e00ce8857fc9347ea4173b9aa655308eeb78a22b
[libav.git] / libavcodec / h264.h
1 /*
2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4 *
5 * This file is part of Libav.
6 *
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * H.264 / AVC / MPEG4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28 #ifndef AVCODEC_H264_H
29 #define AVCODEC_H264_H
30
31 #include "libavutil/intreadwrite.h"
32 #include "libavutil/thread.h"
33 #include "cabac.h"
34 #include "error_resilience.h"
35 #include "get_bits.h"
36 #include "h264_parse.h"
37 #include "h264chroma.h"
38 #include "h264dsp.h"
39 #include "h264pred.h"
40 #include "h264qpel.h"
41 #include "internal.h"
42 #include "mpegutils.h"
43 #include "parser.h"
44 #include "qpeldsp.h"
45 #include "rectangle.h"
46 #include "videodsp.h"
47
48 #define H264_MAX_PICTURE_COUNT 32
49 #define H264_MAX_THREADS 16
50
51 #define MAX_SPS_COUNT 32
52 #define MAX_PPS_COUNT 256
53
54 #define MAX_MMCO_COUNT 66
55
56 #define MAX_DELAYED_PIC_COUNT 16
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 CHROMA422(h) (h->sps.chroma_format_idc == 2)
98 #define CHROMA444(h) (h->sps.chroma_format_idc == 3)
99
100 #define EXTENDED_SAR 255
101
102 #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
103 #define MB_TYPE_8x8DCT 0x01000000
104 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
105 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
106
107 #define QP_MAX_NUM (51 + 2 * 6) // The maximum supported qp
108
109 /* NAL unit types */
110 enum {
111 NAL_SLICE = 1,
112 NAL_DPA = 2,
113 NAL_DPB = 3,
114 NAL_DPC = 4,
115 NAL_IDR_SLICE = 5,
116 NAL_SEI = 6,
117 NAL_SPS = 7,
118 NAL_PPS = 8,
119 NAL_AUD = 9,
120 NAL_END_SEQUENCE = 10,
121 NAL_END_STREAM = 11,
122 NAL_FILLER_DATA = 12,
123 NAL_SPS_EXT = 13,
124 NAL_AUXILIARY_SLICE = 19,
125 NAL_FF_IGNORE = 0xff0f001,
126 };
127
128 /**
129 * SEI message types
130 */
131 typedef enum {
132 SEI_TYPE_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
133 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
134 SEI_TYPE_USER_DATA_REGISTERED = 4, ///< registered user data as specified by Rec. ITU-T T.35
135 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
136 SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
137 SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
138 SEI_TYPE_DISPLAY_ORIENTATION = 47, ///< display orientation
139 } SEI_Type;
140
141 /**
142 * pic_struct in picture timing SEI message
143 */
144 typedef enum {
145 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
146 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
147 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
148 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
149 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
150 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
151 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
152 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
153 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
154 } SEI_PicStructType;
155
156 /**
157 * Sequence parameter set
158 */
159 typedef struct SPS {
160 unsigned int sps_id;
161 int profile_idc;
162 int level_idc;
163 int chroma_format_idc;
164 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
165 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
166 int poc_type; ///< pic_order_cnt_type
167 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
168 int delta_pic_order_always_zero_flag;
169 int offset_for_non_ref_pic;
170 int offset_for_top_to_bottom_field;
171 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
172 int ref_frame_count; ///< num_ref_frames
173 int gaps_in_frame_num_allowed_flag;
174 int mb_width; ///< pic_width_in_mbs_minus1 + 1
175 int mb_height; ///< pic_height_in_map_units_minus1 + 1
176 int frame_mbs_only_flag;
177 int mb_aff; ///< mb_adaptive_frame_field_flag
178 int direct_8x8_inference_flag;
179 int crop; ///< frame_cropping_flag
180
181 /* those 4 are already in luma samples */
182 unsigned int crop_left; ///< frame_cropping_rect_left_offset
183 unsigned int crop_right; ///< frame_cropping_rect_right_offset
184 unsigned int crop_top; ///< frame_cropping_rect_top_offset
185 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
186 int vui_parameters_present_flag;
187 AVRational sar;
188 int video_signal_type_present_flag;
189 int full_range;
190 int colour_description_present_flag;
191 enum AVColorPrimaries color_primaries;
192 enum AVColorTransferCharacteristic color_trc;
193 enum AVColorSpace colorspace;
194 int timing_info_present_flag;
195 uint32_t num_units_in_tick;
196 uint32_t time_scale;
197 int fixed_frame_rate_flag;
198 short offset_for_ref_frame[256]; // FIXME dyn aloc?
199 int bitstream_restriction_flag;
200 int num_reorder_frames;
201 int scaling_matrix_present;
202 uint8_t scaling_matrix4[6][16];
203 uint8_t scaling_matrix8[6][64];
204 int nal_hrd_parameters_present_flag;
205 int vcl_hrd_parameters_present_flag;
206 int pic_struct_present_flag;
207 int time_offset_length;
208 int cpb_cnt; ///< See H.264 E.1.2
209 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
210 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
211 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
212 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
213 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
214 int residual_color_transform_flag; ///< residual_colour_transform_flag
215 int constraint_set_flags; ///< constraint_set[0-3]_flag
216 int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
217 } SPS;
218
219 /**
220 * Picture parameter set
221 */
222 typedef struct PPS {
223 unsigned int sps_id;
224 int cabac; ///< entropy_coding_mode_flag
225 int pic_order_present; ///< pic_order_present_flag
226 int slice_group_count; ///< num_slice_groups_minus1 + 1
227 int mb_slice_group_map_type;
228 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
229 int weighted_pred; ///< weighted_pred_flag
230 int weighted_bipred_idc;
231 int init_qp; ///< pic_init_qp_minus26 + 26
232 int init_qs; ///< pic_init_qs_minus26 + 26
233 int chroma_qp_index_offset[2];
234 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
235 int constrained_intra_pred; ///< constrained_intra_pred_flag
236 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
237 int transform_8x8_mode; ///< transform_8x8_mode_flag
238 uint8_t scaling_matrix4[6][16];
239 uint8_t scaling_matrix8[6][64];
240 uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
241 int chroma_qp_diff;
242 } PPS;
243
244 /**
245 * Memory management control operation opcode.
246 */
247 typedef enum MMCOOpcode {
248 MMCO_END = 0,
249 MMCO_SHORT2UNUSED,
250 MMCO_LONG2UNUSED,
251 MMCO_SHORT2LONG,
252 MMCO_SET_MAX_LONG,
253 MMCO_RESET,
254 MMCO_LONG,
255 } MMCOOpcode;
256
257 /**
258 * Memory management control operation.
259 */
260 typedef struct MMCO {
261 MMCOOpcode opcode;
262 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
263 int long_arg; ///< index, pic_num, or num long refs depending on opcode
264 } MMCO;
265
266 typedef struct H264Picture {
267 AVFrame *f;
268 ThreadFrame tf;
269
270 AVBufferRef *qscale_table_buf;
271 int8_t *qscale_table;
272
273 AVBufferRef *motion_val_buf[2];
274 int16_t (*motion_val[2])[2];
275
276 AVBufferRef *mb_type_buf;
277 uint32_t *mb_type;
278
279 AVBufferRef *hwaccel_priv_buf;
280 void *hwaccel_picture_private; ///< hardware accelerator private data
281
282 AVBufferRef *ref_index_buf[2];
283 int8_t *ref_index[2];
284
285 int field_poc[2]; ///< top/bottom POC
286 int poc; ///< frame POC
287 int frame_num; ///< frame_num (raw frame_num from slice header)
288 int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
289 not mix pictures before and after MMCO_RESET. */
290 int pic_id; /**< pic_num (short -> no wrap version of pic_num,
291 pic_num & max_pic_num; long -> long_pic_num) */
292 int long_ref; ///< 1->long term reference 0->short term reference
293 int ref_poc[2][2][32]; ///< POCs of the frames used as reference (FIXME need per slice)
294 int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
295 int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
296 int field_picture; ///< whether or not picture was encoded in separate fields
297
298 int reference;
299 int recovered; ///< picture at IDR or recovery point + recovery count
300 } H264Picture;
301
302 typedef struct H264Ref {
303 uint8_t *data[3];
304 int linesize[3];
305
306 int reference;
307 int poc;
308 int pic_id;
309
310 H264Picture *parent;
311 } H264Ref;
312
313 typedef struct H264SliceContext {
314 struct H264Context *h264;
315 GetBitContext gb;
316 ERContext er;
317
318 int slice_num;
319 int slice_type;
320 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
321 int slice_type_fixed;
322
323 int qscale;
324 int chroma_qp[2]; // QPc
325 int qp_thresh; ///< QP threshold to skip loopfilter
326 int last_qscale_diff;
327
328 // deblock
329 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
330 int slice_alpha_c0_offset;
331 int slice_beta_offset;
332
333 H264PredWeightTable pwt;
334
335 int prev_mb_skipped;
336 int next_mb_skipped;
337
338 int chroma_pred_mode;
339 int intra16x16_pred_mode;
340
341 int8_t intra4x4_pred_mode_cache[5 * 8];
342 int8_t(*intra4x4_pred_mode);
343
344 int topleft_mb_xy;
345 int top_mb_xy;
346 int topright_mb_xy;
347 int left_mb_xy[LEFT_MBS];
348
349 int topleft_type;
350 int top_type;
351 int topright_type;
352 int left_type[LEFT_MBS];
353
354 const uint8_t *left_block;
355 int topleft_partition;
356
357 unsigned int topleft_samples_available;
358 unsigned int top_samples_available;
359 unsigned int topright_samples_available;
360 unsigned int left_samples_available;
361
362 ptrdiff_t linesize, uvlinesize;
363 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
364 ptrdiff_t mb_uvlinesize;
365
366 int mb_x, mb_y;
367 int mb_xy;
368 int resync_mb_x;
369 int resync_mb_y;
370 // index of the first MB of the next slice
371 int next_slice_idx;
372 int mb_skip_run;
373 int is_complex;
374
375 int mb_field_decoding_flag;
376 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
377
378 int redundant_pic_count;
379
380 /**
381 * number of neighbors (top and/or left) that used 8x8 dct
382 */
383 int neighbor_transform_size;
384
385 int direct_spatial_mv_pred;
386 int col_parity;
387 int col_fieldoff;
388
389 int cbp;
390 int top_cbp;
391 int left_cbp;
392
393 int dist_scale_factor[32];
394 int dist_scale_factor_field[2][32];
395 int map_col_to_list0[2][16 + 32];
396 int map_col_to_list0_field[2][2][16 + 32];
397
398 /**
399 * num_ref_idx_l0/1_active_minus1 + 1
400 */
401 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
402 unsigned int list_count;
403 H264Ref ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
404 * Reordered version of default_ref_list
405 * according to picture reordering in slice header */
406 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
407
408 const uint8_t *intra_pcm_ptr;
409 int16_t *dc_val_base;
410
411 uint8_t *bipred_scratchpad;
412 uint8_t *edge_emu_buffer;
413 uint8_t (*top_borders[2])[(16 * 3) * 2];
414 int bipred_scratchpad_allocated;
415 int edge_emu_buffer_allocated;
416 int top_borders_allocated[2];
417
418 /**
419 * non zero coeff count cache.
420 * is 64 if not available.
421 */
422 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
423
424 /**
425 * Motion vector cache.
426 */
427 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
428 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
429 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
430 uint8_t direct_cache[5 * 8];
431
432 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
433
434 ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
435 DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2];
436 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
437 ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either
438 ///< check that i is not too large or ensure that there is some unused stuff after mb
439 int16_t mb_padding[256 * 2];
440
441 uint8_t (*mvd_table[2])[2];
442
443 /**
444 * Cabac
445 */
446 CABACContext cabac;
447 uint8_t cabac_state[1024];
448 int cabac_init_idc;
449
450 // rbsp buffer used for this slice
451 uint8_t *rbsp_buffer;
452 unsigned int rbsp_buffer_size;
453 } H264SliceContext;
454
455 /**
456 * H264Context
457 */
458 typedef struct H264Context {
459 const AVClass *class;
460 AVCodecContext *avctx;
461 VideoDSPContext vdsp;
462 H264DSPContext h264dsp;
463 H264ChromaContext h264chroma;
464 H264QpelContext h264qpel;
465 GetBitContext gb;
466
467 H264Picture DPB[H264_MAX_PICTURE_COUNT];
468 H264Picture *cur_pic_ptr;
469 H264Picture cur_pic;
470
471 H264SliceContext *slice_ctx;
472 int nb_slice_ctx;
473
474 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
475
476 /* coded dimensions -- 16 * mb w/h */
477 int width, height;
478 int chroma_x_shift, chroma_y_shift;
479
480 int droppable;
481 int coded_picture_number;
482 int low_delay;
483
484 int context_initialized;
485 int flags;
486 int workaround_bugs;
487
488 int8_t(*intra4x4_pred_mode);
489 H264PredContext hpc;
490
491 uint8_t (*non_zero_count)[48];
492
493 #define LIST_NOT_USED -1 // FIXME rename?
494 #define PART_NOT_AVAILABLE -2
495
496 /**
497 * block_offset[ 0..23] for frame macroblocks
498 * block_offset[24..47] for field macroblocks
499 */
500 int block_offset[2 * (16 * 3)];
501
502 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
503 uint32_t *mb2br_xy;
504 int b_stride; // FIXME use s->b4_stride
505
506 SPS sps; ///< current sps
507 PPS pps; ///< current pps
508
509 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
510 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
511 uint32_t(*dequant4_coeff[6])[16];
512 uint32_t(*dequant8_coeff[6])[64];
513
514 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
515
516 // interlacing specific flags
517 int mb_aff_frame;
518 int picture_structure;
519 int first_field;
520
521 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
522
523 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
524 uint16_t *cbp_table;
525
526 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
527 uint8_t *chroma_pred_mode_table;
528 uint8_t (*mvd_table[2])[2];
529 uint8_t *direct_table;
530
531 uint8_t zigzag_scan[16];
532 uint8_t zigzag_scan8x8[64];
533 uint8_t zigzag_scan8x8_cavlc[64];
534 uint8_t field_scan[16];
535 uint8_t field_scan8x8[64];
536 uint8_t field_scan8x8_cavlc[64];
537 const uint8_t *zigzag_scan_q0;
538 const uint8_t *zigzag_scan8x8_q0;
539 const uint8_t *zigzag_scan8x8_cavlc_q0;
540 const uint8_t *field_scan_q0;
541 const uint8_t *field_scan8x8_q0;
542 const uint8_t *field_scan8x8_cavlc_q0;
543
544 int x264_build;
545
546 int mb_y;
547 int mb_height, mb_width;
548 int mb_stride;
549 int mb_num;
550
551 // =============================================================
552 // Things below are not used in the MB or more inner code
553
554 int nal_ref_idc;
555 int nal_unit_type;
556
557 /**
558 * Used to parse AVC variant of h264
559 */
560 int is_avc; ///< this flag is != 0 if codec is avc1
561 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
562
563 int bit_depth_luma; ///< luma bit depth from sps to detect changes
564 int chroma_format_idc; ///< chroma format from sps to detect changes
565
566 SPS *sps_buffers[MAX_SPS_COUNT];
567 PPS *pps_buffers[MAX_PPS_COUNT];
568
569 int dequant_coeff_pps; ///< reinit tables when pps changes
570
571 uint16_t *slice_table_base;
572
573 // POC stuff
574 int poc_lsb;
575 int poc_msb;
576 int delta_poc_bottom;
577 int delta_poc[2];
578 int frame_num;
579 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
580 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
581 int frame_num_offset; ///< for POC type 2
582 int prev_frame_num_offset; ///< for POC type 2
583 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
584
585 /**
586 * frame_num for frames or 2 * frame_num + 1 for field pics.
587 */
588 int curr_pic_num;
589
590 /**
591 * max_frame_num or 2 * max_frame_num for field pics.
592 */
593 int max_pic_num;
594
595 H264Picture *short_ref[32];
596 H264Picture *long_ref[32];
597 H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
598 int last_pocs[MAX_DELAYED_PIC_COUNT];
599 H264Picture *next_output_pic;
600 int next_outputed_poc;
601
602 /**
603 * memory management control operations buffer.
604 */
605 MMCO mmco[MAX_MMCO_COUNT];
606 int mmco_index;
607 int mmco_reset;
608
609 int long_ref_count; ///< number of actual long term references
610 int short_ref_count; ///< number of actual short term references
611
612 /**
613 * @name Members for slice based multithreading
614 * @{
615 */
616 /**
617 * current slice number, used to initalize slice_num of each thread/context
618 */
619 int current_slice;
620
621 /**
622 * Max number of threads / contexts.
623 * This is equal to AVCodecContext.thread_count unless
624 * multithreaded decoding is impossible, in which case it is
625 * reduced to 1.
626 */
627 int max_contexts;
628
629 int slice_context_count;
630
631 /**
632 * 1 if the single thread fallback warning has already been
633 * displayed, 0 otherwise.
634 */
635 int single_decode_warning;
636
637 enum AVPictureType pict_type;
638
639 /** @} */
640
641 /**
642 * pic_struct in picture timing SEI message
643 */
644 SEI_PicStructType sei_pic_struct;
645
646 /**
647 * Complement sei_pic_struct
648 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
649 * However, soft telecined frames may have these values.
650 * This is used in an attempt to flag soft telecine progressive.
651 */
652 int prev_interlaced_frame;
653
654 /**
655 * frame_packing_arrangment SEI message
656 */
657 int sei_frame_packing_present;
658 int frame_packing_arrangement_type;
659 int content_interpretation_type;
660 int quincunx_subsampling;
661
662 /**
663 * display orientation SEI message
664 */
665 int sei_display_orientation_present;
666 int sei_anticlockwise_rotation;
667 int sei_hflip, sei_vflip;
668
669 /**
670 * User data registered by Rec. ITU-T T.35 SEI
671 */
672 int sei_reguserdata_afd_present;
673 uint8_t active_format_description;
674 int a53_caption_size;
675 uint8_t *a53_caption;
676
677 /**
678 * Bit set of clock types for fields/frames in picture timing SEI message.
679 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
680 * interlaced).
681 */
682 int sei_ct_type;
683
684 /**
685 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
686 */
687 int sei_dpb_output_delay;
688
689 /**
690 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
691 */
692 int sei_cpb_removal_delay;
693
694 /**
695 * recovery_frame_cnt from SEI message
696 *
697 * Set to -1 if no recovery point SEI message found or to number of frames
698 * before playback synchronizes. Frames having recovery point are key
699 * frames.
700 */
701 int sei_recovery_frame_cnt;
702
703 /**
704 * recovery_frame is the frame_num at which the next frame should
705 * be fully constructed.
706 *
707 * Set to -1 when not expecting a recovery point.
708 */
709 int recovery_frame;
710
711 /**
712 * We have seen an IDR, so all the following frames in coded order are correctly
713 * decodable.
714 */
715 #define FRAME_RECOVERED_IDR (1 << 0)
716 /**
717 * Sufficient number of frames have been decoded since a SEI recovery point,
718 * so all the following frames in presentation order are correct.
719 */
720 #define FRAME_RECOVERED_SEI (1 << 1)
721
722 int frame_recovered; ///< Initial frame has been completely recovered
723
724 /* for frame threading, this is set to 1
725 * after finish_setup() has been called, so we cannot modify
726 * some context properties (which are supposed to stay constant between
727 * slices) anymore */
728 int setup_finished;
729
730 // Timestamp stuff
731 int sei_buffering_period_present; ///< Buffering period SEI flag
732 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
733
734 int enable_er;
735
736 AVBufferPool *qscale_table_pool;
737 AVBufferPool *mb_type_pool;
738 AVBufferPool *motion_val_pool;
739 AVBufferPool *ref_index_pool;
740
741 /* Motion Estimation */
742 qpel_mc_func (*qpel_put)[16];
743 qpel_mc_func (*qpel_avg)[16];
744 } H264Context;
745
746 extern const uint8_t ff_h264_chroma_qp[3][QP_MAX_NUM + 1]; ///< One chroma qp table for each supported bit depth (8, 9, 10).
747 extern const uint16_t ff_h264_mb_sizes[4];
748
749 /**
750 * Decode SEI
751 */
752 int ff_h264_decode_sei(H264Context *h);
753
754 /**
755 * Decode SPS
756 */
757 int ff_h264_decode_seq_parameter_set(H264Context *h);
758
759 /**
760 * compute profile from sps
761 */
762 int ff_h264_get_profile(SPS *sps);
763
764 /**
765 * Decode PPS
766 */
767 int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
768
769 /**
770 * Decode a network abstraction layer unit.
771 * @param consumed is the number of bytes used as input
772 * @param length is the length of the array
773 * @param dst_length is the number of decoded bytes FIXME here
774 * or a decode rbsp tailing?
775 * @return decoded bytes, might be src+1 if no escapes
776 */
777 const uint8_t *ff_h264_decode_nal(H264Context *h, H264SliceContext *sl, const uint8_t *src,
778 int *dst_length, int *consumed, int length);
779
780 /**
781 * Free any data that may have been allocated in the H264 context
782 * like SPS, PPS etc.
783 */
784 void ff_h264_free_context(H264Context *h);
785
786 /**
787 * Reconstruct bitstream slice_type.
788 */
789 int ff_h264_get_slice_type(const H264SliceContext *sl);
790
791 /**
792 * Allocate tables.
793 * needs width/height
794 */
795 int ff_h264_alloc_tables(H264Context *h);
796
797 int ff_h264_decode_ref_pic_list_reordering(H264Context *h, H264SliceContext *sl);
798 void ff_h264_fill_mbaff_ref_list(H264Context *h, H264SliceContext *sl);
799 void ff_h264_remove_all_refs(H264Context *h);
800
801 /**
802 * Execute the reference picture marking (memory management control operations).
803 */
804 int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
805
806 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
807 int first_slice);
808
809 int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
810
811 void ff_h264_hl_decode_mb(const H264Context *h, H264SliceContext *sl);
812 int ff_h264_decode_extradata(H264Context *h);
813 int ff_h264_decode_init(AVCodecContext *avctx);
814 void ff_h264_decode_init_vlc(void);
815
816 /**
817 * Decode a macroblock
818 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
819 */
820 int ff_h264_decode_mb_cavlc(const H264Context *h, H264SliceContext *sl);
821
822 /**
823 * Decode a CABAC coded macroblock
824 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
825 */
826 int ff_h264_decode_mb_cabac(const H264Context *h, H264SliceContext *sl);
827
828 void ff_h264_init_cabac_states(const H264Context *h, H264SliceContext *sl);
829
830 void h264_init_dequant_tables(H264Context *h);
831
832 void ff_h264_direct_dist_scale_factor(const H264Context *const h, H264SliceContext *sl);
833 void ff_h264_direct_ref_list_init(const H264Context *const h, H264SliceContext *sl);
834 void ff_h264_pred_direct_motion(const H264Context *const h, H264SliceContext *sl,
835 int *mb_type);
836
837 void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
838 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
839 unsigned int linesize, unsigned int uvlinesize);
840 void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
841 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
842 unsigned int linesize, unsigned int uvlinesize);
843
844 /**
845 * Reset SEI values at the beginning of the frame.
846 *
847 * @param h H.264 context.
848 */
849 void ff_h264_reset_sei(H264Context *h);
850
851 /*
852 * o-o o-o
853 * / / /
854 * o-o o-o
855 * ,---'
856 * o-o o-o
857 * / / /
858 * o-o o-o
859 */
860
861 /* Scan8 organization:
862 * 0 1 2 3 4 5 6 7
863 * 0 DY y y y y y
864 * 1 y Y Y Y Y
865 * 2 y Y Y Y Y
866 * 3 y Y Y Y Y
867 * 4 y Y Y Y Y
868 * 5 DU u u u u u
869 * 6 u U U U U
870 * 7 u U U U U
871 * 8 u U U U U
872 * 9 u U U U U
873 * 10 DV v v v v v
874 * 11 v V V V V
875 * 12 v V V V V
876 * 13 v V V V V
877 * 14 v V V V V
878 * DY/DU/DV are for luma/chroma DC.
879 */
880
881 #define LUMA_DC_BLOCK_INDEX 48
882 #define CHROMA_DC_BLOCK_INDEX 49
883
884 // This table must be here because scan8[constant] must be known at compiletime
885 static const uint8_t scan8[16 * 3 + 3] = {
886 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
887 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
888 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
889 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
890 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
891 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
892 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
893 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
894 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
895 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
896 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
897 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
898 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
899 };
900
901 static av_always_inline uint32_t pack16to32(int a, int b)
902 {
903 #if HAVE_BIGENDIAN
904 return (b & 0xFFFF) + (a << 16);
905 #else
906 return (a & 0xFFFF) + (b << 16);
907 #endif
908 }
909
910 static av_always_inline uint16_t pack8to16(int a, int b)
911 {
912 #if HAVE_BIGENDIAN
913 return (b & 0xFF) + (a << 8);
914 #else
915 return (a & 0xFF) + (b << 8);
916 #endif
917 }
918
919 /**
920 * Get the chroma qp.
921 */
922 static av_always_inline int get_chroma_qp(const H264Context *h, int t, int qscale)
923 {
924 return h->pps.chroma_qp_table[t][qscale];
925 }
926
927 /**
928 * Get the predicted intra4x4 prediction mode.
929 */
930 static av_always_inline int pred_intra_mode(const H264Context *h,
931 H264SliceContext *sl, int n)
932 {
933 const int index8 = scan8[n];
934 const int left = sl->intra4x4_pred_mode_cache[index8 - 1];
935 const int top = sl->intra4x4_pred_mode_cache[index8 - 8];
936 const int min = FFMIN(left, top);
937
938 ff_tlog(h->avctx, "mode:%d %d min:%d\n", left, top, min);
939
940 if (min < 0)
941 return DC_PRED;
942 else
943 return min;
944 }
945
946 static av_always_inline void write_back_intra_pred_mode(const H264Context *h,
947 H264SliceContext *sl)
948 {
949 int8_t *i4x4 = sl->intra4x4_pred_mode + h->mb2br_xy[sl->mb_xy];
950 int8_t *i4x4_cache = sl->intra4x4_pred_mode_cache;
951
952 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
953 i4x4[4] = i4x4_cache[7 + 8 * 3];
954 i4x4[5] = i4x4_cache[7 + 8 * 2];
955 i4x4[6] = i4x4_cache[7 + 8 * 1];
956 }
957
958 static av_always_inline void write_back_non_zero_count(const H264Context *h,
959 H264SliceContext *sl)
960 {
961 const int mb_xy = sl->mb_xy;
962 uint8_t *nnz = h->non_zero_count[mb_xy];
963 uint8_t *nnz_cache = sl->non_zero_count_cache;
964
965 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
966 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
967 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
968 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
969 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
970 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
971 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
972 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
973
974 if (!h->chroma_y_shift) {
975 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
976 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
977 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
978 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
979 }
980 }
981
982 static av_always_inline void write_back_motion_list(const H264Context *h,
983 H264SliceContext *sl,
984 int b_stride,
985 int b_xy, int b8_xy,
986 int mb_type, int list)
987 {
988 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
989 int16_t(*mv_src)[2] = &sl->mv_cache[list][scan8[0]];
990 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
991 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
992 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
993 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
994 if (CABAC(h)) {
995 uint8_t (*mvd_dst)[2] = &sl->mvd_table[list][FMO ? 8 * sl->mb_xy
996 : h->mb2br_xy[sl->mb_xy]];
997 uint8_t(*mvd_src)[2] = &sl->mvd_cache[list][scan8[0]];
998 if (IS_SKIP(mb_type)) {
999 AV_ZERO128(mvd_dst);
1000 } else {
1001 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
1002 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
1003 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
1004 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
1005 }
1006 }
1007
1008 {
1009 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
1010 int8_t *ref_cache = sl->ref_cache[list];
1011 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
1012 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
1013 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
1014 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
1015 }
1016 }
1017
1018 static av_always_inline void write_back_motion(const H264Context *h,
1019 H264SliceContext *sl,
1020 int mb_type)
1021 {
1022 const int b_stride = h->b_stride;
1023 const int b_xy = 4 * sl->mb_x + 4 * sl->mb_y * h->b_stride; // try mb2b(8)_xy
1024 const int b8_xy = 4 * sl->mb_xy;
1025
1026 if (USES_LIST(mb_type, 0)) {
1027 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 0);
1028 } else {
1029 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
1030 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
1031 }
1032 if (USES_LIST(mb_type, 1))
1033 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 1);
1034
1035 if (sl->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
1036 if (IS_8X8(mb_type)) {
1037 uint8_t *direct_table = &h->direct_table[4 * sl->mb_xy];
1038 direct_table[1] = sl->sub_mb_type[1] >> 1;
1039 direct_table[2] = sl->sub_mb_type[2] >> 1;
1040 direct_table[3] = sl->sub_mb_type[3] >> 1;
1041 }
1042 }
1043 }
1044
1045 static av_always_inline int get_dct8x8_allowed(const H264Context *h, H264SliceContext *sl)
1046 {
1047 if (h->sps.direct_8x8_inference_flag)
1048 return !(AV_RN64A(sl->sub_mb_type) &
1049 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
1050 0x0001000100010001ULL));
1051 else
1052 return !(AV_RN64A(sl->sub_mb_type) &
1053 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
1054 0x0001000100010001ULL));
1055 }
1056
1057 int ff_h264_field_end(H264Context *h, H264SliceContext *sl, int in_setup);
1058
1059 int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
1060 void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
1061
1062 int ff_h264_slice_context_init(H264Context *h, H264SliceContext *sl);
1063
1064 void ff_h264_draw_horiz_band(const H264Context *h, H264SliceContext *sl, int y, int height);
1065 int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
1066 int ff_set_ref_count(H264Context *h, H264SliceContext *sl);
1067
1068 int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl);
1069 int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count);
1070 int ff_h264_update_thread_context(AVCodecContext *dst,
1071 const AVCodecContext *src);
1072
1073 void ff_h264_flush_change(H264Context *h);
1074
1075 void ff_h264_free_tables(H264Context *h);
1076
1077 #endif /* AVCODEC_H264_H */