h264: Split h264 slice decoding from nal decoding
[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 "cabac.h"
33 #include "dsputil.h"
34 #include "error_resilience.h"
35 #include "get_bits.h"
36 #include "h264chroma.h"
37 #include "h264dsp.h"
38 #include "h264pred.h"
39 #include "h264qpel.h"
40 #include "mpegutils.h"
41 #include "parser.h"
42 #include "rectangle.h"
43 #include "videodsp.h"
44
45 #define H264_MAX_PICTURE_COUNT 32
46 #define H264_MAX_THREADS 16
47
48 #define MAX_SPS_COUNT 32
49 #define MAX_PPS_COUNT 256
50
51 #define MAX_MMCO_COUNT 66
52
53 #define MAX_DELAYED_PIC_COUNT 16
54
55 /* Compiling in interlaced support reduces the speed
56 * of progressive decoding by about 2%. */
57 #define ALLOW_INTERLACE
58
59 #define FMO 0
60
61 /**
62 * The maximum number of slices supported by the decoder.
63 * must be a power of 2
64 */
65 #define MAX_SLICES 16
66
67 #ifdef ALLOW_INTERLACE
68 #define MB_MBAFF(h) h->mb_mbaff
69 #define MB_FIELD(h) h->mb_field_decoding_flag
70 #define FRAME_MBAFF(h) h->mb_aff_frame
71 #define FIELD_PICTURE(h) (h->picture_structure != PICT_FRAME)
72 #define LEFT_MBS 2
73 #define LTOP 0
74 #define LBOT 1
75 #define LEFT(i) (i)
76 #else
77 #define MB_MBAFF(h) 0
78 #define MB_FIELD(h) 0
79 #define FRAME_MBAFF(h) 0
80 #define FIELD_PICTURE(h) 0
81 #undef IS_INTERLACED
82 #define IS_INTERLACED(mb_type) 0
83 #define LEFT_MBS 1
84 #define LTOP 0
85 #define LBOT 0
86 #define LEFT(i) 0
87 #endif
88 #define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
89
90 #ifndef CABAC
91 #define CABAC(h) h->pps.cabac
92 #endif
93
94 #define CHROMA422(h) (h->sps.chroma_format_idc == 2)
95 #define CHROMA444(h) (h->sps.chroma_format_idc == 3)
96
97 #define EXTENDED_SAR 255
98
99 #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
100 #define MB_TYPE_8x8DCT 0x01000000
101 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
102 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
103
104 #define QP_MAX_NUM (51 + 2 * 6) // The maximum supported qp
105
106 /* NAL unit types */
107 enum {
108 NAL_SLICE = 1,
109 NAL_DPA = 2,
110 NAL_DPB = 3,
111 NAL_DPC = 4,
112 NAL_IDR_SLICE = 5,
113 NAL_SEI = 6,
114 NAL_SPS = 7,
115 NAL_PPS = 8,
116 NAL_AUD = 9,
117 NAL_END_SEQUENCE = 10,
118 NAL_END_STREAM = 11,
119 NAL_FILLER_DATA = 12,
120 NAL_SPS_EXT = 13,
121 NAL_AUXILIARY_SLICE = 19,
122 NAL_FF_IGNORE = 0xff0f001,
123 };
124
125 /**
126 * SEI message types
127 */
128 typedef enum {
129 SEI_TYPE_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
130 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
131 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
132 SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
133 SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
134 } SEI_Type;
135
136 /**
137 * pic_struct in picture timing SEI message
138 */
139 typedef enum {
140 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
141 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
142 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
143 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
144 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
145 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
146 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
147 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
148 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
149 } SEI_PicStructType;
150
151 /**
152 * Sequence parameter set
153 */
154 typedef struct SPS {
155 unsigned int sps_id;
156 int profile_idc;
157 int level_idc;
158 int chroma_format_idc;
159 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
160 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
161 int poc_type; ///< pic_order_cnt_type
162 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
163 int delta_pic_order_always_zero_flag;
164 int offset_for_non_ref_pic;
165 int offset_for_top_to_bottom_field;
166 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
167 int ref_frame_count; ///< num_ref_frames
168 int gaps_in_frame_num_allowed_flag;
169 int mb_width; ///< pic_width_in_mbs_minus1 + 1
170 int mb_height; ///< pic_height_in_map_units_minus1 + 1
171 int frame_mbs_only_flag;
172 int mb_aff; ///< mb_adaptive_frame_field_flag
173 int direct_8x8_inference_flag;
174 int crop; ///< frame_cropping_flag
175
176 /* those 4 are already in luma samples */
177 unsigned int crop_left; ///< frame_cropping_rect_left_offset
178 unsigned int crop_right; ///< frame_cropping_rect_right_offset
179 unsigned int crop_top; ///< frame_cropping_rect_top_offset
180 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
181 int vui_parameters_present_flag;
182 AVRational sar;
183 int video_signal_type_present_flag;
184 int full_range;
185 int colour_description_present_flag;
186 enum AVColorPrimaries color_primaries;
187 enum AVColorTransferCharacteristic color_trc;
188 enum AVColorSpace colorspace;
189 int timing_info_present_flag;
190 uint32_t num_units_in_tick;
191 uint32_t time_scale;
192 int fixed_frame_rate_flag;
193 short offset_for_ref_frame[256]; // FIXME dyn aloc?
194 int bitstream_restriction_flag;
195 int num_reorder_frames;
196 int scaling_matrix_present;
197 uint8_t scaling_matrix4[6][16];
198 uint8_t scaling_matrix8[6][64];
199 int nal_hrd_parameters_present_flag;
200 int vcl_hrd_parameters_present_flag;
201 int pic_struct_present_flag;
202 int time_offset_length;
203 int cpb_cnt; ///< See H.264 E.1.2
204 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
205 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
206 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
207 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
208 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
209 int residual_color_transform_flag; ///< residual_colour_transform_flag
210 int constraint_set_flags; ///< constraint_set[0-3]_flag
211 int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
212 } SPS;
213
214 /**
215 * Picture parameter set
216 */
217 typedef struct PPS {
218 unsigned int sps_id;
219 int cabac; ///< entropy_coding_mode_flag
220 int pic_order_present; ///< pic_order_present_flag
221 int slice_group_count; ///< num_slice_groups_minus1 + 1
222 int mb_slice_group_map_type;
223 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
224 int weighted_pred; ///< weighted_pred_flag
225 int weighted_bipred_idc;
226 int init_qp; ///< pic_init_qp_minus26 + 26
227 int init_qs; ///< pic_init_qs_minus26 + 26
228 int chroma_qp_index_offset[2];
229 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
230 int constrained_intra_pred; ///< constrained_intra_pred_flag
231 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
232 int transform_8x8_mode; ///< transform_8x8_mode_flag
233 uint8_t scaling_matrix4[6][16];
234 uint8_t scaling_matrix8[6][64];
235 uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
236 int chroma_qp_diff;
237 } PPS;
238
239 /**
240 * Memory management control operation opcode.
241 */
242 typedef enum MMCOOpcode {
243 MMCO_END = 0,
244 MMCO_SHORT2UNUSED,
245 MMCO_LONG2UNUSED,
246 MMCO_SHORT2LONG,
247 MMCO_SET_MAX_LONG,
248 MMCO_RESET,
249 MMCO_LONG,
250 } MMCOOpcode;
251
252 /**
253 * Memory management control operation.
254 */
255 typedef struct MMCO {
256 MMCOOpcode opcode;
257 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
258 int long_arg; ///< index, pic_num, or num long refs depending on opcode
259 } MMCO;
260
261 typedef struct H264Picture {
262 struct AVFrame f;
263 ThreadFrame tf;
264
265 AVBufferRef *qscale_table_buf;
266 int8_t *qscale_table;
267
268 AVBufferRef *motion_val_buf[2];
269 int16_t (*motion_val[2])[2];
270
271 AVBufferRef *mb_type_buf;
272 uint32_t *mb_type;
273
274 AVBufferRef *hwaccel_priv_buf;
275 void *hwaccel_picture_private; ///< hardware accelerator private data
276
277 AVBufferRef *ref_index_buf[2];
278 int8_t *ref_index[2];
279
280 int field_poc[2]; ///< top/bottom POC
281 int poc; ///< frame POC
282 int frame_num; ///< frame_num (raw frame_num from slice header)
283 int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
284 not mix pictures before and after MMCO_RESET. */
285 int pic_id; /**< pic_num (short -> no wrap version of pic_num,
286 pic_num & max_pic_num; long -> long_pic_num) */
287 int long_ref; ///< 1->long term reference 0->short term reference
288 int ref_poc[2][2][32]; ///< POCs of the frames used as reference (FIXME need per slice)
289 int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
290 int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
291 int field_picture; ///< whether or not picture was encoded in separate fields
292
293 int needs_realloc; ///< picture needs to be reallocated (eg due to a frame size change)
294 int reference;
295 int recovered; ///< picture at IDR or recovery point + recovery count
296 } H264Picture;
297
298 /**
299 * H264Context
300 */
301 typedef struct H264Context {
302 AVCodecContext *avctx;
303 DSPContext dsp;
304 VideoDSPContext vdsp;
305 H264DSPContext h264dsp;
306 H264ChromaContext h264chroma;
307 H264QpelContext h264qpel;
308 ParseContext parse_context;
309 GetBitContext gb;
310 ERContext er;
311
312 H264Picture *DPB;
313 H264Picture *cur_pic_ptr;
314 H264Picture cur_pic;
315
316 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
317 int chroma_qp[2]; // QPc
318
319 int qp_thresh; ///< QP threshold to skip loopfilter
320
321 /* coded dimensions -- 16 * mb w/h */
322 int width, height;
323 ptrdiff_t linesize, uvlinesize;
324 int chroma_x_shift, chroma_y_shift;
325
326 int qscale;
327 int droppable;
328 int data_partitioning;
329 int coded_picture_number;
330 int low_delay;
331
332 int context_initialized;
333 int flags;
334 int workaround_bugs;
335
336 int prev_mb_skipped;
337 int next_mb_skipped;
338
339 // prediction stuff
340 int chroma_pred_mode;
341 int intra16x16_pred_mode;
342
343 int topleft_mb_xy;
344 int top_mb_xy;
345 int topright_mb_xy;
346 int left_mb_xy[LEFT_MBS];
347
348 int topleft_type;
349 int top_type;
350 int topright_type;
351 int left_type[LEFT_MBS];
352
353 const uint8_t *left_block;
354 int topleft_partition;
355
356 int8_t intra4x4_pred_mode_cache[5 * 8];
357 int8_t(*intra4x4_pred_mode);
358 H264PredContext hpc;
359 unsigned int topleft_samples_available;
360 unsigned int top_samples_available;
361 unsigned int topright_samples_available;
362 unsigned int left_samples_available;
363 uint8_t (*top_borders[2])[(16 * 3) * 2];
364
365 /**
366 * non zero coeff count cache.
367 * is 64 if not available.
368 */
369 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
370
371 uint8_t (*non_zero_count)[48];
372
373 /**
374 * Motion vector cache.
375 */
376 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
377 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
378 #define LIST_NOT_USED -1 // FIXME rename?
379 #define PART_NOT_AVAILABLE -2
380
381 /**
382 * number of neighbors (top and/or left) that used 8x8 dct
383 */
384 int neighbor_transform_size;
385
386 /**
387 * block_offset[ 0..23] for frame macroblocks
388 * block_offset[24..47] for field macroblocks
389 */
390 int block_offset[2 * (16 * 3)];
391
392 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
393 uint32_t *mb2br_xy;
394 int b_stride; // FIXME use s->b4_stride
395
396 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
397 ptrdiff_t mb_uvlinesize;
398
399 SPS sps; ///< current sps
400 PPS pps; ///< current pps
401
402 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
403 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
404 uint32_t(*dequant4_coeff[6])[16];
405 uint32_t(*dequant8_coeff[6])[64];
406
407 int slice_num;
408 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
409 int slice_type;
410 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
411 int slice_type_fixed;
412
413 // interlacing specific flags
414 int mb_aff_frame;
415 int mb_field_decoding_flag;
416 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
417 int picture_structure;
418 int first_field;
419
420 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
421
422 // Weighted pred stuff
423 int use_weight;
424 int use_weight_chroma;
425 int luma_log2_weight_denom;
426 int chroma_log2_weight_denom;
427 // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
428 int luma_weight[48][2][2];
429 int chroma_weight[48][2][2][2];
430 int implicit_weight[48][48][2];
431
432 int direct_spatial_mv_pred;
433 int col_parity;
434 int col_fieldoff;
435 int dist_scale_factor[32];
436 int dist_scale_factor_field[2][32];
437 int map_col_to_list0[2][16 + 32];
438 int map_col_to_list0_field[2][2][16 + 32];
439
440 /**
441 * num_ref_idx_l0/1_active_minus1 + 1
442 */
443 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
444 unsigned int list_count;
445 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
446 H264Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
447 * Reordered version of default_ref_list
448 * according to picture reordering in slice header */
449 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
450
451 // data partitioning
452 GetBitContext intra_gb;
453 GetBitContext inter_gb;
454 GetBitContext *intra_gb_ptr;
455 GetBitContext *inter_gb_ptr;
456
457 const uint8_t *intra_pcm_ptr;
458 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.
459 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
460 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
461
462 /**
463 * Cabac
464 */
465 CABACContext cabac;
466 uint8_t cabac_state[1024];
467
468 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
469 uint16_t *cbp_table;
470 int cbp;
471 int top_cbp;
472 int left_cbp;
473 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
474 uint8_t *chroma_pred_mode_table;
475 int last_qscale_diff;
476 uint8_t (*mvd_table[2])[2];
477 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
478 uint8_t *direct_table;
479 uint8_t direct_cache[5 * 8];
480
481 uint8_t zigzag_scan[16];
482 uint8_t zigzag_scan8x8[64];
483 uint8_t zigzag_scan8x8_cavlc[64];
484 uint8_t field_scan[16];
485 uint8_t field_scan8x8[64];
486 uint8_t field_scan8x8_cavlc[64];
487 const uint8_t *zigzag_scan_q0;
488 const uint8_t *zigzag_scan8x8_q0;
489 const uint8_t *zigzag_scan8x8_cavlc_q0;
490 const uint8_t *field_scan_q0;
491 const uint8_t *field_scan8x8_q0;
492 const uint8_t *field_scan8x8_cavlc_q0;
493
494 int x264_build;
495
496 int mb_x, mb_y;
497 int resync_mb_x;
498 int resync_mb_y;
499 int mb_skip_run;
500 int mb_height, mb_width;
501 int mb_stride;
502 int mb_num;
503 int mb_xy;
504
505 int is_complex;
506
507 // deblock
508 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
509 int slice_alpha_c0_offset;
510 int slice_beta_offset;
511
512 // =============================================================
513 // Things below are not used in the MB or more inner code
514
515 int nal_ref_idc;
516 int nal_unit_type;
517 uint8_t *rbsp_buffer[2];
518 unsigned int rbsp_buffer_size[2];
519
520 /**
521 * Used to parse AVC variant of h264
522 */
523 int is_avc; ///< this flag is != 0 if codec is avc1
524 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
525 int got_first; ///< this flag is != 0 if we've parsed a frame
526
527 int bit_depth_luma; ///< luma bit depth from sps to detect changes
528 int chroma_format_idc; ///< chroma format from sps to detect changes
529
530 SPS *sps_buffers[MAX_SPS_COUNT];
531 PPS *pps_buffers[MAX_PPS_COUNT];
532
533 int dequant_coeff_pps; ///< reinit tables when pps changes
534
535 uint16_t *slice_table_base;
536
537 // POC stuff
538 int poc_lsb;
539 int poc_msb;
540 int delta_poc_bottom;
541 int delta_poc[2];
542 int frame_num;
543 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
544 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
545 int frame_num_offset; ///< for POC type 2
546 int prev_frame_num_offset; ///< for POC type 2
547 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
548
549 /**
550 * frame_num for frames or 2 * frame_num + 1 for field pics.
551 */
552 int curr_pic_num;
553
554 /**
555 * max_frame_num or 2 * max_frame_num for field pics.
556 */
557 int max_pic_num;
558
559 int redundant_pic_count;
560
561 H264Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
562 H264Picture *short_ref[32];
563 H264Picture *long_ref[32];
564 H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
565 int last_pocs[MAX_DELAYED_PIC_COUNT];
566 H264Picture *next_output_pic;
567 int outputed_poc;
568 int next_outputed_poc;
569
570 /**
571 * memory management control operations buffer.
572 */
573 MMCO mmco[MAX_MMCO_COUNT];
574 int mmco_index;
575 int mmco_reset;
576
577 int long_ref_count; ///< number of actual long term references
578 int short_ref_count; ///< number of actual short term references
579
580 int cabac_init_idc;
581
582 /**
583 * @name Members for slice based multithreading
584 * @{
585 */
586 struct H264Context *thread_context[H264_MAX_THREADS];
587
588 /**
589 * current slice number, used to initalize slice_num of each thread/context
590 */
591 int current_slice;
592
593 /**
594 * Max number of threads / contexts.
595 * This is equal to AVCodecContext.thread_count unless
596 * multithreaded decoding is impossible, in which case it is
597 * reduced to 1.
598 */
599 int max_contexts;
600
601 int slice_context_count;
602
603 /**
604 * 1 if the single thread fallback warning has already been
605 * displayed, 0 otherwise.
606 */
607 int single_decode_warning;
608
609 enum AVPictureType pict_type;
610
611 int last_slice_type;
612 /** @} */
613
614 /**
615 * pic_struct in picture timing SEI message
616 */
617 SEI_PicStructType sei_pic_struct;
618
619 /**
620 * Complement sei_pic_struct
621 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
622 * However, soft telecined frames may have these values.
623 * This is used in an attempt to flag soft telecine progressive.
624 */
625 int prev_interlaced_frame;
626
627 /**
628 * frame_packing_arrangment SEI message
629 */
630 int sei_frame_packing_present;
631 int frame_packing_arrangement_type;
632 int content_interpretation_type;
633 int quincunx_subsampling;
634
635 /**
636 * Bit set of clock types for fields/frames in picture timing SEI message.
637 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
638 * interlaced).
639 */
640 int sei_ct_type;
641
642 /**
643 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
644 */
645 int sei_dpb_output_delay;
646
647 /**
648 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
649 */
650 int sei_cpb_removal_delay;
651
652 /**
653 * recovery_frame_cnt from SEI message
654 *
655 * Set to -1 if no recovery point SEI message found or to number of frames
656 * before playback synchronizes. Frames having recovery point are key
657 * frames.
658 */
659 int sei_recovery_frame_cnt;
660
661 /**
662 * recovery_frame is the frame_num at which the next frame should
663 * be fully constructed.
664 *
665 * Set to -1 when not expecting a recovery point.
666 */
667 int recovery_frame;
668
669 /**
670 * We have seen an IDR, so all the following frames in coded order are correctly
671 * decodable.
672 */
673 #define FRAME_RECOVERED_IDR (1 << 0)
674 /**
675 * Sufficient number of frames have been decoded since a SEI recovery point,
676 * so all the following frames in presentation order are correct.
677 */
678 #define FRAME_RECOVERED_SEI (1 << 1)
679
680 int frame_recovered; ///< Initial frame has been completely recovered
681
682 int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
683 int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
684
685 // Timestamp stuff
686 int sei_buffering_period_present; ///< Buffering period SEI flag
687 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
688
689 int cur_chroma_format_idc;
690 uint8_t *bipred_scratchpad;
691 uint8_t *edge_emu_buffer;
692 int16_t *dc_val_base;
693
694 AVBufferPool *qscale_table_pool;
695 AVBufferPool *mb_type_pool;
696 AVBufferPool *motion_val_pool;
697 AVBufferPool *ref_index_pool;
698
699 /* Motion Estimation */
700 qpel_mc_func (*qpel_put)[16];
701 qpel_mc_func (*qpel_avg)[16];
702 } H264Context;
703
704 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).
705 extern const uint16_t ff_h264_mb_sizes[4];
706
707 /**
708 * Decode SEI
709 */
710 int ff_h264_decode_sei(H264Context *h);
711
712 /**
713 * Decode SPS
714 */
715 int ff_h264_decode_seq_parameter_set(H264Context *h);
716
717 /**
718 * compute profile from sps
719 */
720 int ff_h264_get_profile(SPS *sps);
721
722 /**
723 * Decode PPS
724 */
725 int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
726
727 /**
728 * Decode a network abstraction layer unit.
729 * @param consumed is the number of bytes used as input
730 * @param length is the length of the array
731 * @param dst_length is the number of decoded bytes FIXME here
732 * or a decode rbsp tailing?
733 * @return decoded bytes, might be src+1 if no escapes
734 */
735 const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
736 int *dst_length, int *consumed, int length);
737
738 /**
739 * Free any data that may have been allocated in the H264 context
740 * like SPS, PPS etc.
741 */
742 void ff_h264_free_context(H264Context *h);
743
744 /**
745 * Reconstruct bitstream slice_type.
746 */
747 int ff_h264_get_slice_type(const H264Context *h);
748
749 /**
750 * Allocate tables.
751 * needs width/height
752 */
753 int ff_h264_alloc_tables(H264Context *h);
754
755 /**
756 * Fill the default_ref_list.
757 */
758 int ff_h264_fill_default_ref_list(H264Context *h);
759
760 int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
761 void ff_h264_fill_mbaff_ref_list(H264Context *h);
762 void ff_h264_remove_all_refs(H264Context *h);
763
764 /**
765 * Execute the reference picture marking (memory management control operations).
766 */
767 int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
768
769 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
770 int first_slice);
771
772 int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
773
774 /**
775 * Check if the top & left blocks are available if needed & change the
776 * dc mode so it only uses the available blocks.
777 */
778 int ff_h264_check_intra4x4_pred_mode(H264Context *h);
779
780 /**
781 * Check if the top & left blocks are available if needed & change the
782 * dc mode so it only uses the available blocks.
783 */
784 int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
785
786 void ff_h264_hl_decode_mb(H264Context *h);
787 int ff_h264_decode_extradata(H264Context *h);
788 int ff_h264_decode_init(AVCodecContext *avctx);
789 void ff_h264_decode_init_vlc(void);
790
791 /**
792 * Decode a macroblock
793 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
794 */
795 int ff_h264_decode_mb_cavlc(H264Context *h);
796
797 /**
798 * Decode a CABAC coded macroblock
799 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
800 */
801 int ff_h264_decode_mb_cabac(H264Context *h);
802
803 void ff_h264_init_cabac_states(H264Context *h);
804
805 void h264_init_dequant_tables(H264Context *h);
806
807 void ff_h264_direct_dist_scale_factor(H264Context *const h);
808 void ff_h264_direct_ref_list_init(H264Context *const h);
809 void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type);
810
811 void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y,
812 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
813 unsigned int linesize, unsigned int uvlinesize);
814 void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
815 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
816 unsigned int linesize, unsigned int uvlinesize);
817
818 /**
819 * Reset SEI values at the beginning of the frame.
820 *
821 * @param h H.264 context.
822 */
823 void ff_h264_reset_sei(H264Context *h);
824
825 /*
826 * o-o o-o
827 * / / /
828 * o-o o-o
829 * ,---'
830 * o-o o-o
831 * / / /
832 * o-o o-o
833 */
834
835 /* Scan8 organization:
836 * 0 1 2 3 4 5 6 7
837 * 0 DY y y y y y
838 * 1 y Y Y Y Y
839 * 2 y Y Y Y Y
840 * 3 y Y Y Y Y
841 * 4 y Y Y Y Y
842 * 5 DU u u u u u
843 * 6 u U U U U
844 * 7 u U U U U
845 * 8 u U U U U
846 * 9 u U U U U
847 * 10 DV v v v v v
848 * 11 v V V V V
849 * 12 v V V V V
850 * 13 v V V V V
851 * 14 v V V V V
852 * DY/DU/DV are for luma/chroma DC.
853 */
854
855 #define LUMA_DC_BLOCK_INDEX 48
856 #define CHROMA_DC_BLOCK_INDEX 49
857
858 // This table must be here because scan8[constant] must be known at compiletime
859 static const uint8_t scan8[16 * 3 + 3] = {
860 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
861 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
862 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
863 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
864 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
865 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
866 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
867 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
868 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
869 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
870 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
871 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
872 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
873 };
874
875 static av_always_inline uint32_t pack16to32(int a, int b)
876 {
877 #if HAVE_BIGENDIAN
878 return (b & 0xFFFF) + (a << 16);
879 #else
880 return (a & 0xFFFF) + (b << 16);
881 #endif
882 }
883
884 static av_always_inline uint16_t pack8to16(int a, int b)
885 {
886 #if HAVE_BIGENDIAN
887 return (b & 0xFF) + (a << 8);
888 #else
889 return (a & 0xFF) + (b << 8);
890 #endif
891 }
892
893 /**
894 * Get the chroma qp.
895 */
896 static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
897 {
898 return h->pps.chroma_qp_table[t][qscale];
899 }
900
901 /**
902 * Get the predicted intra4x4 prediction mode.
903 */
904 static av_always_inline int pred_intra_mode(H264Context *h, int n)
905 {
906 const int index8 = scan8[n];
907 const int left = h->intra4x4_pred_mode_cache[index8 - 1];
908 const int top = h->intra4x4_pred_mode_cache[index8 - 8];
909 const int min = FFMIN(left, top);
910
911 tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min);
912
913 if (min < 0)
914 return DC_PRED;
915 else
916 return min;
917 }
918
919 static av_always_inline void write_back_intra_pred_mode(H264Context *h)
920 {
921 int8_t *i4x4 = h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
922 int8_t *i4x4_cache = h->intra4x4_pred_mode_cache;
923
924 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
925 i4x4[4] = i4x4_cache[7 + 8 * 3];
926 i4x4[5] = i4x4_cache[7 + 8 * 2];
927 i4x4[6] = i4x4_cache[7 + 8 * 1];
928 }
929
930 static av_always_inline void write_back_non_zero_count(H264Context *h)
931 {
932 const int mb_xy = h->mb_xy;
933 uint8_t *nnz = h->non_zero_count[mb_xy];
934 uint8_t *nnz_cache = h->non_zero_count_cache;
935
936 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
937 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
938 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
939 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
940 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
941 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
942 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
943 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
944
945 if (!h->chroma_y_shift) {
946 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
947 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
948 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
949 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
950 }
951 }
952
953 static av_always_inline void write_back_motion_list(H264Context *h,
954 int b_stride,
955 int b_xy, int b8_xy,
956 int mb_type, int list)
957 {
958 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
959 int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
960 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
961 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
962 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
963 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
964 if (CABAC(h)) {
965 uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8 * h->mb_xy
966 : h->mb2br_xy[h->mb_xy]];
967 uint8_t(*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
968 if (IS_SKIP(mb_type)) {
969 AV_ZERO128(mvd_dst);
970 } else {
971 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
972 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
973 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
974 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
975 }
976 }
977
978 {
979 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
980 int8_t *ref_cache = h->ref_cache[list];
981 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
982 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
983 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
984 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
985 }
986 }
987
988 static av_always_inline void write_back_motion(H264Context *h, int mb_type)
989 {
990 const int b_stride = h->b_stride;
991 const int b_xy = 4 * h->mb_x + 4 * h->mb_y * h->b_stride; // try mb2b(8)_xy
992 const int b8_xy = 4 * h->mb_xy;
993
994 if (USES_LIST(mb_type, 0)) {
995 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 0);
996 } else {
997 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
998 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
999 }
1000 if (USES_LIST(mb_type, 1))
1001 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 1);
1002
1003 if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
1004 if (IS_8X8(mb_type)) {
1005 uint8_t *direct_table = &h->direct_table[4 * h->mb_xy];
1006 direct_table[1] = h->sub_mb_type[1] >> 1;
1007 direct_table[2] = h->sub_mb_type[2] >> 1;
1008 direct_table[3] = h->sub_mb_type[3] >> 1;
1009 }
1010 }
1011 }
1012
1013 static av_always_inline int get_dct8x8_allowed(H264Context *h)
1014 {
1015 if (h->sps.direct_8x8_inference_flag)
1016 return !(AV_RN64A(h->sub_mb_type) &
1017 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
1018 0x0001000100010001ULL));
1019 else
1020 return !(AV_RN64A(h->sub_mb_type) &
1021 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
1022 0x0001000100010001ULL));
1023 }
1024
1025 int ff_h264_field_end(H264Context *h, int in_setup);
1026
1027 int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
1028 void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
1029
1030 int ff_h264_context_init(H264Context *h);
1031 int ff_h264_set_parameter_from_sps(H264Context *h);
1032
1033 void ff_h264_draw_horiz_band(H264Context *h, int y, int height);
1034 int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
1035 int ff_pred_weight_table(H264Context *h);
1036 int ff_set_ref_count(H264Context *h);
1037
1038 int ff_h264_decode_slice_header(H264Context *h, H264Context *h0);
1039 int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count);
1040 int ff_h264_update_thread_context(AVCodecContext *dst,
1041 const AVCodecContext *src);
1042
1043 void ff_h264_flush_change(H264Context *h);
1044
1045 void ff_h264_free_tables(H264Context *h, int free_rbsp);
1046
1047 #endif /* AVCODEC_H264_H */