Imported Debian version 2.5.0~trusty1.1
[deb_ffmpeg.git] / ffmpeg / 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 FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * H.264 / AVC / MPEG4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28 #ifndef AVCODEC_H264_H
29 #define AVCODEC_H264_H
30
31 #include "libavutil/intreadwrite.h"
32 #include "cabac.h"
33 #include "error_resilience.h"
34 #include "get_bits.h"
35 #include "h264chroma.h"
36 #include "h264dsp.h"
37 #include "h264pred.h"
38 #include "h264qpel.h"
39 #include "internal.h" // for avpriv_find_start_code()
40 #include "me_cmp.h"
41 #include "mpegutils.h"
42 #include "parser.h"
43 #include "qpeldsp.h"
44 #include "rectangle.h"
45 #include "videodsp.h"
46
47 #define H264_MAX_PICTURE_COUNT 36
48 #define H264_MAX_THREADS 32
49
50 #define MAX_SPS_COUNT 32
51 #define MAX_PPS_COUNT 256
52
53 #define MAX_MMCO_COUNT 66
54
55 #define MAX_DELAYED_PIC_COUNT 16
56
57 #define MAX_MBPAIR_SIZE (256*1024) // a tighter bound could be calculated if someone cares about a few bytes
58
59 /* Compiling in interlaced support reduces the speed
60 * of progressive decoding by about 2%. */
61 #define ALLOW_INTERLACE
62
63 #define FMO 0
64
65 /**
66 * The maximum number of slices supported by the decoder.
67 * must be a power of 2
68 */
69 #define MAX_SLICES 16
70
71 #ifdef ALLOW_INTERLACE
72 #define MB_MBAFF(h) (h)->mb_mbaff
73 #define MB_FIELD(h) (h)->mb_field_decoding_flag
74 #define FRAME_MBAFF(h) (h)->mb_aff_frame
75 #define FIELD_PICTURE(h) ((h)->picture_structure != PICT_FRAME)
76 #define LEFT_MBS 2
77 #define LTOP 0
78 #define LBOT 1
79 #define LEFT(i) (i)
80 #else
81 #define MB_MBAFF(h) 0
82 #define MB_FIELD(h) 0
83 #define FRAME_MBAFF(h) 0
84 #define FIELD_PICTURE(h) 0
85 #undef IS_INTERLACED
86 #define IS_INTERLACED(mb_type) 0
87 #define LEFT_MBS 1
88 #define LTOP 0
89 #define LBOT 0
90 #define LEFT(i) 0
91 #endif
92 #define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
93
94 #ifndef CABAC
95 #define CABAC(h) (h)->pps.cabac
96 #endif
97
98 #define CHROMA(h) ((h)->sps.chroma_format_idc)
99 #define CHROMA422(h) ((h)->sps.chroma_format_idc == 2)
100 #define CHROMA444(h) ((h)->sps.chroma_format_idc == 3)
101
102 #define EXTENDED_SAR 255
103
104 #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
105 #define MB_TYPE_8x8DCT 0x01000000
106 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
107 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
108
109 #define QP_MAX_NUM (51 + 6*6) // The maximum supported qp
110
111 /* NAL unit types */
112 enum {
113 NAL_SLICE = 1,
114 NAL_DPA = 2,
115 NAL_DPB = 3,
116 NAL_DPC = 4,
117 NAL_IDR_SLICE = 5,
118 NAL_SEI = 6,
119 NAL_SPS = 7,
120 NAL_PPS = 8,
121 NAL_AUD = 9,
122 NAL_END_SEQUENCE = 10,
123 NAL_END_STREAM = 11,
124 NAL_FILLER_DATA = 12,
125 NAL_SPS_EXT = 13,
126 NAL_AUXILIARY_SLICE = 19,
127 NAL_FF_IGNORE = 0xff0f001,
128 };
129
130 /**
131 * SEI message types
132 */
133 typedef enum {
134 SEI_TYPE_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
135 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
136 SEI_TYPE_USER_DATA_ITU_T_T35 = 4, ///< user data registered by ITU-T Recommendation T.35
137 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
138 SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
139 SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
140 SEI_TYPE_DISPLAY_ORIENTATION = 47, ///< display orientation
141 } SEI_Type;
142
143 /**
144 * pic_struct in picture timing SEI message
145 */
146 typedef enum {
147 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
148 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
149 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
150 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
151 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
152 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
153 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
154 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
155 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
156 } SEI_PicStructType;
157
158 /**
159 * frame_packing_arrangement types
160 */
161 typedef enum {
162 SEI_FPA_TYPE_CHECKERBOARD = 0,
163 SEI_FPA_TYPE_INTERLEAVE_COLUMN = 1,
164 SEI_FPA_TYPE_INTERLEAVE_ROW = 2,
165 SEI_FPA_TYPE_SIDE_BY_SIDE = 3,
166 SEI_FPA_TYPE_TOP_BOTTOM = 4,
167 SEI_FPA_TYPE_INTERLEAVE_TEMPORAL = 5,
168 SEI_FPA_TYPE_2D = 6,
169 } SEI_FpaType;
170
171 /**
172 * Sequence parameter set
173 */
174 typedef struct SPS {
175 unsigned int sps_id;
176 int profile_idc;
177 int level_idc;
178 int chroma_format_idc;
179 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
180 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
181 int poc_type; ///< pic_order_cnt_type
182 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
183 int delta_pic_order_always_zero_flag;
184 int offset_for_non_ref_pic;
185 int offset_for_top_to_bottom_field;
186 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
187 int ref_frame_count; ///< num_ref_frames
188 int gaps_in_frame_num_allowed_flag;
189 int mb_width; ///< pic_width_in_mbs_minus1 + 1
190 int mb_height; ///< pic_height_in_map_units_minus1 + 1
191 int frame_mbs_only_flag;
192 int mb_aff; ///< mb_adaptive_frame_field_flag
193 int direct_8x8_inference_flag;
194 int crop; ///< frame_cropping_flag
195
196 /* those 4 are already in luma samples */
197 unsigned int crop_left; ///< frame_cropping_rect_left_offset
198 unsigned int crop_right; ///< frame_cropping_rect_right_offset
199 unsigned int crop_top; ///< frame_cropping_rect_top_offset
200 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
201 int vui_parameters_present_flag;
202 AVRational sar;
203 int video_signal_type_present_flag;
204 int full_range;
205 int colour_description_present_flag;
206 enum AVColorPrimaries color_primaries;
207 enum AVColorTransferCharacteristic color_trc;
208 enum AVColorSpace colorspace;
209 int timing_info_present_flag;
210 uint32_t num_units_in_tick;
211 uint32_t time_scale;
212 int fixed_frame_rate_flag;
213 short offset_for_ref_frame[256]; // FIXME dyn aloc?
214 int bitstream_restriction_flag;
215 int num_reorder_frames;
216 int scaling_matrix_present;
217 uint8_t scaling_matrix4[6][16];
218 uint8_t scaling_matrix8[6][64];
219 int nal_hrd_parameters_present_flag;
220 int vcl_hrd_parameters_present_flag;
221 int pic_struct_present_flag;
222 int time_offset_length;
223 int cpb_cnt; ///< See H.264 E.1.2
224 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
225 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
226 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
227 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
228 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
229 int residual_color_transform_flag; ///< residual_colour_transform_flag
230 int constraint_set_flags; ///< constraint_set[0-3]_flag
231 int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
232 } SPS;
233
234 /**
235 * Picture parameter set
236 */
237 typedef struct PPS {
238 unsigned int sps_id;
239 int cabac; ///< entropy_coding_mode_flag
240 int pic_order_present; ///< pic_order_present_flag
241 int slice_group_count; ///< num_slice_groups_minus1 + 1
242 int mb_slice_group_map_type;
243 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
244 int weighted_pred; ///< weighted_pred_flag
245 int weighted_bipred_idc;
246 int init_qp; ///< pic_init_qp_minus26 + 26
247 int init_qs; ///< pic_init_qs_minus26 + 26
248 int chroma_qp_index_offset[2];
249 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
250 int constrained_intra_pred; ///< constrained_intra_pred_flag
251 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
252 int transform_8x8_mode; ///< transform_8x8_mode_flag
253 uint8_t scaling_matrix4[6][16];
254 uint8_t scaling_matrix8[6][64];
255 uint8_t chroma_qp_table[2][QP_MAX_NUM+1]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
256 int chroma_qp_diff;
257 } PPS;
258
259 /**
260 * Frame Packing Arrangement Type
261 */
262 typedef struct FPA {
263 int frame_packing_arrangement_id;
264 int frame_packing_arrangement_cancel_flag; ///< is previous arrangement canceled, -1 if never received
265 SEI_FpaType frame_packing_arrangement_type;
266 int frame_packing_arrangement_repetition_period;
267 int content_interpretation_type;
268 int quincunx_sampling_flag;
269 } FPA;
270
271 /**
272 * Memory management control operation opcode.
273 */
274 typedef enum MMCOOpcode {
275 MMCO_END = 0,
276 MMCO_SHORT2UNUSED,
277 MMCO_LONG2UNUSED,
278 MMCO_SHORT2LONG,
279 MMCO_SET_MAX_LONG,
280 MMCO_RESET,
281 MMCO_LONG,
282 } MMCOOpcode;
283
284 /**
285 * Memory management control operation.
286 */
287 typedef struct MMCO {
288 MMCOOpcode opcode;
289 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
290 int long_arg; ///< index, pic_num, or num long refs depending on opcode
291 } MMCO;
292
293 typedef struct H264Picture {
294 struct AVFrame f;
295 uint8_t avframe_padding[1024]; // hack to allow linking to a avutil with larger AVFrame
296 ThreadFrame tf;
297
298 AVBufferRef *qscale_table_buf;
299 int8_t *qscale_table;
300
301 AVBufferRef *motion_val_buf[2];
302 int16_t (*motion_val[2])[2];
303
304 AVBufferRef *mb_type_buf;
305 uint32_t *mb_type;
306
307 AVBufferRef *hwaccel_priv_buf;
308 void *hwaccel_picture_private; ///< hardware accelerator private data
309
310 AVBufferRef *ref_index_buf[2];
311 int8_t *ref_index[2];
312
313 int field_poc[2]; ///< top/bottom POC
314 int poc; ///< frame POC
315 int frame_num; ///< frame_num (raw frame_num from slice header)
316 int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
317 not mix pictures before and after MMCO_RESET. */
318 int pic_id; /**< pic_num (short -> no wrap version of pic_num,
319 pic_num & max_pic_num; long -> long_pic_num) */
320 int long_ref; ///< 1->long term reference 0->short term reference
321 int ref_poc[2][2][32]; ///< POCs of the frames/fields used as reference (FIXME need per slice)
322 int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
323 int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
324 int field_picture; ///< whether or not picture was encoded in separate fields
325
326 int needs_realloc; ///< picture needs to be reallocated (eg due to a frame size change)
327 int reference;
328 int recovered; ///< picture at IDR or recovery point + recovery count
329 int invalid_gap;
330 int sei_recovery_frame_cnt;
331
332 int crop;
333 int crop_left;
334 int crop_top;
335 } H264Picture;
336
337 /**
338 * H264Context
339 */
340 typedef struct H264Context {
341 AVCodecContext *avctx;
342 MECmpContext mecc;
343 VideoDSPContext vdsp;
344 H264DSPContext h264dsp;
345 H264ChromaContext h264chroma;
346 H264QpelContext h264qpel;
347 ParseContext parse_context;
348 GetBitContext gb;
349 ERContext er;
350
351 H264Picture *DPB;
352 H264Picture *cur_pic_ptr;
353 H264Picture cur_pic;
354
355 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
356 int chroma_qp[2]; // QPc
357
358 int qp_thresh; ///< QP threshold to skip loopfilter
359
360 /* coded dimensions -- 16 * mb w/h */
361 int width, height;
362 ptrdiff_t linesize, uvlinesize;
363 int chroma_x_shift, chroma_y_shift;
364
365 int qscale;
366 int droppable;
367 int data_partitioning;
368 int coded_picture_number;
369 int low_delay;
370
371 int context_initialized;
372 int flags;
373 int workaround_bugs;
374
375 int prev_mb_skipped;
376 int next_mb_skipped;
377
378 // prediction stuff
379 int chroma_pred_mode;
380 int intra16x16_pred_mode;
381
382 int topleft_mb_xy;
383 int top_mb_xy;
384 int topright_mb_xy;
385 int left_mb_xy[LEFT_MBS];
386
387 int topleft_type;
388 int top_type;
389 int topright_type;
390 int left_type[LEFT_MBS];
391
392 const uint8_t *left_block;
393 int topleft_partition;
394
395 int8_t intra4x4_pred_mode_cache[5 * 8];
396 int8_t(*intra4x4_pred_mode);
397 H264PredContext hpc;
398 unsigned int topleft_samples_available;
399 unsigned int top_samples_available;
400 unsigned int topright_samples_available;
401 unsigned int left_samples_available;
402 uint8_t (*top_borders[2])[(16 * 3) * 2];
403
404 /**
405 * non zero coeff count cache.
406 * is 64 if not available.
407 */
408 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
409
410 uint8_t (*non_zero_count)[48];
411
412 /**
413 * Motion vector cache.
414 */
415 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
416 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
417 #define LIST_NOT_USED -1 // FIXME rename?
418 #define PART_NOT_AVAILABLE -2
419
420 /**
421 * number of neighbors (top and/or left) that used 8x8 dct
422 */
423 int neighbor_transform_size;
424
425 /**
426 * block_offset[ 0..23] for frame macroblocks
427 * block_offset[24..47] for field macroblocks
428 */
429 int block_offset[2 * (16 * 3)];
430
431 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
432 uint32_t *mb2br_xy;
433 int b_stride; // FIXME use s->b4_stride
434
435 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
436 ptrdiff_t mb_uvlinesize;
437
438 unsigned current_sps_id; ///< id of the current SPS
439 SPS sps; ///< current sps
440 PPS pps; ///< current pps
441
442 int au_pps_id; ///< pps_id of current access unit
443
444 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
445 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
446 uint32_t(*dequant4_coeff[6])[16];
447 uint32_t(*dequant8_coeff[6])[64];
448
449 int slice_num;
450 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
451 int slice_type;
452 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
453 int slice_type_fixed;
454
455 // interlacing specific flags
456 int mb_aff_frame;
457 int mb_field_decoding_flag;
458 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
459 int picture_structure;
460 int first_field;
461
462 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
463
464 // Weighted pred stuff
465 int use_weight;
466 int use_weight_chroma;
467 int luma_log2_weight_denom;
468 int chroma_log2_weight_denom;
469 // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
470 int luma_weight[48][2][2];
471 int chroma_weight[48][2][2][2];
472 int implicit_weight[48][48][2];
473
474 int direct_spatial_mv_pred;
475 int col_parity;
476 int col_fieldoff;
477 int dist_scale_factor[32];
478 int dist_scale_factor_field[2][32];
479 int map_col_to_list0[2][16 + 32];
480 int map_col_to_list0_field[2][2][16 + 32];
481
482 /**
483 * num_ref_idx_l0/1_active_minus1 + 1
484 */
485 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
486 unsigned int list_count;
487 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
488 H264Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
489 * Reordered version of default_ref_list
490 * according to picture reordering in slice header */
491 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
492
493 // data partitioning
494 GetBitContext intra_gb;
495 GetBitContext inter_gb;
496 GetBitContext *intra_gb_ptr;
497 GetBitContext *inter_gb_ptr;
498
499 const uint8_t *intra_pcm_ptr;
500 DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2]; ///< as a dct coefficient is int32_t in high depth, we need to reserve twice the space.
501 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
502 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
503
504 /**
505 * Cabac
506 */
507 CABACContext cabac;
508 uint8_t cabac_state[1024];
509
510 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
511 uint16_t *cbp_table;
512 int cbp;
513 int top_cbp;
514 int left_cbp;
515 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
516 uint8_t *chroma_pred_mode_table;
517 int last_qscale_diff;
518 uint8_t (*mvd_table[2])[2];
519 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
520 uint8_t *direct_table;
521 uint8_t direct_cache[5 * 8];
522
523 uint8_t zigzag_scan[16];
524 uint8_t zigzag_scan8x8[64];
525 uint8_t zigzag_scan8x8_cavlc[64];
526 uint8_t field_scan[16];
527 uint8_t field_scan8x8[64];
528 uint8_t field_scan8x8_cavlc[64];
529 uint8_t zigzag_scan_q0[16];
530 uint8_t zigzag_scan8x8_q0[64];
531 uint8_t zigzag_scan8x8_cavlc_q0[64];
532 uint8_t field_scan_q0[16];
533 uint8_t field_scan8x8_q0[64];
534 uint8_t field_scan8x8_cavlc_q0[64];
535
536 int x264_build;
537
538 int mb_x, mb_y;
539 int resync_mb_x;
540 int resync_mb_y;
541 int mb_skip_run;
542 int mb_height, mb_width;
543 int mb_stride;
544 int mb_num;
545 int mb_xy;
546
547 int is_complex;
548
549 // deblock
550 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
551 int slice_alpha_c0_offset;
552 int slice_beta_offset;
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 uint8_t *rbsp_buffer[2];
560 unsigned int rbsp_buffer_size[2];
561
562 /**
563 * Used to parse AVC variant of h264
564 */
565 int is_avc; ///< this flag is != 0 if codec is avc1
566 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
567 int got_first; ///< this flag is != 0 if we've parsed a frame
568
569 int bit_depth_luma; ///< luma bit depth from sps to detect changes
570 int chroma_format_idc; ///< chroma format from sps to detect changes
571
572 SPS *sps_buffers[MAX_SPS_COUNT];
573 PPS *pps_buffers[MAX_PPS_COUNT];
574
575 int dequant_coeff_pps; ///< reinit tables when pps changes
576
577 uint16_t *slice_table_base;
578
579 // POC stuff
580 int poc_lsb;
581 int poc_msb;
582 int delta_poc_bottom;
583 int delta_poc[2];
584 int frame_num;
585 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
586 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
587 int frame_num_offset; ///< for POC type 2
588 int prev_frame_num_offset; ///< for POC type 2
589 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
590
591 /**
592 * frame_num for frames or 2 * frame_num + 1 for field pics.
593 */
594 int curr_pic_num;
595
596 /**
597 * max_frame_num or 2 * max_frame_num for field pics.
598 */
599 int max_pic_num;
600
601 int redundant_pic_count;
602
603 H264Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
604 H264Picture *short_ref[32];
605 H264Picture *long_ref[32];
606 H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
607 int last_pocs[MAX_DELAYED_PIC_COUNT];
608 H264Picture *next_output_pic;
609 int outputed_poc;
610 int next_outputed_poc;
611
612 /**
613 * memory management control operations buffer.
614 */
615 MMCO mmco[MAX_MMCO_COUNT];
616 int mmco_index;
617 int mmco_reset;
618
619 int long_ref_count; ///< number of actual long term references
620 int short_ref_count; ///< number of actual short term references
621
622 int cabac_init_idc;
623
624 /**
625 * @name Members for slice based multithreading
626 * @{
627 */
628 struct H264Context *thread_context[H264_MAX_THREADS];
629
630 /**
631 * current slice number, used to initialize slice_num of each thread/context
632 */
633 int current_slice;
634
635 /**
636 * Max number of threads / contexts.
637 * This is equal to AVCodecContext.thread_count unless
638 * multithreaded decoding is impossible, in which case it is
639 * reduced to 1.
640 */
641 int max_contexts;
642
643 int slice_context_count;
644
645 /**
646 * 1 if the single thread fallback warning has already been
647 * displayed, 0 otherwise.
648 */
649 int single_decode_warning;
650
651 enum AVPictureType pict_type;
652
653 int last_slice_type;
654 unsigned int last_ref_count[2];
655 /** @} */
656
657 /**
658 * pic_struct in picture timing SEI message
659 */
660 SEI_PicStructType sei_pic_struct;
661
662 /**
663 * Complement sei_pic_struct
664 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
665 * However, soft telecined frames may have these values.
666 * This is used in an attempt to flag soft telecine progressive.
667 */
668 int prev_interlaced_frame;
669
670 /**
671 * frame_packing_arrangment SEI message
672 */
673 int sei_frame_packing_present;
674 int frame_packing_arrangement_type;
675 int content_interpretation_type;
676 int quincunx_subsampling;
677
678 /**
679 * display orientation SEI message
680 */
681 int sei_display_orientation_present;
682 int sei_anticlockwise_rotation;
683 int sei_hflip, sei_vflip;
684
685 /**
686 * Bit set of clock types for fields/frames in picture timing SEI message.
687 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
688 * interlaced).
689 */
690 int sei_ct_type;
691
692 /**
693 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
694 */
695 int sei_dpb_output_delay;
696
697 /**
698 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
699 */
700 int sei_cpb_removal_delay;
701
702 /**
703 * recovery_frame_cnt from SEI message
704 *
705 * Set to -1 if no recovery point SEI message found or to number of frames
706 * before playback synchronizes. Frames having recovery point are key
707 * frames.
708 */
709 int sei_recovery_frame_cnt;
710
711 /**
712 * Are the SEI recovery points looking valid.
713 */
714 int valid_recovery_point;
715
716 FPA sei_fpa;
717
718 /**
719 * recovery_frame is the frame_num at which the next frame should
720 * be fully constructed.
721 *
722 * Set to -1 when not expecting a recovery point.
723 */
724 int recovery_frame;
725
726 /**
727 * We have seen an IDR, so all the following frames in coded order are correctly
728 * decodable.
729 */
730 #define FRAME_RECOVERED_IDR (1 << 0)
731 /**
732 * Sufficient number of frames have been decoded since a SEI recovery point,
733 * so all the following frames in presentation order are correct.
734 */
735 #define FRAME_RECOVERED_SEI (1 << 1)
736
737 int frame_recovered; ///< Initial frame has been completely recovered
738
739 int has_recovery_point;
740
741 int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
742 int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
743
744 // Timestamp stuff
745 int sei_buffering_period_present; ///< Buffering period SEI flag
746 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
747
748 int cur_chroma_format_idc;
749 uint8_t *bipred_scratchpad;
750
751 int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
752
753 uint8_t parse_history[6];
754 int parse_history_count;
755 int parse_last_mb;
756 uint8_t *edge_emu_buffer;
757 int16_t *dc_val_base;
758
759 AVBufferPool *qscale_table_pool;
760 AVBufferPool *mb_type_pool;
761 AVBufferPool *motion_val_pool;
762 AVBufferPool *ref_index_pool;
763
764 /* Motion Estimation */
765 qpel_mc_func (*qpel_put)[16];
766 qpel_mc_func (*qpel_avg)[16];
767 } H264Context;
768
769 extern const uint8_t ff_h264_chroma_qp[7][QP_MAX_NUM + 1]; ///< One chroma qp table for each possible bit depth (8-14).
770 extern const uint16_t ff_h264_mb_sizes[4];
771
772 /**
773 * Decode SEI
774 */
775 int ff_h264_decode_sei(H264Context *h);
776
777 /**
778 * Decode SPS
779 */
780 int ff_h264_decode_seq_parameter_set(H264Context *h);
781
782 /**
783 * compute profile from sps
784 */
785 int ff_h264_get_profile(SPS *sps);
786
787 /**
788 * Decode PPS
789 */
790 int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
791
792 /**
793 * Decode a network abstraction layer unit.
794 * @param consumed is the number of bytes used as input
795 * @param length is the length of the array
796 * @param dst_length is the number of decoded bytes FIXME here
797 * or a decode rbsp tailing?
798 * @return decoded bytes, might be src+1 if no escapes
799 */
800 const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
801 int *dst_length, int *consumed, int length);
802
803 /**
804 * Free any data that may have been allocated in the H264 context
805 * like SPS, PPS etc.
806 */
807 void ff_h264_free_context(H264Context *h);
808
809 /**
810 * Reconstruct bitstream slice_type.
811 */
812 int ff_h264_get_slice_type(const H264Context *h);
813
814 /**
815 * Allocate tables.
816 * needs width/height
817 */
818 int ff_h264_alloc_tables(H264Context *h);
819
820 /**
821 * Fill the default_ref_list.
822 */
823 int ff_h264_fill_default_ref_list(H264Context *h);
824
825 int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
826 void ff_h264_fill_mbaff_ref_list(H264Context *h);
827 void ff_h264_remove_all_refs(H264Context *h);
828
829 /**
830 * Execute the reference picture marking (memory management control operations).
831 */
832 int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
833
834 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
835 int first_slice);
836
837 int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
838
839 /**
840 * Check if the top & left blocks are available if needed & change the
841 * dc mode so it only uses the available blocks.
842 */
843 int ff_h264_check_intra4x4_pred_mode(H264Context *h);
844
845 /**
846 * Check if the top & left blocks are available if needed & change the
847 * dc mode so it only uses the available blocks.
848 */
849 int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
850
851 void ff_h264_hl_decode_mb(H264Context *h);
852 int ff_h264_decode_extradata(H264Context *h, const uint8_t *buf, int size);
853 int ff_h264_decode_init(AVCodecContext *avctx);
854 void ff_h264_decode_init_vlc(void);
855
856 /**
857 * Decode a macroblock
858 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
859 */
860 int ff_h264_decode_mb_cavlc(H264Context *h);
861
862 /**
863 * Decode a CABAC coded macroblock
864 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
865 */
866 int ff_h264_decode_mb_cabac(H264Context *h);
867
868 void ff_h264_init_cabac_states(H264Context *h);
869
870 void h264_init_dequant_tables(H264Context *h);
871
872 void ff_h264_direct_dist_scale_factor(H264Context *const h);
873 void ff_h264_direct_ref_list_init(H264Context *const h);
874 void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type);
875
876 void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y,
877 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
878 unsigned int linesize, unsigned int uvlinesize);
879 void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
880 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
881 unsigned int linesize, unsigned int uvlinesize);
882
883 /**
884 * Reset SEI values at the beginning of the frame.
885 *
886 * @param h H.264 context.
887 */
888 void ff_h264_reset_sei(H264Context *h);
889
890 /**
891 * Get stereo_mode string from the h264 frame_packing_arrangement
892 * @param h H.264 context.
893 */
894 const char* ff_h264_sei_stereo_mode(H264Context *h);
895
896 /*
897 * o-o o-o
898 * / / /
899 * o-o o-o
900 * ,---'
901 * o-o o-o
902 * / / /
903 * o-o o-o
904 */
905
906 /* Scan8 organization:
907 * 0 1 2 3 4 5 6 7
908 * 0 DY y y y y y
909 * 1 y Y Y Y Y
910 * 2 y Y Y Y Y
911 * 3 y Y Y Y Y
912 * 4 y Y Y Y Y
913 * 5 DU u u u u u
914 * 6 u U U U U
915 * 7 u U U U U
916 * 8 u U U U U
917 * 9 u U U U U
918 * 10 DV v v v v v
919 * 11 v V V V V
920 * 12 v V V V V
921 * 13 v V V V V
922 * 14 v V V V V
923 * DY/DU/DV are for luma/chroma DC.
924 */
925
926 #define LUMA_DC_BLOCK_INDEX 48
927 #define CHROMA_DC_BLOCK_INDEX 49
928
929 // This table must be here because scan8[constant] must be known at compiletime
930 static const uint8_t scan8[16 * 3 + 3] = {
931 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
932 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
933 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
934 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
935 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
936 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
937 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
938 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
939 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
940 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
941 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
942 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
943 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
944 };
945
946 static av_always_inline uint32_t pack16to32(int a, int b)
947 {
948 #if HAVE_BIGENDIAN
949 return (b & 0xFFFF) + (a << 16);
950 #else
951 return (a & 0xFFFF) + (b << 16);
952 #endif
953 }
954
955 static av_always_inline uint16_t pack8to16(int a, int b)
956 {
957 #if HAVE_BIGENDIAN
958 return (b & 0xFF) + (a << 8);
959 #else
960 return (a & 0xFF) + (b << 8);
961 #endif
962 }
963
964 /**
965 * Get the chroma qp.
966 */
967 static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
968 {
969 return h->pps.chroma_qp_table[t][qscale];
970 }
971
972 /**
973 * Get the predicted intra4x4 prediction mode.
974 */
975 static av_always_inline int pred_intra_mode(H264Context *h, int n)
976 {
977 const int index8 = scan8[n];
978 const int left = h->intra4x4_pred_mode_cache[index8 - 1];
979 const int top = h->intra4x4_pred_mode_cache[index8 - 8];
980 const int min = FFMIN(left, top);
981
982 tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min);
983
984 if (min < 0)
985 return DC_PRED;
986 else
987 return min;
988 }
989
990 static av_always_inline void write_back_intra_pred_mode(H264Context *h)
991 {
992 int8_t *i4x4 = h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
993 int8_t *i4x4_cache = h->intra4x4_pred_mode_cache;
994
995 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
996 i4x4[4] = i4x4_cache[7 + 8 * 3];
997 i4x4[5] = i4x4_cache[7 + 8 * 2];
998 i4x4[6] = i4x4_cache[7 + 8 * 1];
999 }
1000
1001 static av_always_inline void write_back_non_zero_count(H264Context *h)
1002 {
1003 const int mb_xy = h->mb_xy;
1004 uint8_t *nnz = h->non_zero_count[mb_xy];
1005 uint8_t *nnz_cache = h->non_zero_count_cache;
1006
1007 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
1008 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
1009 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
1010 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
1011 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
1012 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
1013 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
1014 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
1015
1016 if (!h->chroma_y_shift) {
1017 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
1018 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
1019 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
1020 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
1021 }
1022 }
1023
1024 static av_always_inline void write_back_motion_list(H264Context *h,
1025 int b_stride,
1026 int b_xy, int b8_xy,
1027 int mb_type, int list)
1028 {
1029 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
1030 int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
1031 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
1032 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
1033 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
1034 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
1035 if (CABAC(h)) {
1036 uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8 * h->mb_xy
1037 : h->mb2br_xy[h->mb_xy]];
1038 uint8_t(*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
1039 if (IS_SKIP(mb_type)) {
1040 AV_ZERO128(mvd_dst);
1041 } else {
1042 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
1043 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
1044 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
1045 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
1046 }
1047 }
1048
1049 {
1050 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
1051 int8_t *ref_cache = h->ref_cache[list];
1052 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
1053 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
1054 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
1055 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
1056 }
1057 }
1058
1059 static av_always_inline void write_back_motion(H264Context *h, int mb_type)
1060 {
1061 const int b_stride = h->b_stride;
1062 const int b_xy = 4 * h->mb_x + 4 * h->mb_y * h->b_stride; // try mb2b(8)_xy
1063 const int b8_xy = 4 * h->mb_xy;
1064
1065 if (USES_LIST(mb_type, 0)) {
1066 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 0);
1067 } else {
1068 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
1069 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
1070 }
1071 if (USES_LIST(mb_type, 1))
1072 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 1);
1073
1074 if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
1075 if (IS_8X8(mb_type)) {
1076 uint8_t *direct_table = &h->direct_table[4 * h->mb_xy];
1077 direct_table[1] = h->sub_mb_type[1] >> 1;
1078 direct_table[2] = h->sub_mb_type[2] >> 1;
1079 direct_table[3] = h->sub_mb_type[3] >> 1;
1080 }
1081 }
1082 }
1083
1084 static av_always_inline int get_dct8x8_allowed(H264Context *h)
1085 {
1086 if (h->sps.direct_8x8_inference_flag)
1087 return !(AV_RN64A(h->sub_mb_type) &
1088 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
1089 0x0001000100010001ULL));
1090 else
1091 return !(AV_RN64A(h->sub_mb_type) &
1092 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
1093 0x0001000100010001ULL));
1094 }
1095
1096 static inline int find_start_code(const uint8_t *buf, int buf_size,
1097 int buf_index, int next_avc)
1098 {
1099 uint32_t state = -1;
1100
1101 buf_index = avpriv_find_start_code(buf + buf_index, buf + next_avc + 1, &state) - buf - 1;
1102
1103 return FFMIN(buf_index, buf_size);
1104 }
1105
1106 static inline int get_avc_nalsize(H264Context *h, const uint8_t *buf,
1107 int buf_size, int *buf_index)
1108 {
1109 int i, nalsize = 0;
1110
1111 if (*buf_index >= buf_size - h->nal_length_size)
1112 return -1;
1113
1114 for (i = 0; i < h->nal_length_size; i++)
1115 nalsize = ((unsigned)nalsize << 8) | buf[(*buf_index)++];
1116 if (nalsize <= 0 || nalsize > buf_size - *buf_index) {
1117 av_log(h->avctx, AV_LOG_ERROR,
1118 "AVC: nal size %d\n", nalsize);
1119 return -1;
1120 }
1121 return nalsize;
1122 }
1123
1124 int ff_h264_field_end(H264Context *h, int in_setup);
1125
1126 int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
1127 void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
1128
1129 int ff_h264_context_init(H264Context *h);
1130 int ff_h264_set_parameter_from_sps(H264Context *h);
1131
1132 void ff_h264_draw_horiz_band(H264Context *h, int y, int height);
1133 int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
1134 int ff_pred_weight_table(H264Context *h);
1135 int ff_set_ref_count(H264Context *h);
1136
1137 int ff_h264_decode_slice_header(H264Context *h, H264Context *h0);
1138 #define SLICE_SINGLETHREAD 1
1139 #define SLICE_SKIPED 2
1140
1141 int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count);
1142 int ff_h264_update_thread_context(AVCodecContext *dst,
1143 const AVCodecContext *src);
1144
1145 void ff_h264_flush_change(H264Context *h);
1146
1147 void ff_h264_free_tables(H264Context *h, int free_rbsp);
1148
1149 void ff_h264_set_erpic(ERPicture *dst, H264Picture *src);
1150
1151 #endif /* AVCODEC_H264_H */