6eda2a693f5b2301716294633553488a617f1f5f
[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 "get_bits.h"
34 #include "mpegvideo.h"
35 #include "h264chroma.h"
36 #include "h264dsp.h"
37 #include "h264pred.h"
38 #include "h264qpel.h"
39 #include "rectangle.h"
40
41 #define MAX_SPS_COUNT 32
42 #define MAX_PPS_COUNT 256
43
44 #define MAX_MMCO_COUNT 66
45
46 #define MAX_DELAYED_PIC_COUNT 16
47
48 /* Compiling in interlaced support reduces the speed
49 * of progressive decoding by about 2%. */
50 #define ALLOW_INTERLACE
51
52 #define FMO 0
53
54 /**
55 * The maximum number of slices supported by the decoder.
56 * must be a power of 2
57 */
58 #define MAX_SLICES 16
59
60 #ifdef ALLOW_INTERLACE
61 #define MB_MBAFF h->mb_mbaff
62 #define MB_FIELD h->mb_field_decoding_flag
63 #define FRAME_MBAFF h->mb_aff_frame
64 #define FIELD_PICTURE (h->picture_structure != PICT_FRAME)
65 #define LEFT_MBS 2
66 #define LTOP 0
67 #define LBOT 1
68 #define LEFT(i) (i)
69 #else
70 #define MB_MBAFF 0
71 #define MB_FIELD 0
72 #define FRAME_MBAFF 0
73 #define FIELD_PICTURE 0
74 #undef IS_INTERLACED
75 #define IS_INTERLACED(mb_type) 0
76 #define LEFT_MBS 1
77 #define LTOP 0
78 #define LBOT 0
79 #define LEFT(i) 0
80 #endif
81 #define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
82
83 #ifndef CABAC
84 #define CABAC h->pps.cabac
85 #endif
86
87 #define CHROMA422 (h->sps.chroma_format_idc == 2)
88 #define CHROMA444 (h->sps.chroma_format_idc == 3)
89
90 #define EXTENDED_SAR 255
91
92 #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
93 #define MB_TYPE_8x8DCT 0x01000000
94 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
95 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
96
97 #define QP_MAX_NUM (51 + 2 * 6) // The maximum supported qp
98
99 /* NAL unit types */
100 enum {
101 NAL_SLICE = 1,
102 NAL_DPA,
103 NAL_DPB,
104 NAL_DPC,
105 NAL_IDR_SLICE,
106 NAL_SEI,
107 NAL_SPS,
108 NAL_PPS,
109 NAL_AUD,
110 NAL_END_SEQUENCE,
111 NAL_END_STREAM,
112 NAL_FILLER_DATA,
113 NAL_SPS_EXT,
114 NAL_AUXILIARY_SLICE = 19,
115 NAL_FF_IGNORE = 0xff0f001,
116 };
117
118 /**
119 * SEI message types
120 */
121 typedef enum {
122 SEI_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
123 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
124 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
125 SEI_TYPE_RECOVERY_POINT = 6 ///< recovery point (frame # to decoder sync)
126 } SEI_Type;
127
128 /**
129 * pic_struct in picture timing SEI message
130 */
131 typedef enum {
132 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
133 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
134 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
135 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
136 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
137 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
138 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
139 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
140 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
141 } SEI_PicStructType;
142
143 /**
144 * Sequence parameter set
145 */
146 typedef struct SPS {
147 int profile_idc;
148 int level_idc;
149 int chroma_format_idc;
150 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
151 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
152 int poc_type; ///< pic_order_cnt_type
153 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
154 int delta_pic_order_always_zero_flag;
155 int offset_for_non_ref_pic;
156 int offset_for_top_to_bottom_field;
157 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
158 int ref_frame_count; ///< num_ref_frames
159 int gaps_in_frame_num_allowed_flag;
160 int mb_width; ///< pic_width_in_mbs_minus1 + 1
161 int mb_height; ///< pic_height_in_map_units_minus1 + 1
162 int frame_mbs_only_flag;
163 int mb_aff; ///< mb_adaptive_frame_field_flag
164 int direct_8x8_inference_flag;
165 int crop; ///< frame_cropping_flag
166 unsigned int crop_left; ///< frame_cropping_rect_left_offset
167 unsigned int crop_right; ///< frame_cropping_rect_right_offset
168 unsigned int crop_top; ///< frame_cropping_rect_top_offset
169 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
170 int vui_parameters_present_flag;
171 AVRational sar;
172 int video_signal_type_present_flag;
173 int full_range;
174 int colour_description_present_flag;
175 enum AVColorPrimaries color_primaries;
176 enum AVColorTransferCharacteristic color_trc;
177 enum AVColorSpace colorspace;
178 int timing_info_present_flag;
179 uint32_t num_units_in_tick;
180 uint32_t time_scale;
181 int fixed_frame_rate_flag;
182 short offset_for_ref_frame[256]; // FIXME dyn aloc?
183 int bitstream_restriction_flag;
184 int num_reorder_frames;
185 int scaling_matrix_present;
186 uint8_t scaling_matrix4[6][16];
187 uint8_t scaling_matrix8[6][64];
188 int nal_hrd_parameters_present_flag;
189 int vcl_hrd_parameters_present_flag;
190 int pic_struct_present_flag;
191 int time_offset_length;
192 int cpb_cnt; ///< See H.264 E.1.2
193 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
194 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
195 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
196 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
197 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
198 int residual_color_transform_flag; ///< residual_colour_transform_flag
199 int constraint_set_flags; ///< constraint_set[0-3]_flag
200 int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
201 } SPS;
202
203 /**
204 * Picture parameter set
205 */
206 typedef struct PPS {
207 unsigned int sps_id;
208 int cabac; ///< entropy_coding_mode_flag
209 int pic_order_present; ///< pic_order_present_flag
210 int slice_group_count; ///< num_slice_groups_minus1 + 1
211 int mb_slice_group_map_type;
212 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
213 int weighted_pred; ///< weighted_pred_flag
214 int weighted_bipred_idc;
215 int init_qp; ///< pic_init_qp_minus26 + 26
216 int init_qs; ///< pic_init_qs_minus26 + 26
217 int chroma_qp_index_offset[2];
218 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
219 int constrained_intra_pred; ///< constrained_intra_pred_flag
220 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
221 int transform_8x8_mode; ///< transform_8x8_mode_flag
222 uint8_t scaling_matrix4[6][16];
223 uint8_t scaling_matrix8[6][64];
224 uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
225 int chroma_qp_diff;
226 } PPS;
227
228 /**
229 * Memory management control operation opcode.
230 */
231 typedef enum MMCOOpcode {
232 MMCO_END = 0,
233 MMCO_SHORT2UNUSED,
234 MMCO_LONG2UNUSED,
235 MMCO_SHORT2LONG,
236 MMCO_SET_MAX_LONG,
237 MMCO_RESET,
238 MMCO_LONG,
239 } MMCOOpcode;
240
241 /**
242 * Memory management control operation.
243 */
244 typedef struct MMCO {
245 MMCOOpcode opcode;
246 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
247 int long_arg; ///< index, pic_num, or num long refs depending on opcode
248 } MMCO;
249
250 /**
251 * H264Context
252 */
253 typedef struct H264Context {
254 AVCodecContext *avctx;
255 DSPContext dsp;
256 VideoDSPContext vdsp;
257 H264DSPContext h264dsp;
258 H264ChromaContext h264chroma;
259 H264QpelContext h264qpel;
260 MotionEstContext me;
261 ParseContext parse_context;
262 GetBitContext gb;
263 ERContext er;
264
265 Picture *DPB;
266 Picture *cur_pic_ptr;
267 Picture cur_pic;
268 int picture_count;
269 int picture_range_start, picture_range_end;
270
271 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
272 int chroma_qp[2]; // QPc
273
274 int qp_thresh; ///< QP threshold to skip loopfilter
275
276 int width, height;
277 int linesize, uvlinesize;
278 int chroma_x_shift, chroma_y_shift;
279
280 int qscale;
281 int droppable;
282 int data_partitioning;
283 int coded_picture_number;
284 int low_delay;
285
286 int context_initialized;
287 int flags;
288 int workaround_bugs;
289
290 int prev_mb_skipped;
291 int next_mb_skipped;
292
293 // prediction stuff
294 int chroma_pred_mode;
295 int intra16x16_pred_mode;
296
297 int topleft_mb_xy;
298 int top_mb_xy;
299 int topright_mb_xy;
300 int left_mb_xy[LEFT_MBS];
301
302 int topleft_type;
303 int top_type;
304 int topright_type;
305 int left_type[LEFT_MBS];
306
307 const uint8_t *left_block;
308 int topleft_partition;
309
310 int8_t intra4x4_pred_mode_cache[5 * 8];
311 int8_t(*intra4x4_pred_mode);
312 H264PredContext hpc;
313 unsigned int topleft_samples_available;
314 unsigned int top_samples_available;
315 unsigned int topright_samples_available;
316 unsigned int left_samples_available;
317 uint8_t (*top_borders[2])[(16 * 3) * 2];
318
319 /**
320 * non zero coeff count cache.
321 * is 64 if not available.
322 */
323 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
324
325 uint8_t (*non_zero_count)[48];
326
327 /**
328 * Motion vector cache.
329 */
330 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
331 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
332 #define LIST_NOT_USED -1 // FIXME rename?
333 #define PART_NOT_AVAILABLE -2
334
335 /**
336 * number of neighbors (top and/or left) that used 8x8 dct
337 */
338 int neighbor_transform_size;
339
340 /**
341 * block_offset[ 0..23] for frame macroblocks
342 * block_offset[24..47] for field macroblocks
343 */
344 int block_offset[2 * (16 * 3)];
345
346 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
347 uint32_t *mb2br_xy;
348 int b_stride; // FIXME use s->b4_stride
349
350 int mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
351 int mb_uvlinesize;
352
353 unsigned current_sps_id; ///< id of the current SPS
354 SPS sps; ///< current sps
355
356 /**
357 * current pps
358 */
359 PPS pps; // FIXME move to Picture perhaps? (->no) do we need that?
360
361 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
362 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
363 uint32_t(*dequant4_coeff[6])[16];
364 uint32_t(*dequant8_coeff[6])[64];
365
366 int slice_num;
367 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
368 int slice_type;
369 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
370 int slice_type_fixed;
371
372 // interlacing specific flags
373 int mb_aff_frame;
374 int mb_field_decoding_flag;
375 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
376 int picture_structure;
377 int first_field;
378
379 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
380
381 // Weighted pred stuff
382 int use_weight;
383 int use_weight_chroma;
384 int luma_log2_weight_denom;
385 int chroma_log2_weight_denom;
386 // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
387 int luma_weight[48][2][2];
388 int chroma_weight[48][2][2][2];
389 int implicit_weight[48][48][2];
390
391 int direct_spatial_mv_pred;
392 int col_parity;
393 int col_fieldoff;
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 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
405 Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
406 * Reordered version of default_ref_list
407 * according to picture reordering in slice header */
408 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
409
410 // data partitioning
411 GetBitContext intra_gb;
412 GetBitContext inter_gb;
413 GetBitContext *intra_gb_ptr;
414 GetBitContext *inter_gb_ptr;
415
416 const uint8_t *intra_pcm_ptr;
417 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.
418 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
419 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
420
421 /**
422 * Cabac
423 */
424 CABACContext cabac;
425 uint8_t cabac_state[1024];
426
427 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
428 uint16_t *cbp_table;
429 int cbp;
430 int top_cbp;
431 int left_cbp;
432 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
433 uint8_t *chroma_pred_mode_table;
434 int last_qscale_diff;
435 uint8_t (*mvd_table[2])[2];
436 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
437 uint8_t *direct_table;
438 uint8_t direct_cache[5 * 8];
439
440 uint8_t zigzag_scan[16];
441 uint8_t zigzag_scan8x8[64];
442 uint8_t zigzag_scan8x8_cavlc[64];
443 uint8_t field_scan[16];
444 uint8_t field_scan8x8[64];
445 uint8_t field_scan8x8_cavlc[64];
446 const uint8_t *zigzag_scan_q0;
447 const uint8_t *zigzag_scan8x8_q0;
448 const uint8_t *zigzag_scan8x8_cavlc_q0;
449 const uint8_t *field_scan_q0;
450 const uint8_t *field_scan8x8_q0;
451 const uint8_t *field_scan8x8_cavlc_q0;
452
453 int x264_build;
454
455 int mb_x, mb_y;
456 int resync_mb_x;
457 int resync_mb_y;
458 int mb_skip_run;
459 int mb_height, mb_width;
460 int mb_stride;
461 int mb_num;
462 int mb_xy;
463
464 int is_complex;
465
466 // deblock
467 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
468 int slice_alpha_c0_offset;
469 int slice_beta_offset;
470
471 // =============================================================
472 // Things below are not used in the MB or more inner code
473
474 int nal_ref_idc;
475 int nal_unit_type;
476 uint8_t *rbsp_buffer[2];
477 unsigned int rbsp_buffer_size[2];
478
479 /**
480 * Used to parse AVC variant of h264
481 */
482 int is_avc; ///< this flag is != 0 if codec is avc1
483 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
484 int got_first; ///< this flag is != 0 if we've parsed a frame
485
486 int bit_depth_luma; ///< luma bit depth from sps to detect changes
487 int chroma_format_idc; ///< chroma format from sps to detect changes
488
489 SPS *sps_buffers[MAX_SPS_COUNT];
490 PPS *pps_buffers[MAX_PPS_COUNT];
491
492 int dequant_coeff_pps; ///< reinit tables when pps changes
493
494 uint16_t *slice_table_base;
495
496 // POC stuff
497 int poc_lsb;
498 int poc_msb;
499 int delta_poc_bottom;
500 int delta_poc[2];
501 int frame_num;
502 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
503 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
504 int frame_num_offset; ///< for POC type 2
505 int prev_frame_num_offset; ///< for POC type 2
506 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
507
508 /**
509 * frame_num for frames or 2 * frame_num + 1 for field pics.
510 */
511 int curr_pic_num;
512
513 /**
514 * max_frame_num or 2 * max_frame_num for field pics.
515 */
516 int max_pic_num;
517
518 int redundant_pic_count;
519
520 Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
521 Picture *short_ref[32];
522 Picture *long_ref[32];
523 Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
524 int last_pocs[MAX_DELAYED_PIC_COUNT];
525 Picture *next_output_pic;
526 int outputed_poc;
527 int next_outputed_poc;
528
529 /**
530 * memory management control operations buffer.
531 */
532 MMCO mmco[MAX_MMCO_COUNT];
533 int mmco_index;
534 int mmco_reset;
535
536 int long_ref_count; ///< number of actual long term references
537 int short_ref_count; ///< number of actual short term references
538
539 int cabac_init_idc;
540
541 /**
542 * @name Members for slice based multithreading
543 * @{
544 */
545 struct H264Context *thread_context[MAX_THREADS];
546
547 /**
548 * current slice number, used to initalize slice_num of each thread/context
549 */
550 int current_slice;
551
552 /**
553 * Max number of threads / contexts.
554 * This is equal to AVCodecContext.thread_count unless
555 * multithreaded decoding is impossible, in which case it is
556 * reduced to 1.
557 */
558 int max_contexts;
559
560 int slice_context_count;
561
562 /**
563 * 1 if the single thread fallback warning has already been
564 * displayed, 0 otherwise.
565 */
566 int single_decode_warning;
567
568 enum AVPictureType pict_type;
569
570 int last_slice_type;
571 /** @} */
572
573 /**
574 * pic_struct in picture timing SEI message
575 */
576 SEI_PicStructType sei_pic_struct;
577
578 /**
579 * Complement sei_pic_struct
580 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
581 * However, soft telecined frames may have these values.
582 * This is used in an attempt to flag soft telecine progressive.
583 */
584 int prev_interlaced_frame;
585
586 /**
587 * Bit set of clock types for fields/frames in picture timing SEI message.
588 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
589 * interlaced).
590 */
591 int sei_ct_type;
592
593 /**
594 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
595 */
596 int sei_dpb_output_delay;
597
598 /**
599 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
600 */
601 int sei_cpb_removal_delay;
602
603 /**
604 * recovery_frame_cnt from SEI message
605 *
606 * Set to -1 if no recovery point SEI message found or to number of frames
607 * before playback synchronizes. Frames having recovery point are key
608 * frames.
609 */
610 int sei_recovery_frame_cnt;
611
612 int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
613 int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
614
615 // Timestamp stuff
616 int sei_buffering_period_present; ///< Buffering period SEI flag
617 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
618
619 int cur_chroma_format_idc;
620 uint8_t *bipred_scratchpad;
621 uint8_t *edge_emu_buffer;
622 int16_t *dc_val_base;
623 } H264Context;
624
625 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).
626 extern const uint16_t ff_h264_mb_sizes[4];
627
628 /**
629 * Decode SEI
630 */
631 int ff_h264_decode_sei(H264Context *h);
632
633 /**
634 * Decode SPS
635 */
636 int ff_h264_decode_seq_parameter_set(H264Context *h);
637
638 /**
639 * compute profile from sps
640 */
641 int ff_h264_get_profile(SPS *sps);
642
643 /**
644 * Decode PPS
645 */
646 int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
647
648 /**
649 * Decode a network abstraction layer unit.
650 * @param consumed is the number of bytes used as input
651 * @param length is the length of the array
652 * @param dst_length is the number of decoded bytes FIXME here
653 * or a decode rbsp tailing?
654 * @return decoded bytes, might be src+1 if no escapes
655 */
656 const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
657 int *dst_length, int *consumed, int length);
658
659 /**
660 * Free any data that may have been allocated in the H264 context
661 * like SPS, PPS etc.
662 */
663 void ff_h264_free_context(H264Context *h);
664
665 /**
666 * Reconstruct bitstream slice_type.
667 */
668 int ff_h264_get_slice_type(const H264Context *h);
669
670 /**
671 * Allocate tables.
672 * needs width/height
673 */
674 int ff_h264_alloc_tables(H264Context *h);
675
676 /**
677 * Fill the default_ref_list.
678 */
679 int ff_h264_fill_default_ref_list(H264Context *h);
680
681 int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
682 void ff_h264_fill_mbaff_ref_list(H264Context *h);
683 void ff_h264_remove_all_refs(H264Context *h);
684
685 /**
686 * Execute the reference picture marking (memory management control operations).
687 */
688 int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
689
690 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
691 int first_slice);
692
693 int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
694
695 /**
696 * Check if the top & left blocks are available if needed & change the
697 * dc mode so it only uses the available blocks.
698 */
699 int ff_h264_check_intra4x4_pred_mode(H264Context *h);
700
701 /**
702 * Check if the top & left blocks are available if needed & change the
703 * dc mode so it only uses the available blocks.
704 */
705 int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
706
707 void ff_h264_hl_decode_mb(H264Context *h);
708 int ff_h264_frame_start(H264Context *h);
709 int ff_h264_decode_extradata(H264Context *h);
710 int ff_h264_decode_init(AVCodecContext *avctx);
711 void ff_h264_decode_init_vlc(void);
712
713 /**
714 * Decode a macroblock
715 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
716 */
717 int ff_h264_decode_mb_cavlc(H264Context *h);
718
719 /**
720 * Decode a CABAC coded macroblock
721 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
722 */
723 int ff_h264_decode_mb_cabac(H264Context *h);
724
725 void ff_h264_init_cabac_states(H264Context *h);
726
727 void ff_h264_direct_dist_scale_factor(H264Context *const h);
728 void ff_h264_direct_ref_list_init(H264Context *const h);
729 void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type);
730
731 void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y,
732 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
733 unsigned int linesize, unsigned int uvlinesize);
734 void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
735 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
736 unsigned int linesize, unsigned int uvlinesize);
737
738 /**
739 * Reset SEI values at the beginning of the frame.
740 *
741 * @param h H.264 context.
742 */
743 void ff_h264_reset_sei(H264Context *h);
744
745 /*
746 * o-o o-o
747 * / / /
748 * o-o o-o
749 * ,---'
750 * o-o o-o
751 * / / /
752 * o-o o-o
753 */
754
755 /* Scan8 organization:
756 * 0 1 2 3 4 5 6 7
757 * 0 DY y y y y y
758 * 1 y Y Y Y Y
759 * 2 y Y Y Y Y
760 * 3 y Y Y Y Y
761 * 4 y Y Y Y Y
762 * 5 DU u u u u u
763 * 6 u U U U U
764 * 7 u U U U U
765 * 8 u U U U U
766 * 9 u U U U U
767 * 10 DV v v v v v
768 * 11 v V V V V
769 * 12 v V V V V
770 * 13 v V V V V
771 * 14 v V V V V
772 * DY/DU/DV are for luma/chroma DC.
773 */
774
775 #define LUMA_DC_BLOCK_INDEX 48
776 #define CHROMA_DC_BLOCK_INDEX 49
777
778 // This table must be here because scan8[constant] must be known at compiletime
779 static const uint8_t scan8[16 * 3 + 3] = {
780 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
781 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
782 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
783 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
784 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
785 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
786 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
787 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
788 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
789 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
790 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
791 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
792 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
793 };
794
795 static av_always_inline uint32_t pack16to32(int a, int b)
796 {
797 #if HAVE_BIGENDIAN
798 return (b & 0xFFFF) + (a << 16);
799 #else
800 return (a & 0xFFFF) + (b << 16);
801 #endif
802 }
803
804 static av_always_inline uint16_t pack8to16(int a, int b)
805 {
806 #if HAVE_BIGENDIAN
807 return (b & 0xFF) + (a << 8);
808 #else
809 return (a & 0xFF) + (b << 8);
810 #endif
811 }
812
813 /**
814 * Get the chroma qp.
815 */
816 static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
817 {
818 return h->pps.chroma_qp_table[t][qscale];
819 }
820
821 /**
822 * Get the predicted intra4x4 prediction mode.
823 */
824 static av_always_inline int pred_intra_mode(H264Context *h, int n)
825 {
826 const int index8 = scan8[n];
827 const int left = h->intra4x4_pred_mode_cache[index8 - 1];
828 const int top = h->intra4x4_pred_mode_cache[index8 - 8];
829 const int min = FFMIN(left, top);
830
831 tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min);
832
833 if (min < 0)
834 return DC_PRED;
835 else
836 return min;
837 }
838
839 static av_always_inline void write_back_intra_pred_mode(H264Context *h)
840 {
841 int8_t *i4x4 = h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
842 int8_t *i4x4_cache = h->intra4x4_pred_mode_cache;
843
844 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
845 i4x4[4] = i4x4_cache[7 + 8 * 3];
846 i4x4[5] = i4x4_cache[7 + 8 * 2];
847 i4x4[6] = i4x4_cache[7 + 8 * 1];
848 }
849
850 static av_always_inline void write_back_non_zero_count(H264Context *h)
851 {
852 const int mb_xy = h->mb_xy;
853 uint8_t *nnz = h->non_zero_count[mb_xy];
854 uint8_t *nnz_cache = h->non_zero_count_cache;
855
856 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
857 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
858 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
859 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
860 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
861 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
862 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
863 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
864
865 if (!h->chroma_y_shift) {
866 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
867 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
868 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
869 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
870 }
871 }
872
873 static av_always_inline void write_back_motion_list(H264Context *h,
874 int b_stride,
875 int b_xy, int b8_xy,
876 int mb_type, int list)
877 {
878 int16_t(*mv_dst)[2] = &h->cur_pic.f.motion_val[list][b_xy];
879 int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
880 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
881 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
882 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
883 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
884 if (CABAC) {
885 uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8 * h->mb_xy
886 : h->mb2br_xy[h->mb_xy]];
887 uint8_t(*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
888 if (IS_SKIP(mb_type)) {
889 AV_ZERO128(mvd_dst);
890 } else {
891 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
892 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
893 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
894 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
895 }
896 }
897
898 {
899 int8_t *ref_index = &h->cur_pic.f.ref_index[list][b8_xy];
900 int8_t *ref_cache = h->ref_cache[list];
901 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
902 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
903 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
904 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
905 }
906 }
907
908 static av_always_inline void write_back_motion(H264Context *h, int mb_type)
909 {
910 const int b_stride = h->b_stride;
911 const int b_xy = 4 * h->mb_x + 4 * h->mb_y * h->b_stride; // try mb2b(8)_xy
912 const int b8_xy = 4 * h->mb_xy;
913
914 if (USES_LIST(mb_type, 0)) {
915 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 0);
916 } else {
917 fill_rectangle(&h->cur_pic.f.ref_index[0][b8_xy],
918 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
919 }
920 if (USES_LIST(mb_type, 1))
921 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 1);
922
923 if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC) {
924 if (IS_8X8(mb_type)) {
925 uint8_t *direct_table = &h->direct_table[4 * h->mb_xy];
926 direct_table[1] = h->sub_mb_type[1] >> 1;
927 direct_table[2] = h->sub_mb_type[2] >> 1;
928 direct_table[3] = h->sub_mb_type[3] >> 1;
929 }
930 }
931 }
932
933 static av_always_inline int get_dct8x8_allowed(H264Context *h)
934 {
935 if (h->sps.direct_8x8_inference_flag)
936 return !(AV_RN64A(h->sub_mb_type) &
937 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
938 0x0001000100010001ULL));
939 else
940 return !(AV_RN64A(h->sub_mb_type) &
941 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
942 0x0001000100010001ULL));
943 }
944
945 void ff_h264_draw_horiz_band(H264Context *h, int y, int height);
946
947 #endif /* AVCODEC_H264_H */