FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/libavcodec/h264dec.h
Date: 2023-03-22 23:59:29
Exec Total Coverage
Lines: 79 79 100.0%
Functions: 7 7 100.0%
Branches: 16 16 100.0%
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
35 #include "cabac.h"
36 #include "error_resilience.h"
37 #include "h264_parse.h"
38 #include "h264_ps.h"
39 #include "h264_sei.h"
40 #include "h2645_parse.h"
41 #include "h264chroma.h"
42 #include "h264dsp.h"
43 #include "h264pred.h"
44 #include "h264qpel.h"
45 #include "h274.h"
46 #include "mpegutils.h"
47 #include "rectangle.h"
48 #include "videodsp.h"
49
50 #define H264_MAX_PICTURE_COUNT 36
51
52 /* Compiling in interlaced support reduces the speed
53 * of progressive decoding by about 2%. */
54 #define ALLOW_INTERLACE
55
56 #define FMO 0
57
58 /**
59 * The maximum number of slices supported by the decoder.
60 * must be a power of 2
61 */
62 #define MAX_SLICES 32
63
64 #ifdef ALLOW_INTERLACE
65 #define MB_MBAFF(h) (h)->mb_mbaff
66 #define MB_FIELD(sl) (sl)->mb_field_decoding_flag
67 #define FRAME_MBAFF(h) (h)->mb_aff_frame
68 #define FIELD_PICTURE(h) ((h)->picture_structure != PICT_FRAME)
69 #define LEFT_MBS 2
70 #define LTOP 0
71 #define LBOT 1
72 #define LEFT(i) (i)
73 #else
74 #define MB_MBAFF(h) 0
75 #define MB_FIELD(sl) 0
76 #define FRAME_MBAFF(h) 0
77 #define FIELD_PICTURE(h) 0
78 #undef IS_INTERLACED
79 #define IS_INTERLACED(mb_type) 0
80 #define LEFT_MBS 1
81 #define LTOP 0
82 #define LBOT 0
83 #define LEFT(i) 0
84 #endif
85 #define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
86
87 #ifndef CABAC
88 #define CABAC(h) (h)->ps.pps->cabac
89 #endif
90
91 #define CHROMA(h) ((h)->ps.sps->chroma_format_idc)
92 #define CHROMA422(h) ((h)->ps.sps->chroma_format_idc == 2)
93 #define CHROMA444(h) ((h)->ps.sps->chroma_format_idc == 3)
94
95 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
96 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
97
98 /**
99 * Memory management control operation.
100 */
101 typedef struct MMCO {
102 MMCOOpcode opcode;
103 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
104 int long_arg; ///< index, pic_num, or num long refs depending on opcode
105 } MMCO;
106
107 typedef struct H264Picture {
108 AVFrame *f;
109 ThreadFrame tf;
110
111 AVFrame *f_grain;
112
113 AVBufferRef *qscale_table_buf;
114 int8_t *qscale_table;
115
116 AVBufferRef *motion_val_buf[2];
117 int16_t (*motion_val[2])[2];
118
119 AVBufferRef *mb_type_buf;
120 uint32_t *mb_type;
121
122 AVBufferRef *hwaccel_priv_buf;
123 void *hwaccel_picture_private; ///< hardware accelerator private data
124
125 AVBufferRef *ref_index_buf[2];
126 int8_t *ref_index[2];
127
128 int field_poc[2]; ///< top/bottom POC
129 int poc; ///< frame POC
130 int frame_num; ///< frame_num (raw frame_num from slice header)
131 int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
132 not mix pictures before and after MMCO_RESET. */
133 int pic_id; /**< pic_num (short -> no wrap version of pic_num,
134 pic_num & max_pic_num; long -> long_pic_num) */
135 int long_ref; ///< 1->long term reference 0->short term reference
136 int ref_poc[2][2][32]; ///< POCs of the frames/fields used as reference (FIXME need per slice)
137 int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
138 int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
139 int field_picture; ///< whether or not picture was encoded in separate fields
140
141 /**
142 * H264Picture.reference has this flag set,
143 * when the picture is held for delayed output.
144 */
145 #define DELAYED_PIC_REF (1 << 2)
146 int reference;
147 int recovered; ///< picture at IDR or recovery point + recovery count
148 int invalid_gap;
149 int sei_recovery_frame_cnt;
150 int needs_fg; ///< whether picture needs film grain synthesis (see `f_grain`)
151
152 AVBufferRef *pps_buf;
153 const PPS *pps;
154
155 int mb_width, mb_height;
156 int mb_stride;
157 } H264Picture;
158
159 typedef struct H264Ref {
160 uint8_t *data[3];
161 int linesize[3];
162
163 int reference;
164 int poc;
165 int pic_id;
166
167 H264Picture *parent;
168 } H264Ref;
169
170 typedef struct H264SliceContext {
171 const struct H264Context *h264;
172 GetBitContext gb;
173 ERContext *er;
174
175 int slice_num;
176 int slice_type;
177 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
178 int slice_type_fixed;
179
180 int qscale;
181 int chroma_qp[2]; // QPc
182 int qp_thresh; ///< QP threshold to skip loopfilter
183 int last_qscale_diff;
184
185 // deblock
186 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
187 int slice_alpha_c0_offset;
188 int slice_beta_offset;
189
190 H264PredWeightTable pwt;
191
192 int prev_mb_skipped;
193 int next_mb_skipped;
194
195 int chroma_pred_mode;
196 int intra16x16_pred_mode;
197
198 int8_t intra4x4_pred_mode_cache[5 * 8];
199 int8_t(*intra4x4_pred_mode);
200
201 int topleft_mb_xy;
202 int top_mb_xy;
203 int topright_mb_xy;
204 int left_mb_xy[LEFT_MBS];
205
206 int topleft_type;
207 int top_type;
208 int topright_type;
209 int left_type[LEFT_MBS];
210
211 const uint8_t *left_block;
212 int topleft_partition;
213
214 unsigned int topleft_samples_available;
215 unsigned int top_samples_available;
216 unsigned int topright_samples_available;
217 unsigned int left_samples_available;
218
219 ptrdiff_t linesize, uvlinesize;
220 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
221 ptrdiff_t mb_uvlinesize;
222
223 int mb_x, mb_y;
224 int mb_xy;
225 int resync_mb_x;
226 int resync_mb_y;
227 unsigned int first_mb_addr;
228 // index of the first MB of the next slice
229 int next_slice_idx;
230 int mb_skip_run;
231 int is_complex;
232
233 int picture_structure;
234 int mb_field_decoding_flag;
235 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
236
237 int redundant_pic_count;
238
239 /**
240 * number of neighbors (top and/or left) that used 8x8 dct
241 */
242 int neighbor_transform_size;
243
244 int direct_spatial_mv_pred;
245 int col_parity;
246 int col_fieldoff;
247
248 int cbp;
249 int top_cbp;
250 int left_cbp;
251
252 int dist_scale_factor[32];
253 int dist_scale_factor_field[2][32];
254 int map_col_to_list0[2][16 + 32];
255 int map_col_to_list0_field[2][2][16 + 32];
256
257 /**
258 * num_ref_idx_l0/1_active_minus1 + 1
259 */
260 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
261 unsigned int list_count;
262 H264Ref ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
263 * Reordered version of default_ref_list
264 * according to picture reordering in slice header */
265 struct {
266 uint8_t op;
267 uint32_t val;
268 } ref_modifications[2][32];
269 int nb_ref_modifications[2];
270
271 unsigned int pps_id;
272
273 const uint8_t *intra_pcm_ptr;
274
275 uint8_t *bipred_scratchpad;
276 uint8_t *edge_emu_buffer;
277 uint8_t (*top_borders[2])[(16 * 3) * 2];
278 int bipred_scratchpad_allocated;
279 int edge_emu_buffer_allocated;
280 int top_borders_allocated[2];
281
282 /**
283 * non zero coeff count cache.
284 * is 64 if not available.
285 */
286 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
287
288 /**
289 * Motion vector cache.
290 */
291 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
292 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
293 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
294 uint8_t direct_cache[5 * 8];
295
296 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
297
298 ///< as a DCT coefficient is int32_t in high depth, we need to reserve twice the space.
299 DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2];
300 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
301 ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either
302 ///< check that i is not too large or ensure that there is some unused stuff after mb
303 int16_t mb_padding[256 * 2];
304
305 uint8_t (*mvd_table[2])[2];
306
307 /**
308 * Cabac
309 */
310 CABACContext cabac;
311 uint8_t cabac_state[1024];
312 int cabac_init_idc;
313
314 MMCO mmco[H264_MAX_MMCO_COUNT];
315 int nb_mmco;
316 int explicit_ref_marking;
317
318 int frame_num;
319 int idr_pic_id;
320 int poc_lsb;
321 int delta_poc_bottom;
322 int delta_poc[2];
323 int curr_pic_num;
324 int max_pic_num;
325 } H264SliceContext;
326
327 /**
328 * H264Context
329 */
330 typedef struct H264Context {
331 const AVClass *class;
332 AVCodecContext *avctx;
333 VideoDSPContext vdsp;
334 H264DSPContext h264dsp;
335 H264ChromaContext h264chroma;
336 H264QpelContext h264qpel;
337 H274FilmGrainDatabase h274db;
338
339 H264Picture DPB[H264_MAX_PICTURE_COUNT];
340 H264Picture *cur_pic_ptr;
341 H264Picture cur_pic;
342 H264Picture last_pic_for_ec;
343
344 H264SliceContext *slice_ctx;
345 int nb_slice_ctx;
346 int nb_slice_ctx_queued;
347
348 H2645Packet pkt;
349
350 int pixel_shift; ///< 0 for 8-bit H.264, 1 for high-bit-depth H.264
351
352 /* coded dimensions -- 16 * mb w/h */
353 int width, height;
354 int chroma_x_shift, chroma_y_shift;
355
356 int droppable;
357 int coded_picture_number;
358
359 int context_initialized;
360 int flags;
361 int workaround_bugs;
362 int x264_build;
363 /* Set when slice threading is used and at least one slice uses deblocking
364 * mode 1 (i.e. across slice boundaries). Then we disable the loop filter
365 * during normal MB decoding and execute it serially at the end.
366 */
367 int postpone_filter;
368
369 /*
370 * Set to 1 when the current picture is IDR, 0 otherwise.
371 */
372 int picture_idr;
373
374 /*
375 * Set to 1 when the current picture contains only I slices, 0 otherwise.
376 */
377 int picture_intra_only;
378
379 int crop_left;
380 int crop_right;
381 int crop_top;
382 int crop_bottom;
383
384 int8_t(*intra4x4_pred_mode);
385 H264PredContext hpc;
386
387 uint8_t (*non_zero_count)[48];
388
389 #define LIST_NOT_USED -1 // FIXME rename?
390
391 /**
392 * block_offset[ 0..23] for frame macroblocks
393 * block_offset[24..47] for field macroblocks
394 */
395 int block_offset[2 * (16 * 3)];
396
397 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
398 uint32_t *mb2br_xy;
399 int b_stride; // FIXME use s->b4_stride
400
401 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
402
403 // interlacing specific flags
404 int mb_aff_frame;
405 int picture_structure;
406 int first_field;
407
408 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
409
410 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
411 uint16_t *cbp_table;
412
413 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
414 uint8_t *chroma_pred_mode_table;
415 uint8_t (*mvd_table[2])[2];
416 uint8_t *direct_table;
417
418 uint8_t scan_padding[16];
419 uint8_t zigzag_scan[16];
420 uint8_t zigzag_scan8x8[64];
421 uint8_t zigzag_scan8x8_cavlc[64];
422 uint8_t field_scan[16];
423 uint8_t field_scan8x8[64];
424 uint8_t field_scan8x8_cavlc[64];
425 uint8_t zigzag_scan_q0[16];
426 uint8_t zigzag_scan8x8_q0[64];
427 uint8_t zigzag_scan8x8_cavlc_q0[64];
428 uint8_t field_scan_q0[16];
429 uint8_t field_scan8x8_q0[64];
430 uint8_t field_scan8x8_cavlc_q0[64];
431
432 int mb_y;
433 int mb_height, mb_width;
434 int mb_stride;
435 int mb_num;
436
437 // =============================================================
438 // Things below are not used in the MB or more inner code
439
440 int nal_ref_idc;
441 int nal_unit_type;
442
443 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()
444
445 /**
446 * Used to parse AVC variant of H.264
447 */
448 int is_avc; ///< this flag is != 0 if codec is avc1
449 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
450
451 int bit_depth_luma; ///< luma bit depth from sps to detect changes
452 int chroma_format_idc; ///< chroma format from sps to detect changes
453
454 H264ParamSets ps;
455
456 uint16_t *slice_table_base;
457
458 H264POCContext poc;
459
460 H264Ref default_ref[2];
461 H264Picture *short_ref[32];
462 H264Picture *long_ref[32];
463 H264Picture *delayed_pic[H264_MAX_DPB_FRAMES + 2]; // FIXME size?
464 int last_pocs[H264_MAX_DPB_FRAMES];
465 H264Picture *next_output_pic;
466 int next_outputed_poc;
467 int poc_offset; ///< PicOrderCnt_offset from SMPTE RDD-2006
468
469 /**
470 * memory management control operations buffer.
471 */
472 MMCO mmco[H264_MAX_MMCO_COUNT];
473 int nb_mmco;
474 int mmco_reset;
475 int explicit_ref_marking;
476
477 int long_ref_count; ///< number of actual long term references
478 int short_ref_count; ///< number of actual short term references
479
480 /**
481 * @name Members for slice based multithreading
482 * @{
483 */
484 /**
485 * current slice number, used to initialize slice_num of each thread/context
486 */
487 int current_slice;
488
489 /** @} */
490
491 /**
492 * Complement sei_pic_struct
493 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
494 * However, soft telecined frames may have these values.
495 * This is used in an attempt to flag soft telecine progressive.
496 */
497 int prev_interlaced_frame;
498
499 /**
500 * Are the SEI recovery points looking valid.
501 */
502 int valid_recovery_point;
503
504 /**
505 * recovery_frame is the frame_num at which the next frame should
506 * be fully constructed.
507 *
508 * Set to -1 when not expecting a recovery point.
509 */
510 int recovery_frame;
511
512 /**
513 * We have seen an IDR, so all the following frames in coded order are correctly
514 * decodable.
515 */
516 #define FRAME_RECOVERED_IDR (1 << 0)
517 /**
518 * Sufficient number of frames have been decoded since a SEI recovery point,
519 * so all the following frames in presentation order are correct.
520 */
521 #define FRAME_RECOVERED_SEI (1 << 1)
522
523 int frame_recovered; ///< Initial frame has been completely recovered
524
525 int has_recovery_point;
526
527 int missing_fields;
528
529 /* for frame threading, this is set to 1
530 * after finish_setup() has been called, so we cannot modify
531 * some context properties (which are supposed to stay constant between
532 * slices) anymore */
533 int setup_finished;
534
535 int cur_chroma_format_idc;
536 int cur_bit_depth_luma;
537 int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
538
539 /* original AVCodecContext dimensions, used to handle container
540 * cropping */
541 int width_from_caller;
542 int height_from_caller;
543
544 int enable_er;
545 ERContext er;
546 int16_t *dc_val_base;
547
548 H264SEIContext sei;
549
550 AVBufferPool *qscale_table_pool;
551 AVBufferPool *mb_type_pool;
552 AVBufferPool *motion_val_pool;
553 AVBufferPool *ref_index_pool;
554 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
555 } H264Context;
556
557 extern const uint16_t ff_h264_mb_sizes[4];
558
559 /**
560 * Reconstruct bitstream slice_type.
561 */
562 int ff_h264_get_slice_type(const H264SliceContext *sl);
563
564 /**
565 * Allocate tables.
566 * needs width/height
567 */
568 int ff_h264_alloc_tables(H264Context *h);
569
570 int ff_h264_decode_ref_pic_list_reordering(H264SliceContext *sl, void *logctx);
571 int ff_h264_build_ref_list(H264Context *h, H264SliceContext *sl);
572 void ff_h264_remove_all_refs(H264Context *h);
573
574 /**
575 * Execute the reference picture marking (memory management control operations).
576 */
577 int ff_h264_execute_ref_pic_marking(H264Context *h);
578
579 int ff_h264_decode_ref_pic_marking(H264SliceContext *sl, GetBitContext *gb,
580 const H2645NAL *nal, void *logctx);
581
582 void ff_h264_hl_decode_mb(const H264Context *h, H264SliceContext *sl);
583 void ff_h264_decode_init_vlc(void);
584
585 /**
586 * Decode a macroblock
587 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
588 */
589 int ff_h264_decode_mb_cavlc(const H264Context *h, H264SliceContext *sl);
590
591 /**
592 * Decode a CABAC coded macroblock
593 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
594 */
595 int ff_h264_decode_mb_cabac(const H264Context *h, H264SliceContext *sl);
596
597 void ff_h264_init_cabac_states(const H264Context *h, H264SliceContext *sl);
598
599 void ff_h264_direct_dist_scale_factor(const H264Context *const h, H264SliceContext *sl);
600 void ff_h264_direct_ref_list_init(const H264Context *const h, H264SliceContext *sl);
601 void ff_h264_pred_direct_motion(const H264Context *const h, H264SliceContext *sl,
602 int *mb_type);
603
604 void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
605 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
606 unsigned int linesize, unsigned int uvlinesize);
607 void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
608 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
609 unsigned int linesize, unsigned int uvlinesize);
610
611 /*
612 * o-o o-o
613 * / / /
614 * o-o o-o
615 * ,---'
616 * o-o o-o
617 * / / /
618 * o-o o-o
619 */
620
621 /* Scan8 organization:
622 * 0 1 2 3 4 5 6 7
623 * 0 DY y y y y y
624 * 1 y Y Y Y Y
625 * 2 y Y Y Y Y
626 * 3 y Y Y Y Y
627 * 4 y Y Y Y Y
628 * 5 DU u u u u u
629 * 6 u U U U U
630 * 7 u U U U U
631 * 8 u U U U U
632 * 9 u U U U U
633 * 10 DV v v v v v
634 * 11 v V V V V
635 * 12 v V V V V
636 * 13 v V V V V
637 * 14 v V V V V
638 * DY/DU/DV are for luma/chroma DC.
639 */
640
641 #define LUMA_DC_BLOCK_INDEX 48
642 #define CHROMA_DC_BLOCK_INDEX 49
643
644 /**
645 * Get the chroma qp.
646 */
647 62159762 static av_always_inline int get_chroma_qp(const PPS *pps, int t, int qscale)
648 {
649 62159762 return pps->chroma_qp_table[t][qscale];
650 }
651
652 /**
653 * Get the predicted intra4x4 prediction mode.
654 */
655 30247800 static av_always_inline int pred_intra_mode(const H264Context *h,
656 H264SliceContext *sl, int n)
657 {
658 30247800 const int index8 = scan8[n];
659 30247800 const int left = sl->intra4x4_pred_mode_cache[index8 - 1];
660 30247800 const int top = sl->intra4x4_pred_mode_cache[index8 - 8];
661 30247800 const int min = FFMIN(left, top);
662
663 ff_tlog(h->avctx, "mode:%d %d min:%d\n", left, top, min);
664
665
2/2
✓ Branch 0 taken 847409 times.
✓ Branch 1 taken 29400391 times.
 30247800 if (min < 0)
666 847409 return DC_PRED;
667 else
668 29400391 return min;
669 }
670
671 3034386 static av_always_inline void write_back_intra_pred_mode(const H264Context *h,
672 H264SliceContext *sl)
673 {
674 3034386 int8_t *i4x4 = sl->intra4x4_pred_mode + h->mb2br_xy[sl->mb_xy];
675 3034386 int8_t *i4x4_cache = sl->intra4x4_pred_mode_cache;
676
677 3034386 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
678 3034386 i4x4[4] = i4x4_cache[7 + 8 * 3];
679 3034386 i4x4[5] = i4x4_cache[7 + 8 * 2];
680 3034386 i4x4[6] = i4x4_cache[7 + 8 * 1];
681 3034386 }
682
683 10593095 static av_always_inline void write_back_non_zero_count(const H264Context *h,
684 H264SliceContext *sl)
685 {
686 10593095 const int mb_xy = sl->mb_xy;
687 10593095 uint8_t *nnz = h->non_zero_count[mb_xy];
688 10593095 uint8_t *nnz_cache = sl->non_zero_count_cache;
689
690 10593095 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
691 10593095 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
692 10593095 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
693 10593095 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
694 10593095 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
695 10593095 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
696 10593095 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
697 10593095 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
698
699
2/2
✓ Branch 0 taken 971539 times.
✓ Branch 1 taken 9621556 times.
 10593095 if (!h->chroma_y_shift) {
700 971539 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
701 971539 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
702 971539 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
703 971539 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
704 }
705 10593095 }
706
707 14129878 static av_always_inline void write_back_motion_list(const H264Context *h,
708 H264SliceContext *sl,
709 int b_stride,
710 int b_xy, int b8_xy,
711 int mb_type, int list)
712 {
713 14129878 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
714 14129878 int16_t(*mv_src)[2] = &sl->mv_cache[list][scan8[0]];
715 14129878 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
716 14129878 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
717 14129878 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
718 14129878 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
719 if (CABAC(h)) {
720 10254015 uint8_t (*mvd_dst)[2] = &sl->mvd_table[list][FMO ? 8 * sl->mb_xy
721 10254015 : h->mb2br_xy[sl->mb_xy]];
722 10254015 uint8_t(*mvd_src)[2] = &sl->mvd_cache[list][scan8[0]];
723
2/2
✓ Branch 0 taken 4658019 times.
✓ Branch 1 taken 5595996 times.
 10254015 if (IS_SKIP(mb_type)) {
724 4658019 AV_ZERO128(mvd_dst);
725 } else {
726 5595996 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
727 5595996 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
728 5595996 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
729 5595996 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
730 }
731 }
732
733 {
734 14129878 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
735 14129878 int8_t *ref_cache = sl->ref_cache[list];
736 14129878 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
737 14129878 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
738 14129878 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
739 14129878 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
740 }
741 14129878 }
742
743 10502061 static av_always_inline void write_back_motion(const H264Context *h,
744 H264SliceContext *sl,
745 int mb_type)
746 {
747 10502061 const int b_stride = h->b_stride;
748 10502061 const int b_xy = 4 * sl->mb_x + 4 * sl->mb_y * h->b_stride; // try mb2b(8)_xy
749 10502061 const int b8_xy = 4 * sl->mb_xy;
750
751
2/2
✓ Branch 0 taken 9727385 times.
✓ Branch 1 taken 774676 times.
 10502061 if (USES_LIST(mb_type, 0)) {
752 9727385 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 0);
753 } else {
754 774676 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
755 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
756 }
757
2/2
✓ Branch 0 taken 4402493 times.
✓ Branch 1 taken 6099568 times.
 10502061 if (USES_LIST(mb_type, 1))
758 4402493 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 1);
759
760
2/2
✓ Branch 0 taken 4361961 times.
✓ Branch 1 taken 2918239 times.
 7280200 if (sl->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
761
2/2
✓ Branch 0 taken 420200 times.
✓ Branch 1 taken 3941761 times.
 4361961 if (IS_8X8(mb_type)) {
762 420200 uint8_t *direct_table = &h->direct_table[4 * sl->mb_xy];
763 420200 direct_table[1] = sl->sub_mb_type[1] >> 1;
764 420200 direct_table[2] = sl->sub_mb_type[2] >> 1;
765 420200 direct_table[3] = sl->sub_mb_type[3] >> 1;
766 }
767 }
768 10502061 }
769
770 272274 static av_always_inline int get_dct8x8_allowed(const H264Context *h, H264SliceContext *sl)
771 {
772
2/2
✓ Branch 0 taken 267397 times.
✓ Branch 1 taken 4877 times.
 272274 if (h->ps.sps->direct_8x8_inference_flag)
773 267397 return !(AV_RN64A(sl->sub_mb_type) &
774 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
775 0x0001000100010001ULL));
776 else
777 4877 return !(AV_RN64A(sl->sub_mb_type) &
778 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
779 0x0001000100010001ULL));
780 }
781
782 int ff_h264_field_end(H264Context *h, H264SliceContext *sl, int in_setup);
783
784 int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
785 int ff_h264_replace_picture(H264Context *h, H264Picture *dst, const H264Picture *src);
786 void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
787
788 void ff_h264_slice_context_init(H264Context *h, H264SliceContext *sl);
789
790 void ff_h264_draw_horiz_band(const H264Context *h, H264SliceContext *sl, int y, int height);
791
792 /**
793 * Submit a slice for decoding.
794 *
795 * Parse the slice header, starting a new field/frame if necessary. If any
796 * slices are queued for the previous field, they are decoded.
797 */
798 int ff_h264_queue_decode_slice(H264Context *h, const H2645NAL *nal);
799 int ff_h264_execute_decode_slices(H264Context *h);
800 int ff_h264_update_thread_context(AVCodecContext *dst,
801 const AVCodecContext *src);
802 int ff_h264_update_thread_context_for_user(AVCodecContext *dst,
803 const AVCodecContext *src);
804
805 void ff_h264_flush_change(H264Context *h);
806
807 void ff_h264_free_tables(H264Context *h);
808
809 void ff_h264_set_erpic(ERPicture *dst, H264Picture *src);
810
811 #endif /* AVCODEC_H264DEC_H */
812