a7b1d58e3b89e91ded26e22d8e8fe658c99f2703
[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 "h2645_parse.h"
38 #include "h264chroma.h"
39 #include "h264dsp.h"
40 #include "h264pred.h"
41 #include "h264qpel.h"
42 #include "internal.h"
43 #include "mpegutils.h"
44 #include "parser.h"
45 #include "qpeldsp.h"
46 #include "rectangle.h"
47 #include "videodsp.h"
48
49 #define H264_MAX_PICTURE_COUNT 32
50 #define H264_MAX_THREADS 16
51
52 #define MAX_SPS_COUNT 32
53 #define MAX_PPS_COUNT 256
54
55 #define MAX_MMCO_COUNT 66
56
57 #define MAX_DELAYED_PIC_COUNT 16
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 32
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 CHROMA422(h) (h->sps.chroma_format_idc == 2)
99 #define CHROMA444(h) (h->sps.chroma_format_idc == 3)
100
101 #define EXTENDED_SAR 255
102
103 #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
104 #define MB_TYPE_8x8DCT 0x01000000
105 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
106 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
107
108 #define QP_MAX_NUM (51 + 2 * 6) // The maximum supported qp
109
110 /* NAL unit types */
111 enum {
112 NAL_SLICE = 1,
113 NAL_DPA = 2,
114 NAL_DPB = 3,
115 NAL_DPC = 4,
116 NAL_IDR_SLICE = 5,
117 NAL_SEI = 6,
118 NAL_SPS = 7,
119 NAL_PPS = 8,
120 NAL_AUD = 9,
121 NAL_END_SEQUENCE = 10,
122 NAL_END_STREAM = 11,
123 NAL_FILLER_DATA = 12,
124 NAL_SPS_EXT = 13,
125 NAL_AUXILIARY_SLICE = 19,
126 NAL_FF_IGNORE = 0xff0f001,
127 };
128
129 /**
130 * SEI message types
131 */
132 typedef enum {
133 SEI_TYPE_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
134 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
135 SEI_TYPE_USER_DATA_REGISTERED = 4, ///< registered user data as specified by Rec. ITU-T T.35
136 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
137 SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
138 SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
139 SEI_TYPE_DISPLAY_ORIENTATION = 47, ///< display orientation
140 } SEI_Type;
141
142 /**
143 * pic_struct in picture timing SEI message
144 */
145 typedef enum {
146 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
147 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
148 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
149 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
150 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
151 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
152 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
153 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
154 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
155 } SEI_PicStructType;
156
157 /**
158 * Sequence parameter set
159 */
160 typedef struct SPS {
161 unsigned int sps_id;
162 int profile_idc;
163 int level_idc;
164 int chroma_format_idc;
165 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
166 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
167 int poc_type; ///< pic_order_cnt_type
168 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
169 int delta_pic_order_always_zero_flag;
170 int offset_for_non_ref_pic;
171 int offset_for_top_to_bottom_field;
172 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
173 int ref_frame_count; ///< num_ref_frames
174 int gaps_in_frame_num_allowed_flag;
175 int mb_width; ///< pic_width_in_mbs_minus1 + 1
176 int mb_height; ///< pic_height_in_map_units_minus1 + 1
177 int frame_mbs_only_flag;
178 int mb_aff; ///< mb_adaptive_frame_field_flag
179 int direct_8x8_inference_flag;
180 int crop; ///< frame_cropping_flag
181
182 /* those 4 are already in luma samples */
183 unsigned int crop_left; ///< frame_cropping_rect_left_offset
184 unsigned int crop_right; ///< frame_cropping_rect_right_offset
185 unsigned int crop_top; ///< frame_cropping_rect_top_offset
186 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
187 int vui_parameters_present_flag;
188 AVRational sar;
189 int video_signal_type_present_flag;
190 int full_range;
191 int colour_description_present_flag;
192 enum AVColorPrimaries color_primaries;
193 enum AVColorTransferCharacteristic color_trc;
194 enum AVColorSpace colorspace;
195 int timing_info_present_flag;
196 uint32_t num_units_in_tick;
197 uint32_t time_scale;
198 int fixed_frame_rate_flag;
199 short offset_for_ref_frame[256]; // FIXME dyn aloc?
200 int bitstream_restriction_flag;
201 int num_reorder_frames;
202 int scaling_matrix_present;
203 uint8_t scaling_matrix4[6][16];
204 uint8_t scaling_matrix8[6][64];
205 int nal_hrd_parameters_present_flag;
206 int vcl_hrd_parameters_present_flag;
207 int pic_struct_present_flag;
208 int time_offset_length;
209 int cpb_cnt; ///< See H.264 E.1.2
210 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
211 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
212 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
213 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
214 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
215 int residual_color_transform_flag; ///< residual_colour_transform_flag
216 int constraint_set_flags; ///< constraint_set[0-3]_flag
217 int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
218 } SPS;
219
220 /**
221 * Picture parameter set
222 */
223 typedef struct PPS {
224 unsigned int sps_id;
225 int cabac; ///< entropy_coding_mode_flag
226 int pic_order_present; ///< pic_order_present_flag
227 int slice_group_count; ///< num_slice_groups_minus1 + 1
228 int mb_slice_group_map_type;
229 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
230 int weighted_pred; ///< weighted_pred_flag
231 int weighted_bipred_idc;
232 int init_qp; ///< pic_init_qp_minus26 + 26
233 int init_qs; ///< pic_init_qs_minus26 + 26
234 int chroma_qp_index_offset[2];
235 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
236 int constrained_intra_pred; ///< constrained_intra_pred_flag
237 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
238 int transform_8x8_mode; ///< transform_8x8_mode_flag
239 uint8_t scaling_matrix4[6][16];
240 uint8_t scaling_matrix8[6][64];
241 uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
242 int chroma_qp_diff;
243 } PPS;
244
245 /**
246 * Memory management control operation opcode.
247 */
248 typedef enum MMCOOpcode {
249 MMCO_END = 0,
250 MMCO_SHORT2UNUSED,
251 MMCO_LONG2UNUSED,
252 MMCO_SHORT2LONG,
253 MMCO_SET_MAX_LONG,
254 MMCO_RESET,
255 MMCO_LONG,
256 } MMCOOpcode;
257
258 /**
259 * Memory management control operation.
260 */
261 typedef struct MMCO {
262 MMCOOpcode opcode;
263 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
264 int long_arg; ///< index, pic_num, or num long refs depending on opcode
265 } MMCO;
266
267 typedef struct H264Picture {
268 AVFrame *f;
269 ThreadFrame tf;
270
271 AVBufferRef *qscale_table_buf;
272 int8_t *qscale_table;
273
274 AVBufferRef *motion_val_buf[2];
275 int16_t (*motion_val[2])[2];
276
277 AVBufferRef *mb_type_buf;
278 uint32_t *mb_type;
279
280 AVBufferRef *hwaccel_priv_buf;
281 void *hwaccel_picture_private; ///< hardware accelerator private data
282
283 AVBufferRef *ref_index_buf[2];
284 int8_t *ref_index[2];
285
286 int field_poc[2]; ///< top/bottom POC
287 int poc; ///< frame POC
288 int frame_num; ///< frame_num (raw frame_num from slice header)
289 int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
290 not mix pictures before and after MMCO_RESET. */
291 int pic_id; /**< pic_num (short -> no wrap version of pic_num,
292 pic_num & max_pic_num; long -> long_pic_num) */
293 int long_ref; ///< 1->long term reference 0->short term reference
294 int ref_poc[2][2][32]; ///< POCs of the frames used as reference (FIXME need per slice)
295 int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
296 int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
297 int field_picture; ///< whether or not picture was encoded in separate fields
298
299 int reference;
300 int recovered; ///< picture at IDR or recovery point + recovery count
301 } H264Picture;
302
303 typedef struct H264Ref {
304 uint8_t *data[3];
305 int linesize[3];
306
307 int reference;
308 int poc;
309 int pic_id;
310
311 H264Picture *parent;
312 } H264Ref;
313
314 typedef struct H264SliceContext {
315 struct H264Context *h264;
316 GetBitContext gb;
317 ERContext er;
318
319 int slice_num;
320 int slice_type;
321 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
322 int slice_type_fixed;
323
324 int qscale;
325 int chroma_qp[2]; // QPc
326 int qp_thresh; ///< QP threshold to skip loopfilter
327 int last_qscale_diff;
328
329 // deblock
330 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
331 int slice_alpha_c0_offset;
332 int slice_beta_offset;
333
334 H264PredWeightTable pwt;
335
336 int prev_mb_skipped;
337 int next_mb_skipped;
338
339 int chroma_pred_mode;
340 int intra16x16_pred_mode;
341
342 int8_t intra4x4_pred_mode_cache[5 * 8];
343 int8_t(*intra4x4_pred_mode);
344
345 int topleft_mb_xy;
346 int top_mb_xy;
347 int topright_mb_xy;
348 int left_mb_xy[LEFT_MBS];
349
350 int topleft_type;
351 int top_type;
352 int topright_type;
353 int left_type[LEFT_MBS];
354
355 const uint8_t *left_block;
356 int topleft_partition;
357
358 unsigned int topleft_samples_available;
359 unsigned int top_samples_available;
360 unsigned int topright_samples_available;
361 unsigned int left_samples_available;
362
363 ptrdiff_t linesize, uvlinesize;
364 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
365 ptrdiff_t mb_uvlinesize;
366
367 int mb_x, mb_y;
368 int mb_xy;
369 int resync_mb_x;
370 int resync_mb_y;
371 // index of the first MB of the next slice
372 int next_slice_idx;
373 int mb_skip_run;
374 int is_complex;
375
376 int mb_field_decoding_flag;
377 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
378
379 int redundant_pic_count;
380
381 /**
382 * number of neighbors (top and/or left) that used 8x8 dct
383 */
384 int neighbor_transform_size;
385
386 int direct_spatial_mv_pred;
387 int col_parity;
388 int col_fieldoff;
389
390 int cbp;
391 int top_cbp;
392 int left_cbp;
393
394 int dist_scale_factor[32];
395 int dist_scale_factor_field[2][32];
396 int map_col_to_list0[2][16 + 32];
397 int map_col_to_list0_field[2][2][16 + 32];
398
399 /**
400 * num_ref_idx_l0/1_active_minus1 + 1
401 */
402 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
403 unsigned int list_count;
404 H264Ref ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
405 * Reordered version of default_ref_list
406 * according to picture reordering in slice header */
407 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
408
409 const uint8_t *intra_pcm_ptr;
410 int16_t *dc_val_base;
411
412 uint8_t *bipred_scratchpad;
413 uint8_t *edge_emu_buffer;
414 uint8_t (*top_borders[2])[(16 * 3) * 2];
415 int bipred_scratchpad_allocated;
416 int edge_emu_buffer_allocated;
417 int top_borders_allocated[2];
418
419 /**
420 * non zero coeff count cache.
421 * is 64 if not available.
422 */
423 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
424
425 /**
426 * Motion vector cache.
427 */
428 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
429 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
430 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
431 uint8_t direct_cache[5 * 8];
432
433 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
434
435 ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
436 DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2];
437 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
438 ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either
439 ///< check that i is not too large or ensure that there is some unused stuff after mb
440 int16_t mb_padding[256 * 2];
441
442 uint8_t (*mvd_table[2])[2];
443
444 /**
445 * Cabac
446 */
447 CABACContext cabac;
448 uint8_t cabac_state[1024];
449 int cabac_init_idc;
450
451 // rbsp buffer used for this slice
452 uint8_t *rbsp_buffer;
453 unsigned int rbsp_buffer_size;
454 } H264SliceContext;
455
456 /**
457 * H264Context
458 */
459 typedef struct H264Context {
460 const AVClass *class;
461 AVCodecContext *avctx;
462 VideoDSPContext vdsp;
463 H264DSPContext h264dsp;
464 H264ChromaContext h264chroma;
465 H264QpelContext h264qpel;
466 GetBitContext gb;
467
468 H264Picture DPB[H264_MAX_PICTURE_COUNT];
469 H264Picture *cur_pic_ptr;
470 H264Picture cur_pic;
471
472 H264SliceContext *slice_ctx;
473 int nb_slice_ctx;
474
475 H2645Packet pkt;
476
477 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
478
479 /* coded dimensions -- 16 * mb w/h */
480 int width, height;
481 int chroma_x_shift, chroma_y_shift;
482
483 int droppable;
484 int coded_picture_number;
485 int low_delay;
486
487 int context_initialized;
488 int flags;
489 int workaround_bugs;
490
491 int8_t(*intra4x4_pred_mode);
492 H264PredContext hpc;
493
494 uint8_t (*non_zero_count)[48];
495
496 #define LIST_NOT_USED -1 // FIXME rename?
497 #define PART_NOT_AVAILABLE -2
498
499 /**
500 * block_offset[ 0..23] for frame macroblocks
501 * block_offset[24..47] for field macroblocks
502 */
503 int block_offset[2 * (16 * 3)];
504
505 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
506 uint32_t *mb2br_xy;
507 int b_stride; // FIXME use s->b4_stride
508
509 SPS sps; ///< current sps
510 PPS pps; ///< current pps
511
512 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
513 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
514 uint32_t(*dequant4_coeff[6])[16];
515 uint32_t(*dequant8_coeff[6])[64];
516
517 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
518
519 // interlacing specific flags
520 int mb_aff_frame;
521 int picture_structure;
522 int first_field;
523
524 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
525
526 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
527 uint16_t *cbp_table;
528
529 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
530 uint8_t *chroma_pred_mode_table;
531 uint8_t (*mvd_table[2])[2];
532 uint8_t *direct_table;
533
534 uint8_t zigzag_scan[16];
535 uint8_t zigzag_scan8x8[64];
536 uint8_t zigzag_scan8x8_cavlc[64];
537 uint8_t field_scan[16];
538 uint8_t field_scan8x8[64];
539 uint8_t field_scan8x8_cavlc[64];
540 const uint8_t *zigzag_scan_q0;
541 const uint8_t *zigzag_scan8x8_q0;
542 const uint8_t *zigzag_scan8x8_cavlc_q0;
543 const uint8_t *field_scan_q0;
544 const uint8_t *field_scan8x8_q0;
545 const uint8_t *field_scan8x8_cavlc_q0;
546
547 int x264_build;
548
549 int mb_y;
550 int mb_height, mb_width;
551 int mb_stride;
552 int mb_num;
553
554 // =============================================================
555 // Things below are not used in the MB or more inner code
556
557 int nal_ref_idc;
558 int nal_unit_type;
559
560 /**
561 * Used to parse AVC variant of h264
562 */
563 int is_avc; ///< this flag is != 0 if codec is avc1
564 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
565
566 int bit_depth_luma; ///< luma bit depth from sps to detect changes
567 int chroma_format_idc; ///< chroma format from sps to detect changes
568
569 SPS *sps_buffers[MAX_SPS_COUNT];
570 PPS *pps_buffers[MAX_PPS_COUNT];
571
572 int dequant_coeff_pps; ///< reinit tables when pps changes
573
574 uint16_t *slice_table_base;
575
576 // POC stuff
577 int poc_lsb;
578 int poc_msb;
579 int delta_poc_bottom;
580 int delta_poc[2];
581 int frame_num;
582 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
583 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
584 int frame_num_offset; ///< for POC type 2
585 int prev_frame_num_offset; ///< for POC type 2
586 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
587
588 /**
589 * frame_num for frames or 2 * frame_num + 1 for field pics.
590 */
591 int curr_pic_num;
592
593 /**
594 * max_frame_num or 2 * max_frame_num for field pics.
595 */
596 int max_pic_num;
597
598 H264Picture *short_ref[32];
599 H264Picture *long_ref[32];
600 H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
601 int last_pocs[MAX_DELAYED_PIC_COUNT];
602 H264Picture *next_output_pic;
603 int next_outputed_poc;
604
605 /**
606 * memory management control operations buffer.
607 */
608 MMCO mmco[MAX_MMCO_COUNT];
609 int mmco_index;
610 int mmco_reset;
611
612 int long_ref_count; ///< number of actual long term references
613 int short_ref_count; ///< number of actual short term references
614
615 /**
616 * @name Members for slice based multithreading
617 * @{
618 */
619 /**
620 * current slice number, used to initalize slice_num of each thread/context
621 */
622 int current_slice;
623
624 /**
625 * Max number of threads / contexts.
626 * This is equal to AVCodecContext.thread_count unless
627 * multithreaded decoding is impossible, in which case it is
628 * reduced to 1.
629 */
630 int max_contexts;
631
632 int slice_context_count;
633
634 /**
635 * 1 if the single thread fallback warning has already been
636 * displayed, 0 otherwise.
637 */
638 int single_decode_warning;
639
640 enum AVPictureType pict_type;
641
642 /** @} */
643
644 /**
645 * pic_struct in picture timing SEI message
646 */
647 SEI_PicStructType sei_pic_struct;
648
649 /**
650 * Complement sei_pic_struct
651 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
652 * However, soft telecined frames may have these values.
653 * This is used in an attempt to flag soft telecine progressive.
654 */
655 int prev_interlaced_frame;
656
657 /**
658 * frame_packing_arrangment SEI message
659 */
660 int sei_frame_packing_present;
661 int frame_packing_arrangement_type;
662 int content_interpretation_type;
663 int quincunx_subsampling;
664
665 /**
666 * display orientation SEI message
667 */
668 int sei_display_orientation_present;
669 int sei_anticlockwise_rotation;
670 int sei_hflip, sei_vflip;
671
672 /**
673 * User data registered by Rec. ITU-T T.35 SEI
674 */
675 int sei_reguserdata_afd_present;
676 uint8_t active_format_description;
677 int a53_caption_size;
678 uint8_t *a53_caption;
679
680 /**
681 * Bit set of clock types for fields/frames in picture timing SEI message.
682 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
683 * interlaced).
684 */
685 int sei_ct_type;
686
687 /**
688 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
689 */
690 int sei_dpb_output_delay;
691
692 /**
693 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
694 */
695 int sei_cpb_removal_delay;
696
697 /**
698 * recovery_frame_cnt from SEI message
699 *
700 * Set to -1 if no recovery point SEI message found or to number of frames
701 * before playback synchronizes. Frames having recovery point are key
702 * frames.
703 */
704 int sei_recovery_frame_cnt;
705
706 /**
707 * recovery_frame is the frame_num at which the next frame should
708 * be fully constructed.
709 *
710 * Set to -1 when not expecting a recovery point.
711 */
712 int recovery_frame;
713
714 /**
715 * We have seen an IDR, so all the following frames in coded order are correctly
716 * decodable.
717 */
718 #define FRAME_RECOVERED_IDR (1 << 0)
719 /**
720 * Sufficient number of frames have been decoded since a SEI recovery point,
721 * so all the following frames in presentation order are correct.
722 */
723 #define FRAME_RECOVERED_SEI (1 << 1)
724
725 int frame_recovered; ///< Initial frame has been completely recovered
726
727 /* for frame threading, this is set to 1
728 * after finish_setup() has been called, so we cannot modify
729 * some context properties (which are supposed to stay constant between
730 * slices) anymore */
731 int setup_finished;
732
733 // Timestamp stuff
734 int sei_buffering_period_present; ///< Buffering period SEI flag
735 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
736
737 int enable_er;
738
739 AVBufferPool *qscale_table_pool;
740 AVBufferPool *mb_type_pool;
741 AVBufferPool *motion_val_pool;
742 AVBufferPool *ref_index_pool;
743
744 /* Motion Estimation */
745 qpel_mc_func (*qpel_put)[16];
746 qpel_mc_func (*qpel_avg)[16];
747 } H264Context;
748
749 extern const uint16_t ff_h264_mb_sizes[4];
750
751 /**
752 * Decode SEI
753 */
754 int ff_h264_decode_sei(H264Context *h);
755
756 /**
757 * Decode SPS
758 */
759 int ff_h264_decode_seq_parameter_set(H264Context *h);
760
761 /**
762 * compute profile from sps
763 */
764 int ff_h264_get_profile(SPS *sps);
765
766 /**
767 * Decode PPS
768 */
769 int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
770
771 /**
772 * Decode a network abstraction layer unit.
773 * @param consumed is the number of bytes used as input
774 * @param length is the length of the array
775 * @param dst_length is the number of decoded bytes FIXME here
776 * or a decode rbsp tailing?
777 * @return decoded bytes, might be src+1 if no escapes
778 */
779 const uint8_t *ff_h264_decode_nal(H264Context *h, H264SliceContext *sl, const uint8_t *src,
780 int *dst_length, int *consumed, int length);
781
782 /**
783 * Free any data that may have been allocated in the H264 context
784 * like SPS, PPS etc.
785 */
786 void ff_h264_free_context(H264Context *h);
787
788 /**
789 * Reconstruct bitstream slice_type.
790 */
791 int ff_h264_get_slice_type(const H264SliceContext *sl);
792
793 /**
794 * Allocate tables.
795 * needs width/height
796 */
797 int ff_h264_alloc_tables(H264Context *h);
798
799 int ff_h264_decode_ref_pic_list_reordering(H264Context *h, H264SliceContext *sl);
800 void ff_h264_fill_mbaff_ref_list(H264Context *h, H264SliceContext *sl);
801 void ff_h264_remove_all_refs(H264Context *h);
802
803 /**
804 * Execute the reference picture marking (memory management control operations).
805 */
806 int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
807
808 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
809 int first_slice);
810
811 int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
812
813 void ff_h264_hl_decode_mb(const H264Context *h, H264SliceContext *sl);
814 int ff_h264_decode_extradata(H264Context *h);
815 int ff_h264_decode_init(AVCodecContext *avctx);
816 void ff_h264_decode_init_vlc(void);
817
818 /**
819 * Decode a macroblock
820 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
821 */
822 int ff_h264_decode_mb_cavlc(const H264Context *h, H264SliceContext *sl);
823
824 /**
825 * Decode a CABAC coded macroblock
826 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
827 */
828 int ff_h264_decode_mb_cabac(const H264Context *h, H264SliceContext *sl);
829
830 void ff_h264_init_cabac_states(const H264Context *h, H264SliceContext *sl);
831
832 void h264_init_dequant_tables(H264Context *h);
833
834 void ff_h264_direct_dist_scale_factor(const H264Context *const h, H264SliceContext *sl);
835 void ff_h264_direct_ref_list_init(const H264Context *const h, H264SliceContext *sl);
836 void ff_h264_pred_direct_motion(const H264Context *const h, H264SliceContext *sl,
837 int *mb_type);
838
839 void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
840 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
841 unsigned int linesize, unsigned int uvlinesize);
842 void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
843 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
844 unsigned int linesize, unsigned int uvlinesize);
845
846 /**
847 * Reset SEI values at the beginning of the frame.
848 *
849 * @param h H.264 context.
850 */
851 void ff_h264_reset_sei(H264Context *h);
852
853 /*
854 * o-o o-o
855 * / / /
856 * o-o o-o
857 * ,---'
858 * o-o o-o
859 * / / /
860 * o-o o-o
861 */
862
863 /* Scan8 organization:
864 * 0 1 2 3 4 5 6 7
865 * 0 DY y y y y y
866 * 1 y Y Y Y Y
867 * 2 y Y Y Y Y
868 * 3 y Y Y Y Y
869 * 4 y Y Y Y Y
870 * 5 DU u u u u u
871 * 6 u U U U U
872 * 7 u U U U U
873 * 8 u U U U U
874 * 9 u U U U U
875 * 10 DV v v v v v
876 * 11 v V V V V
877 * 12 v V V V V
878 * 13 v V V V V
879 * 14 v V V V V
880 * DY/DU/DV are for luma/chroma DC.
881 */
882
883 #define LUMA_DC_BLOCK_INDEX 48
884 #define CHROMA_DC_BLOCK_INDEX 49
885
886 // This table must be here because scan8[constant] must be known at compiletime
887 static const uint8_t scan8[16 * 3 + 3] = {
888 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
889 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
890 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
891 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
892 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
893 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
894 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
895 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
896 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
897 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
898 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
899 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
900 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
901 };
902
903 static av_always_inline uint32_t pack16to32(int a, int b)
904 {
905 #if HAVE_BIGENDIAN
906 return (b & 0xFFFF) + (a << 16);
907 #else
908 return (a & 0xFFFF) + (b << 16);
909 #endif
910 }
911
912 static av_always_inline uint16_t pack8to16(int a, int b)
913 {
914 #if HAVE_BIGENDIAN
915 return (b & 0xFF) + (a << 8);
916 #else
917 return (a & 0xFF) + (b << 8);
918 #endif
919 }
920
921 /**
922 * Get the chroma qp.
923 */
924 static av_always_inline int get_chroma_qp(const H264Context *h, int t, int qscale)
925 {
926 return h->pps.chroma_qp_table[t][qscale];
927 }
928
929 /**
930 * Get the predicted intra4x4 prediction mode.
931 */
932 static av_always_inline int pred_intra_mode(const H264Context *h,
933 H264SliceContext *sl, int n)
934 {
935 const int index8 = scan8[n];
936 const int left = sl->intra4x4_pred_mode_cache[index8 - 1];
937 const int top = sl->intra4x4_pred_mode_cache[index8 - 8];
938 const int min = FFMIN(left, top);
939
940 ff_tlog(h->avctx, "mode:%d %d min:%d\n", left, top, min);
941
942 if (min < 0)
943 return DC_PRED;
944 else
945 return min;
946 }
947
948 static av_always_inline void write_back_intra_pred_mode(const H264Context *h,
949 H264SliceContext *sl)
950 {
951 int8_t *i4x4 = sl->intra4x4_pred_mode + h->mb2br_xy[sl->mb_xy];
952 int8_t *i4x4_cache = sl->intra4x4_pred_mode_cache;
953
954 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
955 i4x4[4] = i4x4_cache[7 + 8 * 3];
956 i4x4[5] = i4x4_cache[7 + 8 * 2];
957 i4x4[6] = i4x4_cache[7 + 8 * 1];
958 }
959
960 static av_always_inline void write_back_non_zero_count(const H264Context *h,
961 H264SliceContext *sl)
962 {
963 const int mb_xy = sl->mb_xy;
964 uint8_t *nnz = h->non_zero_count[mb_xy];
965 uint8_t *nnz_cache = sl->non_zero_count_cache;
966
967 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
968 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
969 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
970 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
971 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
972 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
973 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
974 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
975
976 if (!h->chroma_y_shift) {
977 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
978 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
979 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
980 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
981 }
982 }
983
984 static av_always_inline void write_back_motion_list(const H264Context *h,
985 H264SliceContext *sl,
986 int b_stride,
987 int b_xy, int b8_xy,
988 int mb_type, int list)
989 {
990 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
991 int16_t(*mv_src)[2] = &sl->mv_cache[list][scan8[0]];
992 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
993 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
994 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
995 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
996 if (CABAC(h)) {
997 uint8_t (*mvd_dst)[2] = &sl->mvd_table[list][FMO ? 8 * sl->mb_xy
998 : h->mb2br_xy[sl->mb_xy]];
999 uint8_t(*mvd_src)[2] = &sl->mvd_cache[list][scan8[0]];
1000 if (IS_SKIP(mb_type)) {
1001 AV_ZERO128(mvd_dst);
1002 } else {
1003 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
1004 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
1005 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
1006 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
1007 }
1008 }
1009
1010 {
1011 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
1012 int8_t *ref_cache = sl->ref_cache[list];
1013 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
1014 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
1015 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
1016 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
1017 }
1018 }
1019
1020 static av_always_inline void write_back_motion(const H264Context *h,
1021 H264SliceContext *sl,
1022 int mb_type)
1023 {
1024 const int b_stride = h->b_stride;
1025 const int b_xy = 4 * sl->mb_x + 4 * sl->mb_y * h->b_stride; // try mb2b(8)_xy
1026 const int b8_xy = 4 * sl->mb_xy;
1027
1028 if (USES_LIST(mb_type, 0)) {
1029 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 0);
1030 } else {
1031 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
1032 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
1033 }
1034 if (USES_LIST(mb_type, 1))
1035 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 1);
1036
1037 if (sl->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
1038 if (IS_8X8(mb_type)) {
1039 uint8_t *direct_table = &h->direct_table[4 * sl->mb_xy];
1040 direct_table[1] = sl->sub_mb_type[1] >> 1;
1041 direct_table[2] = sl->sub_mb_type[2] >> 1;
1042 direct_table[3] = sl->sub_mb_type[3] >> 1;
1043 }
1044 }
1045 }
1046
1047 static av_always_inline int get_dct8x8_allowed(const H264Context *h, H264SliceContext *sl)
1048 {
1049 if (h->sps.direct_8x8_inference_flag)
1050 return !(AV_RN64A(sl->sub_mb_type) &
1051 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
1052 0x0001000100010001ULL));
1053 else
1054 return !(AV_RN64A(sl->sub_mb_type) &
1055 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
1056 0x0001000100010001ULL));
1057 }
1058
1059 int ff_h264_field_end(H264Context *h, H264SliceContext *sl, int in_setup);
1060
1061 int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
1062 void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
1063
1064 int ff_h264_slice_context_init(H264Context *h, H264SliceContext *sl);
1065
1066 void ff_h264_draw_horiz_band(const H264Context *h, H264SliceContext *sl, int y, int height);
1067 int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
1068 int ff_set_ref_count(H264Context *h, H264SliceContext *sl);
1069
1070 int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl);
1071 int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count);
1072 int ff_h264_update_thread_context(AVCodecContext *dst,
1073 const AVCodecContext *src);
1074
1075 void ff_h264_flush_change(H264Context *h);
1076
1077 void ff_h264_free_tables(H264Context *h);
1078
1079 #endif /* AVCODEC_H264_H */