FFmpeg coverage


Directory: ../../../ffmpeg/
File: src/libavcodec/h264dec.h
Date: 2021-09-26 18:22:30
Exec Total Coverage
Lines: 87 88 98.9%
Branches: 16 20 80.0%

Line Branch Exec Source
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 / MPEG-4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28 #ifndef AVCODEC_H264DEC_H
29 #define AVCODEC_H264DEC_H
30
31 #include "libavutil/buffer.h"
32 #include "libavutil/intreadwrite.h"
33 #include "libavutil/mem_internal.h"
34 #include "libavutil/thread.h"
35
36 #include "cabac.h"
37 #include "error_resilience.h"
38 #include "h264_parse.h"
39 #include "h264_ps.h"
40 #include "h264_sei.h"
41 #include "h2645_parse.h"
42 #include "h264chroma.h"
43 #include "h264dsp.h"
44 #include "h264pred.h"
45 #include "h264qpel.h"
46 #include "h274.h"
47 #include "internal.h"
48 #include "mpegutils.h"
49 #include "parser.h"
50 #include "qpeldsp.h"
51 #include "rectangle.h"
52 #include "videodsp.h"
53
54 #define H264_MAX_PICTURE_COUNT 36
55
56 #define MAX_MMCO_COUNT 66
57
58 #define MAX_DELAYED_PIC_COUNT 16
59
60 /* Compiling in interlaced support reduces the speed
61 * of progressive decoding by about 2%. */
62 #define ALLOW_INTERLACE
63
64 #define FMO 0
65
66 /**
67 * The maximum number of slices supported by the decoder.
68 * must be a power of 2
69 */
70 #define MAX_SLICES 32
71
72 #ifdef ALLOW_INTERLACE
73 #define MB_MBAFF(h) (h)->mb_mbaff
74 #define MB_FIELD(sl) (sl)->mb_field_decoding_flag
75 #define FRAME_MBAFF(h) (h)->mb_aff_frame
76 #define FIELD_PICTURE(h) ((h)->picture_structure != PICT_FRAME)
77 #define LEFT_MBS 2
78 #define LTOP 0
79 #define LBOT 1
80 #define LEFT(i) (i)
81 #else
82 #define MB_MBAFF(h) 0
83 #define MB_FIELD(sl) 0
84 #define FRAME_MBAFF(h) 0
85 #define FIELD_PICTURE(h) 0
86 #undef IS_INTERLACED
87 #define IS_INTERLACED(mb_type) 0
88 #define LEFT_MBS 1
89 #define LTOP 0
90 #define LBOT 0
91 #define LEFT(i) 0
92 #endif
93 #define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
94
95 #ifndef CABAC
96 #define CABAC(h) (h)->ps.pps->cabac
97 #endif
98
99 #define CHROMA(h) ((h)->ps.sps->chroma_format_idc)
100 #define CHROMA422(h) ((h)->ps.sps->chroma_format_idc == 2)
101 #define CHROMA444(h) ((h)->ps.sps->chroma_format_idc == 3)
102
103 #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
104 #define MB_TYPE_8x8DCT 0x01000000
105 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
106 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
107
108 /**
109 * Memory management control operation opcode.
110 */
111 typedef enum MMCOOpcode {
112 MMCO_END = 0,
113 MMCO_SHORT2UNUSED,
114 MMCO_LONG2UNUSED,
115 MMCO_SHORT2LONG,
116 MMCO_SET_MAX_LONG,
117 MMCO_RESET,
118 MMCO_LONG,
119 } MMCOOpcode;
120
121 /**
122 * Memory management control operation.
123 */
124 typedef struct MMCO {
125 MMCOOpcode opcode;
126 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
127 int long_arg; ///< index, pic_num, or num long refs depending on opcode
128 } MMCO;
129
130 typedef struct H264Picture {
131 AVFrame *f;
132 ThreadFrame tf;
133
134 AVFrame *f_grain;
135 ThreadFrame tf_grain;
136
137 AVBufferRef *qscale_table_buf;
138 int8_t *qscale_table;
139
140 AVBufferRef *motion_val_buf[2];
141 int16_t (*motion_val[2])[2];
142
143 AVBufferRef *mb_type_buf;
144 uint32_t *mb_type;
145
146 AVBufferRef *hwaccel_priv_buf;
147 void *hwaccel_picture_private; ///< hardware accelerator private data
148
149 AVBufferRef *ref_index_buf[2];
150 int8_t *ref_index[2];
151
152 int field_poc[2]; ///< top/bottom POC
153 int poc; ///< frame POC
154 int frame_num; ///< frame_num (raw frame_num from slice header)
155 int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
156 not mix pictures before and after MMCO_RESET. */
157 int pic_id; /**< pic_num (short -> no wrap version of pic_num,
158 pic_num & max_pic_num; long -> long_pic_num) */
159 int long_ref; ///< 1->long term reference 0->short term reference
160 int ref_poc[2][2][32]; ///< POCs of the frames/fields used as reference (FIXME need per slice)
161 int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
162 int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
163 int field_picture; ///< whether or not picture was encoded in separate fields
164
165 int reference;
166 int recovered; ///< picture at IDR or recovery point + recovery count
167 int invalid_gap;
168 int sei_recovery_frame_cnt;
169 int needs_fg; ///< whether picture needs film grain synthesis (see `f_grain`)
170
171 AVBufferRef *pps_buf;
172 const PPS *pps;
173
174 int mb_width, mb_height;
175 int mb_stride;
176 } H264Picture;
177
178 typedef struct H264Ref {
179 uint8_t *data[3];
180 int linesize[3];
181
182 int reference;
183 int poc;
184 int pic_id;
185
186 H264Picture *parent;
187 } H264Ref;
188
189 typedef struct H264SliceContext {
190 struct H264Context *h264;
191 GetBitContext gb;
192 ERContext er;
193
194 int slice_num;
195 int slice_type;
196 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
197 int slice_type_fixed;
198
199 int qscale;
200 int chroma_qp[2]; // QPc
201 int qp_thresh; ///< QP threshold to skip loopfilter
202 int last_qscale_diff;
203
204 // deblock
205 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
206 int slice_alpha_c0_offset;
207 int slice_beta_offset;
208
209 H264PredWeightTable pwt;
210
211 int prev_mb_skipped;
212 int next_mb_skipped;
213
214 int chroma_pred_mode;
215 int intra16x16_pred_mode;
216
217 int8_t intra4x4_pred_mode_cache[5 * 8];
218 int8_t(*intra4x4_pred_mode);
219
220 int topleft_mb_xy;
221 int top_mb_xy;
222 int topright_mb_xy;
223 int left_mb_xy[LEFT_MBS];
224
225 int topleft_type;
226 int top_type;
227 int topright_type;
228 int left_type[LEFT_MBS];
229
230 const uint8_t *left_block;
231 int topleft_partition;
232
233 unsigned int topleft_samples_available;
234 unsigned int top_samples_available;
235 unsigned int topright_samples_available;
236 unsigned int left_samples_available;
237
238 ptrdiff_t linesize, uvlinesize;
239 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
240 ptrdiff_t mb_uvlinesize;
241
242 int mb_x, mb_y;
243 int mb_xy;
244 int resync_mb_x;
245 int resync_mb_y;
246 unsigned int first_mb_addr;
247 // index of the first MB of the next slice
248 int next_slice_idx;
249 int mb_skip_run;
250 int is_complex;
251
252 int picture_structure;
253 int mb_field_decoding_flag;
254 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
255
256 int redundant_pic_count;
257
258 /**
259 * number of neighbors (top and/or left) that used 8x8 dct
260 */
261 int neighbor_transform_size;
262
263 int direct_spatial_mv_pred;
264 int col_parity;
265 int col_fieldoff;
266
267 int cbp;
268 int top_cbp;
269 int left_cbp;
270
271 int dist_scale_factor[32];
272 int dist_scale_factor_field[2][32];
273 int map_col_to_list0[2][16 + 32];
274 int map_col_to_list0_field[2][2][16 + 32];
275
276 /**
277 * num_ref_idx_l0/1_active_minus1 + 1
278 */
279 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
280 unsigned int list_count;
281 H264Ref ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
282 * Reordered version of default_ref_list
283 * according to picture reordering in slice header */
284 struct {
285 uint8_t op;
286 uint32_t val;
287 } ref_modifications[2][32];
288 int nb_ref_modifications[2];
289
290 unsigned int pps_id;
291
292 const uint8_t *intra_pcm_ptr;
293 int16_t *dc_val_base;
294
295 uint8_t *bipred_scratchpad;
296 uint8_t *edge_emu_buffer;
297 uint8_t (*top_borders[2])[(16 * 3) * 2];
298 int bipred_scratchpad_allocated;
299 int edge_emu_buffer_allocated;
300 int top_borders_allocated[2];
301
302 /**
303 * non zero coeff count cache.
304 * is 64 if not available.
305 */
306 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
307
308 /**
309 * Motion vector cache.
310 */
311 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
312 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
313 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
314 uint8_t direct_cache[5 * 8];
315
316 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
317
318 ///< as a DCT coefficient is int32_t in high depth, we need to reserve twice the space.
319 DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2];
320 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
321 ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either
322 ///< check that i is not too large or ensure that there is some unused stuff after mb
323 int16_t mb_padding[256 * 2];
324
325 uint8_t (*mvd_table[2])[2];
326
327 /**
328 * Cabac
329 */
330 CABACContext cabac;
331 uint8_t cabac_state[1024];
332 int cabac_init_idc;
333
334 MMCO mmco[MAX_MMCO_COUNT];
335 int nb_mmco;
336 int explicit_ref_marking;
337
338 int frame_num;
339 int idr_pic_id;
340 int poc_lsb;
341 int delta_poc_bottom;
342 int delta_poc[2];
343 int curr_pic_num;
344 int max_pic_num;
345 } H264SliceContext;
346
347 /**
348 * H264Context
349 */
350 typedef struct H264Context {
351 const AVClass *class;
352 AVCodecContext *avctx;
353 VideoDSPContext vdsp;
354 H264DSPContext h264dsp;
355 H264ChromaContext h264chroma;
356 H264QpelContext h264qpel;
357 H274FilmGrainDatabase h274db;
358
359 H264Picture DPB[H264_MAX_PICTURE_COUNT];
360 H264Picture *cur_pic_ptr;
361 H264Picture cur_pic;
362 H264Picture last_pic_for_ec;
363
364 H264SliceContext *slice_ctx;
365 int nb_slice_ctx;
366 int nb_slice_ctx_queued;
367
368 H2645Packet pkt;
369
370 int pixel_shift; ///< 0 for 8-bit H.264, 1 for high-bit-depth H.264
371
372 /* coded dimensions -- 16 * mb w/h */
373 int width, height;
374 int chroma_x_shift, chroma_y_shift;
375
376 int droppable;
377 int coded_picture_number;
378
379 int context_initialized;
380 int flags;
381 int workaround_bugs;
382 int x264_build;
383 /* Set when slice threading is used and at least one slice uses deblocking
384 * mode 1 (i.e. across slice boundaries). Then we disable the loop filter
385 * during normal MB decoding and execute it serially at the end.
386 */
387 int postpone_filter;
388
389 /*
390 * Set to 1 when the current picture is IDR, 0 otherwise.
391 */
392 int picture_idr;
393
394 /*
395 * Set to 1 when the current picture contains only I slices, 0 otherwise.
396 */
397 int picture_intra_only;
398
399 int crop_left;
400 int crop_right;
401 int crop_top;
402 int crop_bottom;
403
404 int8_t(*intra4x4_pred_mode);
405 H264PredContext hpc;
406
407 uint8_t (*non_zero_count)[48];
408
409 #define LIST_NOT_USED -1 // FIXME rename?
410 #define PART_NOT_AVAILABLE -2
411
412 /**
413 * block_offset[ 0..23] for frame macroblocks
414 * block_offset[24..47] for field macroblocks
415 */
416 int block_offset[2 * (16 * 3)];
417
418 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
419 uint32_t *mb2br_xy;
420 int b_stride; // FIXME use s->b4_stride
421
422 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
423
424 // interlacing specific flags
425 int mb_aff_frame;
426 int picture_structure;
427 int first_field;
428
429 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
430
431 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
432 uint16_t *cbp_table;
433
434 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
435 uint8_t *chroma_pred_mode_table;
436 uint8_t (*mvd_table[2])[2];
437 uint8_t *direct_table;
438
439 uint8_t scan_padding[16];
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 uint8_t zigzag_scan_q0[16];
447 uint8_t zigzag_scan8x8_q0[64];
448 uint8_t zigzag_scan8x8_cavlc_q0[64];
449 uint8_t field_scan_q0[16];
450 uint8_t field_scan8x8_q0[64];
451 uint8_t field_scan8x8_cavlc_q0[64];
452
453 int mb_y;
454 int mb_height, mb_width;
455 int mb_stride;
456 int mb_num;
457
458 // =============================================================
459 // Things below are not used in the MB or more inner code
460
461 int nal_ref_idc;
462 int nal_unit_type;
463
464 int has_slice; ///< slice NAL is found in the packet, set by decode_nal_units, its state does not need to be preserved outside h264_decode_frame()
465
466 /**
467 * Used to parse AVC variant of H.264
468 */
469 int is_avc; ///< this flag is != 0 if codec is avc1
470 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
471
472 int bit_depth_luma; ///< luma bit depth from sps to detect changes
473 int chroma_format_idc; ///< chroma format from sps to detect changes
474
475 H264ParamSets ps;
476
477 uint16_t *slice_table_base;
478
479 H264POCContext poc;
480
481 H264Ref default_ref[2];
482 H264Picture *short_ref[32];
483 H264Picture *long_ref[32];
484 H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
485 int last_pocs[MAX_DELAYED_PIC_COUNT];
486 H264Picture *next_output_pic;
487 int next_outputed_poc;
488 int poc_offset; ///< PicOrderCnt_offset from SMPTE RDD-2006
489
490 /**
491 * memory management control operations buffer.
492 */
493 MMCO mmco[MAX_MMCO_COUNT];
494 int nb_mmco;
495 int mmco_reset;
496 int explicit_ref_marking;
497
498 int long_ref_count; ///< number of actual long term references
499 int short_ref_count; ///< number of actual short term references
500
501 /**
502 * @name Members for slice based multithreading
503 * @{
504 */
505 /**
506 * current slice number, used to initialize slice_num of each thread/context
507 */
508 int current_slice;
509
510 /** @} */
511
512 /**
513 * Complement sei_pic_struct
514 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
515 * However, soft telecined frames may have these values.
516 * This is used in an attempt to flag soft telecine progressive.
517 */
518 int prev_interlaced_frame;
519
520 /**
521 * Are the SEI recovery points looking valid.
522 */
523 int valid_recovery_point;
524
525 /**
526 * recovery_frame is the frame_num at which the next frame should
527 * be fully constructed.
528 *
529 * Set to -1 when not expecting a recovery point.
530 */
531 int recovery_frame;
532
533 /**
534 * We have seen an IDR, so all the following frames in coded order are correctly
535 * decodable.
536 */
537 #define FRAME_RECOVERED_IDR (1 << 0)
538 /**
539 * Sufficient number of frames have been decoded since a SEI recovery point,
540 * so all the following frames in presentation order are correct.
541 */
542 #define FRAME_RECOVERED_SEI (1 << 1)
543
544 int frame_recovered; ///< Initial frame has been completely recovered
545
546 int has_recovery_point;
547
548 int missing_fields;
549
550 /* for frame threading, this is set to 1
551 * after finish_setup() has been called, so we cannot modify
552 * some context properties (which are supposed to stay constant between
553 * slices) anymore */
554 int setup_finished;
555
556 int cur_chroma_format_idc;
557 int cur_bit_depth_luma;
558 int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
559
560 /* original AVCodecContext dimensions, used to handle container
561 * cropping */
562 int width_from_caller;
563 int height_from_caller;
564
565 int enable_er;
566
567 H264SEIContext sei;
568
569 AVBufferPool *qscale_table_pool;
570 AVBufferPool *mb_type_pool;
571 AVBufferPool *motion_val_pool;
572 AVBufferPool *ref_index_pool;
573 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
574 } H264Context;
575
576 extern const uint16_t ff_h264_mb_sizes[4];
577
578 /**
579 * Reconstruct bitstream slice_type.
580 */
581 int ff_h264_get_slice_type(const H264SliceContext *sl);
582
583 /**
584 * Allocate tables.
585 * needs width/height
586 */
587 int ff_h264_alloc_tables(H264Context *h);
588
589 int ff_h264_decode_ref_pic_list_reordering(H264SliceContext *sl, void *logctx);
590 int ff_h264_build_ref_list(H264Context *h, H264SliceContext *sl);
591 void ff_h264_remove_all_refs(H264Context *h);
592
593 /**
594 * Execute the reference picture marking (memory management control operations).
595 */
596 int ff_h264_execute_ref_pic_marking(H264Context *h);
597
598 int ff_h264_decode_ref_pic_marking(H264SliceContext *sl, GetBitContext *gb,
599 const H2645NAL *nal, void *logctx);
600
601 void ff_h264_hl_decode_mb(const H264Context *h, H264SliceContext *sl);
602 void ff_h264_decode_init_vlc(void);
603
604 /**
605 * Decode a macroblock
606 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
607 */
608 int ff_h264_decode_mb_cavlc(const H264Context *h, H264SliceContext *sl);
609
610 /**
611 * Decode a CABAC coded macroblock
612 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
613 */
614 int ff_h264_decode_mb_cabac(const H264Context *h, H264SliceContext *sl);
615
616 void ff_h264_init_cabac_states(const H264Context *h, H264SliceContext *sl);
617
618 void ff_h264_direct_dist_scale_factor(const H264Context *const h, H264SliceContext *sl);
619 void ff_h264_direct_ref_list_init(const H264Context *const h, H264SliceContext *sl);
620 void ff_h264_pred_direct_motion(const H264Context *const h, H264SliceContext *sl,
621 int *mb_type);
622
623 void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
624 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
625 unsigned int linesize, unsigned int uvlinesize);
626 void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
627 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
628 unsigned int linesize, unsigned int uvlinesize);
629
630 /*
631 * o-o o-o
632 * / / /
633 * o-o o-o
634 * ,---'
635 * o-o o-o
636 * / / /
637 * o-o o-o
638 */
639
640 /* Scan8 organization:
641 * 0 1 2 3 4 5 6 7
642 * 0 DY y y y y y
643 * 1 y Y Y Y Y
644 * 2 y Y Y Y Y
645 * 3 y Y Y Y Y
646 * 4 y Y Y Y Y
647 * 5 DU u u u u u
648 * 6 u U U U U
649 * 7 u U U U U
650 * 8 u U U U U
651 * 9 u U U U U
652 * 10 DV v v v v v
653 * 11 v V V V V
654 * 12 v V V V V
655 * 13 v V V V V
656 * 14 v V V V V
657 * DY/DU/DV are for luma/chroma DC.
658 */
659
660 #define LUMA_DC_BLOCK_INDEX 48
661 #define CHROMA_DC_BLOCK_INDEX 49
662
663 // This table must be here because scan8[constant] must be known at compiletime
664 static const uint8_t scan8[16 * 3 + 3] = {
665 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
666 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
667 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
668 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
669 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
670 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
671 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
672 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
673 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
674 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
675 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
676 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
677 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
678 };
679
680 34269250 static av_always_inline uint32_t pack16to32(unsigned a, unsigned b)
681 {
682 #if HAVE_BIGENDIAN
683 return (b & 0xFFFF) + (a << 16);
684 #else
685 34269250 return (a & 0xFFFF) + (b << 16);
686 #endif
687 }
688
689 3897119 static av_always_inline uint16_t pack8to16(unsigned a, unsigned b)
690 {
691 #if HAVE_BIGENDIAN
692 return (b & 0xFF) + (a << 8);
693 #else
694 3897119 return (a & 0xFF) + (b << 8);
695 #endif
696 }
697
698 /**
699 * Get the chroma qp.
700 */
701 60081759 static av_always_inline int get_chroma_qp(const PPS *pps, int t, int qscale)
702 {
703 60081759 return pps->chroma_qp_table[t][qscale];
704 }
705
706 /**
707 * Get the predicted intra4x4 prediction mode.
708 */
709 29176592 static av_always_inline int pred_intra_mode(const H264Context *h,
710 H264SliceContext *sl, int n)
711 {
712 29176592 const int index8 = scan8[n];
713 29176592 const int left = sl->intra4x4_pred_mode_cache[index8 - 1];
714 29176592 const int top = sl->intra4x4_pred_mode_cache[index8 - 8];
715 29176592 const int min = FFMIN(left, top);
716
717 ff_tlog(h->avctx, "mode:%d %d min:%d\n", left, top, min);
718
719
2/2
✓ Branch 0 taken 818542 times.
✓ Branch 1 taken 28358050 times.
29176592 if (min < 0)
720 818542 return DC_PRED;
721 else
722 28358050 return min;
723 }
724
725 2956505 static av_always_inline void write_back_intra_pred_mode(const H264Context *h,
726 H264SliceContext *sl)
727 {
728 2956505 int8_t *i4x4 = sl->intra4x4_pred_mode + h->mb2br_xy[sl->mb_xy];
729 2956505 int8_t *i4x4_cache = sl->intra4x4_pred_mode_cache;
730
731 2956505 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
732 2956505 i4x4[4] = i4x4_cache[7 + 8 * 3];
733 2956505 i4x4[5] = i4x4_cache[7 + 8 * 2];
734 2956505 i4x4[6] = i4x4_cache[7 + 8 * 1];
735 2956505 }
736
737 10137017 static av_always_inline void write_back_non_zero_count(const H264Context *h,
738 H264SliceContext *sl)
739 {
740 10137017 const int mb_xy = sl->mb_xy;
741 10137017 uint8_t *nnz = h->non_zero_count[mb_xy];
742 10137017 uint8_t *nnz_cache = sl->non_zero_count_cache;
743
744 10137017 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
745 10137017 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
746 10137017 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
747 10137017 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
748 10137017 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
749 10137017 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
750 10137017 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
751 10137017 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
752
753
2/2
✓ Branch 0 taken 958595 times.
✓ Branch 1 taken 9178422 times.
10137017 if (!h->chroma_y_shift) {
754 958595 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
755 958595 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
756 958595 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
757 958595 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
758 }
759 10137017 }
760
761 13330709 static av_always_inline void write_back_motion_list(const H264Context *h,
762 H264SliceContext *sl,
763 int b_stride,
764 int b_xy, int b8_xy,
765 int mb_type, int list)
766 {
767 13330709 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
768 13330709 int16_t(*mv_src)[2] = &sl->mv_cache[list][scan8[0]];
769 13330709 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
770 13330709 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
771 13330709 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
772 13330709 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
773 if (CABAC(h)) {
774 9758096 uint8_t (*mvd_dst)[2] = &sl->mvd_table[list][FMO ? 8 * sl->mb_xy
775 9758096 : h->mb2br_xy[sl->mb_xy]];
776 9758096 uint8_t(*mvd_src)[2] = &sl->mvd_cache[list][scan8[0]];
777
2/2
✓ Branch 0 taken 4383451 times.
✓ Branch 1 taken 5374645 times.
9758096 if (IS_SKIP(mb_type)) {
778 4383451 AV_ZERO128(mvd_dst);
779 } else {
780 5374645 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
781 5374645 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
782 5374645 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
783 5374645 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
784 }
785 }
786
787 {
788 13330709 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
789 13330709 int8_t *ref_cache = sl->ref_cache[list];
790 13330709 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
791 13330709 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
792 13330709 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
793 13330709 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
794 }
795 13330709 }
796
797 9861704 static av_always_inline void write_back_motion(const H264Context *h,
798 H264SliceContext *sl,
799 int mb_type)
800 {
801 9861704 const int b_stride = h->b_stride;
802 9861704 const int b_xy = 4 * sl->mb_x + 4 * sl->mb_y * h->b_stride; // try mb2b(8)_xy
803 9861704 const int b8_xy = 4 * sl->mb_xy;
804
805
2/2
✓ Branch 0 taken 9140929 times.
✓ Branch 1 taken 720775 times.
9861704 if (USES_LIST(mb_type, 0)) {
806 9140929 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 0);
807 } else {
808 720775 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
809 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
810 }
811
2/2
✓ Branch 0 taken 4189780 times.
✓ Branch 1 taken 5671924 times.
9861704 if (USES_LIST(mb_type, 1))
812 4189780 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 1);
813
814
2/4
✓ Branch 0 taken 4129859 times.
✓ Branch 1 taken 2771880 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
6901739 if (sl->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
815
2/2
✓ Branch 0 taken 415008 times.
✓ Branch 1 taken 3714851 times.
4129859 if (IS_8X8(mb_type)) {
816 415008 uint8_t *direct_table = &h->direct_table[4 * sl->mb_xy];
817 415008 direct_table[1] = sl->sub_mb_type[1] >> 1;
818 415008 direct_table[2] = sl->sub_mb_type[2] >> 1;
819 415008 direct_table[3] = sl->sub_mb_type[3] >> 1;
820 }
821 }
822 9861704 }
823
824 265181 static av_always_inline int get_dct8x8_allowed(const H264Context *h, H264SliceContext *sl)
825 {
826
2/2
✓ Branch 0 taken 260304 times.
✓ Branch 1 taken 4877 times.
265181 if (h->ps.sps->direct_8x8_inference_flag)
827 260304 return !(AV_RN64A(sl->sub_mb_type) &
828 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
829 0x0001000100010001ULL));
830 else
831 4877 return !(AV_RN64A(sl->sub_mb_type) &
832 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
833 0x0001000100010001ULL));
834 }
835
836 36351 static inline int find_start_code(const uint8_t *buf, int buf_size,
837 int buf_index, int next_avc)
838 {
839 36351 uint32_t state = -1;
840
841 36351 buf_index = avpriv_find_start_code(buf + buf_index, buf + next_avc + 1, &state) - buf - 1;
842
843 36351 return FFMIN(buf_index, buf_size);
844 }
845
846 int ff_h264_field_end(H264Context *h, H264SliceContext *sl, int in_setup);
847
848 int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
849 int ff_h264_replace_picture(H264Context *h, H264Picture *dst, const H264Picture *src);
850 void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
851
852 int ff_h264_slice_context_init(H264Context *h, H264SliceContext *sl);
853
854 void ff_h264_draw_horiz_band(const H264Context *h, H264SliceContext *sl, int y, int height);
855
856 /**
857 * Submit a slice for decoding.
858 *
859 * Parse the slice header, starting a new field/frame if necessary. If any
860 * slices are queued for the previous field, they are decoded.
861 */
862 int ff_h264_queue_decode_slice(H264Context *h, const H2645NAL *nal);
863 int ff_h264_execute_decode_slices(H264Context *h);
864 int ff_h264_update_thread_context(AVCodecContext *dst,
865 const AVCodecContext *src);
866 int ff_h264_update_thread_context_for_user(AVCodecContext *dst,
867 const AVCodecContext *src);
868
869 void ff_h264_flush_change(H264Context *h);
870
871 void ff_h264_free_tables(H264Context *h);
872
873 void ff_h264_set_erpic(ERPicture *dst, H264Picture *src);
874
875 #endif /* AVCODEC_H264DEC_H */
876