing_arrangement_type;
259 int frame_packing_arrangement_repetition_period;
260 int content_interpretation_type;
261 int quincunx_sampling_flag;
262} FPA;
263
264/**
265 * Memory management control operation opcode.
266 */
267typedef enum MMCOOpcode {
268 MMCO_END = 0,
269 MMCO_SHORT2UNUSED,
270 MMCO_LONG2UNUSED,
271 MMCO_SHORT2LONG,
272 MMCO_SET_MAX_LONG,
273 MMCO_RESET,
274 MMCO_LONG,
275} MMCOOpcode;
276
277/**
278 * Memory management control operation.
279 */
280typedef struct MMCO {
281 MMCOOpcode opcode;
282 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
283 int long_arg; ///< index, pic_num, or num long refs depending on opcode
284} MMCO;
285
286/**
287 * H264Context
288 */
289typedef struct H264Context {
290 AVCodecContext *avctx;
291 VideoDSPContext vdsp;
292 H264DSPContext h264dsp;
293 H264ChromaContext h264chroma;
294 H264QpelContext h264qpel;
295 MotionEstContext me;
296 ParseContext parse_context;
297 GetBitContext gb;
298 DSPContext dsp;
299 ERContext er;
300
301 Picture *DPB;
302 Picture *cur_pic_ptr;
303 Picture cur_pic;
304
305 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
306 int chroma_qp[2]; // QPc
307
308 int qp_thresh; ///< QP threshold to skip loopfilter
309
310 /* coded dimensions -- 16 * mb w/h */
311 int width, height;
312 ptrdiff_t linesize, uvlinesize;
313 int chroma_x_shift, chroma_y_shift;
314
315 int qscale;
316 int droppable;
317 int data_partitioning;
318 int coded_picture_number;
319 int low_delay;
320
321 int context_initialized;
322 int flags;
323 int workaround_bugs;
324
325 int prev_mb_skipped;
326 int next_mb_skipped;
327
328 // prediction stuff
329 int chroma_pred_mode;
330 int intra16x16_pred_mode;
331
332 int topleft_mb_xy;
333 int top_mb_xy;
334 int topright_mb_xy;
335 int left_mb_xy[LEFT_MBS];
336
337 int topleft_type;
338 int top_type;
339 int topright_type;
340 int left_type[LEFT_MBS];
341
342 const uint8_t *left_block;
343 int topleft_partition;
344
345 int8_t intra4x4_pred_mode_cache[5 * 8];
346 int8_t(*intra4x4_pred_mode);
347 H264PredContext hpc;
348 unsigned int topleft_samples_available;
349 unsigned int top_samples_available;
350 unsigned int topright_samples_available;
351 unsigned int left_samples_available;
352 uint8_t (*top_borders[2])[(16 * 3) * 2];
353
354 /**
355 * non zero coeff count cache.
356 * is 64 if not available.
357 */
358 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
359
360 uint8_t (*non_zero_count)[48];
361
362 /**
363 * Motion vector cache.
364 */
365 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
366 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
367#define LIST_NOT_USED -1 // FIXME rename?
368#define PART_NOT_AVAILABLE -2
369
370 /**
371 * number of neighbors (top and/or left) that used 8x8 dct
372 */
373 int neighbor_transform_size;
374
375 /**
376 * block_offset[ 0..23] for frame macroblocks
377 * block_offset[24..47] for field macroblocks
378 */
379 int block_offset[2 * (16 * 3)];
380
381 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
382 uint32_t *mb2br_xy;
383 int b_stride; // FIXME use s->b4_stride
384
385 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
386 ptrdiff_t mb_uvlinesize;
387
388 unsigned current_sps_id; ///< id of the current SPS
389 SPS sps; ///< current sps
390
391 /**
392 * current pps
393 */
394 PPS pps; // FIXME move to Picture perhaps? (->no) do we need that?
395
396 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
397 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
398 uint32_t(*dequant4_coeff[6])[16];
399 uint32_t(*dequant8_coeff[6])[64];
400
401 int slice_num;
402 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
403 int slice_type;
404 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
405 int slice_type_fixed;
406
407 // interlacing specific flags
408 int mb_aff_frame;
409 int mb_field_decoding_flag;
410 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
411 int picture_structure;
412 int first_field;
413
414 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
415
416 // Weighted pred stuff
417 int use_weight;
418 int use_weight_chroma;
419 int luma_log2_weight_denom;
420 int chroma_log2_weight_denom;
421 // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
422 int luma_weight[48][2][2];
423 int chroma_weight[48][2][2][2];
424 int implicit_weight[48][48][2];
425
426 int direct_spatial_mv_pred;
427 int col_parity;
428 int col_fieldoff;
429 int dist_scale_factor[32];
430 int dist_scale_factor_field[2][32];
431 int map_col_to_list0[2][16 + 32];
432 int map_col_to_list0_field[2][2][16 + 32];
433
434 /**
435 * num_ref_idx_l0/1_active_minus1 + 1
436 */
437 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
438 unsigned int list_count;
439 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
440 Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
441 * Reordered version of default_ref_list
442 * according to picture reordering in slice header */
443 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
444
445 // data partitioning
446 GetBitContext intra_gb;
447 GetBitContext inter_gb;
448 GetBitContext *intra_gb_ptr;
449 GetBitContext *inter_gb_ptr;
450
451 const uint8_t *intra_pcm_ptr;
452 DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
453 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
454 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
455
456 /**
457 * Cabac
458 */
459 CABACContext cabac;
460 uint8_t cabac_state[1024];
461
462 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
463 uint16_t *cbp_table;
464 int cbp;
465 int top_cbp;
466 int left_cbp;
467 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
468 uint8_t *chroma_pred_mode_table;
469 int last_qscale_diff;
470 uint8_t (*mvd_table[2])[2];
471 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
472 uint8_t *direct_table;
473 uint8_t direct_cache[5 * 8];
474
475 uint8_t zigzag_scan[16];
476 uint8_t zigzag_scan8x8[64];
477 uint8_t zigzag_scan8x8_cavlc[64];
478 uint8_t field_scan[16];
479 uint8_t field_scan8x8[64];
480 uint8_t field_scan8x8_cavlc[64];
481 uint8_t zigzag_scan_q0[16];
482 uint8_t zigzag_scan8x8_q0[64];
483 uint8_t zigzag_scan8x8_cavlc_q0[64];
484 uint8_t field_scan_q0[16];
485 uint8_t field_scan8x8_q0[64];
486 uint8_t field_scan8x8_cavlc_q0[64];
487
488 int x264_build;
489
490 int mb_x, mb_y;
491 int resync_mb_x;
492 int resync_mb_y;
493 int mb_skip_run;
494 int mb_height, mb_width;
495 int mb_stride;
496 int mb_num;
497 int mb_xy;
498
499 int is_complex;
500
501 // deblock
502 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
503 int slice_alpha_c0_offset;
504 int slice_beta_offset;
505
506 // =============================================================
507 // Things below are not used in the MB or more inner code
508
509 int nal_ref_idc;
510 int nal_unit_type;
511 uint8_t *rbsp_buffer[2];
512 unsigned int rbsp_buffer_size[2];
513
514 /**
515 * Used to parse AVC variant of h264
516 */
517 int is_avc; ///< this flag is != 0 if codec is avc1
518 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
519 int got_first; ///< this flag is != 0 if we've parsed a frame
520
521 int bit_depth_luma; ///< luma bit depth from sps to detect changes
522 int chroma_format_idc; ///< chroma format from sps to detect changes
523
524 SPS *sps_buffers[MAX_SPS_COUNT];
525 PPS *pps_buffers[MAX_PPS_COUNT];
526
527 int dequant_coeff_pps; ///< reinit tables when pps changes
528
529 uint16_t *slice_table_base;
530
531 // POC stuff
532 int poc_lsb;
533 int poc_msb;
534 int delta_poc_bottom;
535 int delta_poc[2];
536 int frame_num;
537 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
538 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
539 int frame_num_offset; ///< for POC type 2
540 int prev_frame_num_offset; ///< for POC type 2
541 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
542
543 /**
544 * frame_num for frames or 2 * frame_num + 1 for field pics.
545 */
546 int curr_pic_num;
547
548 /**
549 * max_frame_num or 2 * max_frame_num for field pics.
550 */
551 int max_pic_num;
552
553 int redundant_pic_count;
554
555 Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
556 Picture *short_ref[32];
557 Picture *long_ref[32];
558 Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
559 int last_pocs[MAX_DELAYED_PIC_COUNT];
560 Picture *next_output_pic;
561 int outputed_poc;
562 int next_outputed_poc;
563
564 /**
565 * memory management control operations buffer.
566 */
567 MMCO mmco[MAX_MMCO_COUNT];
568 int mmco_index;
569 int mmco_reset;
570
571 int long_ref_count; ///< number of actual long term references
572 int short_ref_count; ///< number of actual short term references
573
574 int cabac_init_idc;
575
576 /**
577 * @name Members for slice based multithreading
578 * @{
579 */
580 struct H264Context *thread_context[MAX_THREADS];
581
582 /**
583 * current slice number, used to initialize slice_num of each thread/context
584 */
585 int current_slice;
586
587 /**
588 * Max number of threads / contexts.
589 * This is equal to AVCodecContext.thread_count unless
590 * multithreaded decoding is impossible, in which case it is
591 * reduced to 1.
592 */
593 int max_contexts;
594
595 int slice_context_count;
596
597 /**
598 * 1 if the single thread fallback warning has already been
599 * displayed, 0 otherwise.
600 */
601 int single_decode_warning;
602
603 enum AVPictureType pict_type;
604
605 int last_slice_type;
606 unsigned int last_ref_count[2];
607 /** @} */
608
609 /**
610 * pic_struct in picture timing SEI message
611 */
612 SEI_PicStructType sei_pic_struct;
613
614 /**
615 * Complement sei_pic_struct
616 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
617 * However, soft telecined frames may have these values.
618 * This is used in an attempt to flag soft telecine progressive.
619 */
620 int prev_interlaced_frame;
621
622 /**
623 * Bit set of clock types for fields/frames in picture timing SEI message.
624 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
625 * interlaced).
626 */
627 int sei_ct_type;
628
629 /**
630 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
631 */
632 int sei_dpb_output_delay;
633
634 /**
635 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
636 */
637 int sei_cpb_removal_delay;
638
639 /**
640 * recovery_frame_cnt from SEI message
641 *
642 * Set to -1 if no recovery point SEI message found or to number of frames
643 * before playback synchronizes. Frames having recovery point are key
644 * frames.
645 */
646 int sei_recovery_frame_cnt;
647 /**
648 * recovery_frame is the frame_num at which the next frame should
649 * be fully constructed.
650 *
651 * Set to -1 when not expecting a recovery point.
652 */
653 int recovery_frame;
654
655 /**
656 * Are the SEI recovery points looking valid.
657 */
658 int valid_recovery_point;
659
660 FPA sei_fpa;
661
662 int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
663 int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
664
665 // Timestamp stuff
666 int sei_buffering_period_present; ///< Buffering period SEI flag
667 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
668
669 int cur_chroma_format_idc;
670 uint8_t *bipred_scratchpad;
671
672 int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
673
674 int sync; ///< did we had a keyframe or recovery point
675
676 uint8_t parse_history[4];
677 int parse_history_count;
678 int parse_last_mb;
679 uint8_t *edge_emu_buffer;
680 int16_t *dc_val_base;
681
682 uint8_t *visualization_buffer[3]; ///< temporary buffer vor MV visualization
683
684 AVBufferPool *qscale_table_pool;
685 AVBufferPool *mb_type_pool;
686 AVBufferPool *motion_val_pool;
687 AVBufferPool *ref_index_pool;
688} H264Context;
689
690extern const uint8_t ff_h264_chroma_qp[7][QP_MAX_NUM + 1]; ///< One chroma qp table for each possible bit depth (8-14).
691extern const uint16_t ff_h264_mb_sizes[4];
692
693/**
694 * Decode SEI
695 */
696int ff_h264_decode_sei(H264Context *h);
697
698/**
699 * Decode SPS
700 */
701int ff_h264_decode_seq_parameter_set(H264Context *h);
702
703/**
704 * compute profile from sps
705 */
706int ff_h264_get_profile(SPS *sps);
707
708/**
709 * Decode PPS
710 */
711int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
712
713/**
714 * Decode a network abstraction layer unit.
715 * @param consumed is the number of bytes used as input
716 * @param length is the length of the array
717 * @param dst_length is the number of decoded bytes FIXME here
718 * or a decode rbsp tailing?
719 * @return decoded bytes, might be src+1 if no escapes
720 */
721const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
722 int *dst_length, int *consumed, int length);
723
724/**
725 * Free any data that may have been allocated in the H264 context
726 * like SPS, PPS etc.
727 */
728void ff_h264_free_context(H264Context *h);
729
730/**
731 * Reconstruct bitstream slice_type.
732 */
733int ff_h264_get_slice_type(const H264Context *h);
734
735/**
736 * Allocate tables.
737 * needs width/height
738 */
739int ff_h264_alloc_tables(H264Context *h);
740
741/**
742 * Fill the default_ref_list.
743 */
744int ff_h264_fill_default_ref_list(H264Context *h);
745
746int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
747void ff_h264_fill_mbaff_ref_list(H264Context *h);
748void ff_h264_remove_all_refs(H264Context *h);
749
750/**
751 * Execute the reference picture marking (memory management control operations).
752 */
753int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
754
755int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
756 int first_slice);
757
758int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
759
760/**
761 * Check if the top & left blocks are available if needed & change the
762 * dc mode so it only uses the available blocks.
763 */
764int ff_h264_check_intra4x4_pred_mode(H264Context *h);
765
766/**
767 * Check if the top & left blocks are available if needed & change the
768 * dc mode so it only uses the available blocks.
769 */
770int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
771
772void ff_h264_hl_decode_mb(H264Context *h);
773int ff_h264_decode_extradata(H264Context *h, const uint8_t *buf, int size);
774int ff_h264_decode_init(AVCodecContext *avctx);
775void ff_h264_decode_init_vlc(void);
776
777/**
778 * Decode a macroblock
779 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
780 */
781int ff_h264_decode_mb_cavlc(H264Context *h);
782
783/**
784 * Decode a CABAC coded macroblock
785 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
786 */
787int ff_h264_decode_mb_cabac(H264Context *h);
788
789void ff_h264_init_cabac_states(H264Context *h);
790
791void ff_h264_direct_dist_scale_factor(H264Context *const h);
792void ff_h264_direct_ref_list_init(H264Context *const h);
793void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type);
794
795void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y,
796 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
797 unsigned int linesize, unsigned int uvlinesize);
798void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
799 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
800 unsigned int linesize, unsigned int uvlinesize);
801
802/**
803 * Reset SEI values at the beginning of the frame.
804 *
805 * @param h H.264 context.
806 */
807void ff_h264_reset_sei(H264Context *h);
808
809/**
810 * Get stereo_mode string from the h264 frame_packing_arrangement
811 * @param h H.264 context.
812 */
813const char* ff_h264_sei_stereo_mode(H264Context *h);
814
815/*
816 * o-o o-o
817 * / / /
818 * o-o o-o
819 * ,---'
820 * o-o o-o
821 * / / /
822 * o-o o-o
823 */
824
825/* Scan8 organization:
826 * 0 1 2 3 4 5 6 7
827 * 0 DY y y y y y
828 * 1 y Y Y Y Y
829 * 2 y Y Y Y Y
830 * 3 y Y Y Y Y
831 * 4 y Y Y Y Y
832 * 5 DU u u u u u
833 * 6 u U U U U
834 * 7 u U U U U
835 * 8 u U U U U
836 * 9 u U U U U
837 * 10 DV v v v v v
838 * 11 v V V V V
839 * 12 v V V V V
840 * 13 v V V V V
841 * 14 v V V V V
842 * DY/DU/DV are for luma/chroma DC.
843 */
844
845#define LUMA_DC_BLOCK_INDEX 48
846#define CHROMA_DC_BLOCK_INDEX 49
847
848// This table must be here because scan8[constant] must be known at compiletime
849static const uint8_t scan8[16 * 3 + 3] = {
850 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
851 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
852 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
853 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
854 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
855 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
856 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
857 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
858 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
859 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
860 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
861 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
862 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
863};
864
865static av_always_inline uint32_t pack16to32(int a, int b)
866{
867#if HAVE_BIGENDIAN
868 return (b & 0xFFFF) + (a << 16);
869#else
870 return (a & 0xFFFF) + (b << 16);
871#endif
872}
873
874static av_always_inline uint16_t pack8to16(int a, int b)
875{
876#if HAVE_BIGENDIAN
877 return (b & 0xFF) + (a << 8);
878#else
879 return (a & 0xFF) + (b << 8);
880#endif
881}
882
883/**
884 * Get the chroma qp.
885 */
886static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
887{
888 return h->pps.chroma_qp_table[t][qscale];
889}
890
891/**
892 * Get the predicted intra4x4 prediction mode.
893 */
894static av_always_inline int pred_intra_mode(H264Context *h, int n)
895{
896 const int index8 = scan8[n];
897 const int left = h->intra4x4_pred_mode_cache[index8 - 1];
898 const int top = h->intra4x4_pred_mode_cache[index8 - 8];
899 const int min = FFMIN(left, top);
900
901 tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min);
902
903 if (min < 0)
904 return DC_PRED;
905 else
906 return min;
907}
908
909static av_always_inline void write_back_intra_pred_mode(H264Context *h)
910{
911 int8_t *i4x4 = h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
912 int8_t *i4x4_cache = h->intra4x4_pred_mode_cache;
913
914 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
915 i4x4[4] = i4x4_cache[7 + 8 * 3];
916 i4x4[5] = i4x4_cache[7 + 8 * 2];
917 i4x4[6] = i4x4_cache[7 + 8 * 1];
918}
919
920static av_always_inline void write_back_non_zero_count(H264Context *h)
921{
922 const int mb_xy = h->mb_xy;
923 uint8_t *nnz = h->non_zero_count[mb_xy];
924 uint8_t *nnz_cache = h->non_zero_count_cache;
925
926 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
927 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
928 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
929 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
930 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
931 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
932 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
933 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
934
935 if (!h->chroma_y_shift) {
936 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
937 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
938 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
939 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
940 }
941}
942
943static av_always_inline void write_back_motion_list(H264Context *h,
944 int b_stride,
945 int b_xy, int b8_xy,
946 int mb_type, int list)
947{
948 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
949 int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
950 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
951 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
952 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
953 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
954 if (CABAC(h)) {
955 uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8 * h->mb_xy
956 : h->mb2br_xy[h->mb_xy]];
957 uint8_t(*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
958 if (IS_SKIP(mb_type)) {
959 AV_ZERO128(mvd_dst);
960 } else {
961 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
962 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
963 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
964 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
965 }
966 }
967
968 {
969 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
970 int8_t *ref_cache = h->ref_cache[list];
971 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
972 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
973 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
974 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
975 }
976}
977
978static av_always_inline void write_back_motion(H264Context *h, int mb_type)
979{
980 const int b_stride = h->b_stride;
981 const int b_xy = 4 * h->mb_x + 4 * h->mb_y * h->b_stride; // try mb2b(8)_xy
982 const int b8_xy = 4 * h->mb_xy;
983
984 if (USES_LIST(mb_type, 0)) {
985 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 0);
986 } else {
987 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
988 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
989 }
990 if (USES_LIST(mb_type, 1))
991 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 1);
992
993 if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
994 if (IS_8X8(mb_type)) {
995 uint8_t *direct_table = &h->direct_table[4 * h->mb_xy];
996 direct_table[1] = h->sub_mb_type[1] >> 1;
997 direct_table[2] = h->sub_mb_type[2] >> 1;
998 direct_table[3] = h->sub_mb_type[3] >> 1;
999 }
1000 }
1001}
1002
1003static av_always_inline int get_dct8x8_allowed(H264Context *h)
1004{
1005 if (h->sps.direct_8x8_inference_flag)
1006 return !(AV_RN64A(h->sub_mb_type) &
1007 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
1008 0x0001000100010001ULL));
1009 else
1010 return !(AV_RN64A(h->sub_mb_type) &
1011 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
1012 0x0001000100010001ULL));
1013}
1014
1015void ff_h264_draw_horiz_band(H264Context *h, int y, int height);
1016int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
1017int ff_pred_weight_table(H264Context *h);
1018int ff_set_ref_count(H264Context *h);
1019
1020#endif /* AVCODEC_H264_H */
1021
generated by cgit at 2024-05-17 06:30:29 (GMT)