1 /*
2 * HEVC video decoder
3 *
4 * Copyright (C) 2012 - 2013 Guillaume Martres
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 #ifndef AVCODEC_HEVCDEC_H
24 #define AVCODEC_HEVCDEC_H
25
26 #include <stdatomic.h>
27
28 #include "libavutil/buffer.h"
29 #include "libavutil/md5.h"
30
31 #include "avcodec.h"
32 #include "bswapdsp.h"
33 #include "cabac.h"
34 #include "get_bits.h"
35 #include "hevcpred.h"
36 #include "h2645_parse.h"
37 #include "hevc.h"
38 #include "hevc_ps.h"
39 #include "hevc_sei.h"
40 #include "hevcdsp.h"
41 #include "internal.h"
42 #include "thread.h"
43 #include "videodsp.h"
44
45 #define MAX_NB_THREADS 16
46 #define SHIFT_CTB_WPP 2
47
48 //TODO: check if this is really the maximum
49 #define MAX_TRANSFORM_DEPTH 5
50
51 #define MAX_TB_SIZE 32
52 #define MAX_QP 51
53 #define DEFAULT_INTRA_TC_OFFSET 2
54
55 #define HEVC_CONTEXTS 199
56
57 #define MRG_MAX_NUM_CANDS 5
58
59 #define L0 0
60 #define L1 1
61
62 #define EPEL_EXTRA_BEFORE 1
63 #define EPEL_EXTRA_AFTER 2
64 #define EPEL_EXTRA 3
65 #define QPEL_EXTRA_BEFORE 3
66 #define QPEL_EXTRA_AFTER 4
67 #define QPEL_EXTRA 7
68
69 #define EDGE_EMU_BUFFER_STRIDE 80
70
71 /**
72 * Value of the luma sample at position (x, y) in the 2D array tab.
73 */
74 #define SAMPLE(tab, x, y) ((tab)[(y) * s->sps->width + (x)])
75 #define SAMPLE_CTB(tab, x, y) ((tab)[(y) * min_cb_width + (x)])
76
77 #define IS_IDR(s) ((s)->nal_unit_type == HEVC_NAL_IDR_W_RADL || (s)->nal_unit_type == HEVC_NAL_IDR_N_LP)
78 #define IS_BLA(s) ((s)->nal_unit_type == HEVC_NAL_BLA_W_RADL || (s)->nal_unit_type == HEVC_NAL_BLA_W_LP || \
79 (s)->nal_unit_type == HEVC_NAL_BLA_N_LP)
80 #define IS_IRAP(s) ((s)->nal_unit_type >= 16 && (s)->nal_unit_type <= 23)
81
82 enum RPSType {
83 ST_CURR_BEF = 0,
84 ST_CURR_AFT,
85 ST_FOLL,
86 LT_CURR,
87 LT_FOLL,
88 NB_RPS_TYPE,
89 };
90
91 enum SyntaxElement {
92 SAO_MERGE_FLAG = 0,
93 SAO_TYPE_IDX,
94 SAO_EO_CLASS,
95 SAO_BAND_POSITION,
96 SAO_OFFSET_ABS,
97 SAO_OFFSET_SIGN,
98 END_OF_SLICE_FLAG,
99 SPLIT_CODING_UNIT_FLAG,
100 CU_TRANSQUANT_BYPASS_FLAG,
101 SKIP_FLAG,
102 CU_QP_DELTA,
103 PRED_MODE_FLAG,
104 PART_MODE,
105 PCM_FLAG,
106 PREV_INTRA_LUMA_PRED_FLAG,
107 MPM_IDX,
108 REM_INTRA_LUMA_PRED_MODE,
109 INTRA_CHROMA_PRED_MODE,
110 MERGE_FLAG,
111 MERGE_IDX,
112 INTER_PRED_IDC,
113 REF_IDX_L0,
114 REF_IDX_L1,
115 ABS_MVD_GREATER0_FLAG,
116 ABS_MVD_GREATER1_FLAG,
117 ABS_MVD_MINUS2,
118 MVD_SIGN_FLAG,
119 MVP_LX_FLAG,
120 NO_RESIDUAL_DATA_FLAG,
121 SPLIT_TRANSFORM_FLAG,
122 CBF_LUMA,
123 CBF_CB_CR,
124 TRANSFORM_SKIP_FLAG,
125 EXPLICIT_RDPCM_FLAG,
126 EXPLICIT_RDPCM_DIR_FLAG,
127 LAST_SIGNIFICANT_COEFF_X_PREFIX,
128 LAST_SIGNIFICANT_COEFF_Y_PREFIX,
129 LAST_SIGNIFICANT_COEFF_X_SUFFIX,
130 LAST_SIGNIFICANT_COEFF_Y_SUFFIX,
131 SIGNIFICANT_COEFF_GROUP_FLAG,
132 SIGNIFICANT_COEFF_FLAG,
133 COEFF_ABS_LEVEL_GREATER1_FLAG,
134 COEFF_ABS_LEVEL_GREATER2_FLAG,
135 COEFF_ABS_LEVEL_REMAINING,
136 COEFF_SIGN_FLAG,
137 LOG2_RES_SCALE_ABS,
138 RES_SCALE_SIGN_FLAG,
139 CU_CHROMA_QP_OFFSET_FLAG,
140 CU_CHROMA_QP_OFFSET_IDX,
141 };
142
143 enum PartMode {
144 PART_2Nx2N = 0,
145 PART_2NxN = 1,
146 PART_Nx2N = 2,
147 PART_NxN = 3,
148 PART_2NxnU = 4,
149 PART_2NxnD = 5,
150 PART_nLx2N = 6,
151 PART_nRx2N = 7,
152 };
153
154 enum PredMode {
155 MODE_INTER = 0,
156 MODE_INTRA,
157 MODE_SKIP,
158 };
159
160 enum InterPredIdc {
161 PRED_L0 = 0,
162 PRED_L1,
163 PRED_BI,
164 };
165
166 enum PredFlag {
167 PF_INTRA = 0,
168 PF_L0,
169 PF_L1,
170 PF_BI,
171 };
172
173 enum IntraPredMode {
174 INTRA_PLANAR = 0,
175 INTRA_DC,
176 INTRA_ANGULAR_2,
177 INTRA_ANGULAR_3,
178 INTRA_ANGULAR_4,
179 INTRA_ANGULAR_5,
180 INTRA_ANGULAR_6,
181 INTRA_ANGULAR_7,
182 INTRA_ANGULAR_8,
183 INTRA_ANGULAR_9,
184 INTRA_ANGULAR_10,
185 INTRA_ANGULAR_11,
186 INTRA_ANGULAR_12,
187 INTRA_ANGULAR_13,
188 INTRA_ANGULAR_14,
189 INTRA_ANGULAR_15,
190 INTRA_ANGULAR_16,
191 INTRA_ANGULAR_17,
192 INTRA_ANGULAR_18,
193 INTRA_ANGULAR_19,
194 INTRA_ANGULAR_20,
195 INTRA_ANGULAR_21,
196 INTRA_ANGULAR_22,
197 INTRA_ANGULAR_23,
198 INTRA_ANGULAR_24,
199 INTRA_ANGULAR_25,
200 INTRA_ANGULAR_26,
201 INTRA_ANGULAR_27,
202 INTRA_ANGULAR_28,
203 INTRA_ANGULAR_29,
204 INTRA_ANGULAR_30,
205 INTRA_ANGULAR_31,
206 INTRA_ANGULAR_32,
207 INTRA_ANGULAR_33,
208 INTRA_ANGULAR_34,
209 };
210
211 enum SAOType {
212 SAO_NOT_APPLIED = 0,
213 SAO_BAND,
214 SAO_EDGE,
215 SAO_APPLIED
216 };
217
218 enum SAOEOClass {
219 SAO_EO_HORIZ = 0,
220 SAO_EO_VERT,
221 SAO_EO_135D,
222 SAO_EO_45D,
223 };
224
225 enum ScanType {
226 SCAN_DIAG = 0,
227 SCAN_HORIZ,
228 SCAN_VERT,
229 };
230
231 typedef struct RefPicList {
232 struct HEVCFrame *ref[HEVC_MAX_REFS];
233 int list[HEVC_MAX_REFS];
234 int isLongTerm[HEVC_MAX_REFS];
235 int nb_refs;
236 } RefPicList;
237
238 typedef struct RefPicListTab {
239 RefPicList refPicList[2];
240 } RefPicListTab;
241
242 typedef struct CodingUnit {
243 int x;
244 int y;
245
246 enum PredMode pred_mode; ///< PredMode
247 enum PartMode part_mode; ///< PartMode
248
249 // Inferred parameters
250 uint8_t intra_split_flag; ///< IntraSplitFlag
251 uint8_t max_trafo_depth; ///< MaxTrafoDepth
252 uint8_t cu_transquant_bypass_flag;
253 } CodingUnit;
254
255 typedef struct Mv {
256 int16_t x; ///< horizontal component of motion vector
257 int16_t y; ///< vertical component of motion vector
258 } Mv;
259
260 typedef struct MvField {
261 DECLARE_ALIGNED(4, Mv, mv)[2];
262 int8_t ref_idx[2];
263 int8_t pred_flag;
264 } MvField;
265
266 typedef struct NeighbourAvailable {
267 int cand_bottom_left;
268 int cand_left;
269 int cand_up;
270 int cand_up_left;
271 int cand_up_right;
272 int cand_up_right_sap;
273 } NeighbourAvailable;
274
275 typedef struct PredictionUnit {
276 int mpm_idx;
277 int rem_intra_luma_pred_mode;
278 uint8_t intra_pred_mode[4];
279 Mv mvd;
280 uint8_t merge_flag;
281 uint8_t intra_pred_mode_c[4];
282 uint8_t chroma_mode_c[4];
283 } PredictionUnit;
284
285 typedef struct TransformUnit {
286 int cu_qp_delta;
287
288 int res_scale_val;
289
290 // Inferred parameters;
291 int intra_pred_mode;
292 int intra_pred_mode_c;
293 int chroma_mode_c;
294 uint8_t is_cu_qp_delta_coded;
295 uint8_t is_cu_chroma_qp_offset_coded;
296 int8_t cu_qp_offset_cb;
297 int8_t cu_qp_offset_cr;
298 uint8_t cross_pf;
299 } TransformUnit;
300
301 typedef struct DBParams {
302 int beta_offset;
303 int tc_offset;
304 } DBParams;
305
306 #define HEVC_FRAME_FLAG_OUTPUT (1 << 0)
307 #define HEVC_FRAME_FLAG_SHORT_REF (1 << 1)
308 #define HEVC_FRAME_FLAG_LONG_REF (1 << 2)
309 #define HEVC_FRAME_FLAG_BUMPING (1 << 3)
310
311 typedef struct HEVCFrame {
312 AVFrame *frame;
313 ThreadFrame tf;
314 MvField *tab_mvf;
315 RefPicList *refPicList;
316 RefPicListTab **rpl_tab;
317 int ctb_count;
318 int poc;
319 struct HEVCFrame *collocated_ref;
320
321 AVBufferRef *tab_mvf_buf;
322 AVBufferRef *rpl_tab_buf;
323 AVBufferRef *rpl_buf;
324
325 AVBufferRef *hwaccel_priv_buf;
326 void *hwaccel_picture_private;
327
328 /**
329 * A sequence counter, so that old frames are output first
330 * after a POC reset
331 */
332 uint16_t sequence;
333
334 /**
335 * A combination of HEVC_FRAME_FLAG_*
336 */
337 uint8_t flags;
338 } HEVCFrame;
339
340 typedef struct HEVCLocalContext {
341 uint8_t cabac_state[HEVC_CONTEXTS];
342
343 uint8_t stat_coeff[4];
344
345 uint8_t first_qp_group;
346
347 GetBitContext gb;
348 CABACContext cc;
349
350 int8_t qp_y;
351 int8_t curr_qp_y;
352
353 int qPy_pred;
354
355 TransformUnit tu;
356
357 uint8_t ctb_left_flag;
358 uint8_t ctb_up_flag;
359 uint8_t ctb_up_right_flag;
360 uint8_t ctb_up_left_flag;
361 int end_of_tiles_x;
362 int end_of_tiles_y;
363 /* +7 is for subpixel interpolation, *2 for high bit depths */
364 DECLARE_ALIGNED(32, uint8_t, edge_emu_buffer)[(MAX_PB_SIZE + 7) * EDGE_EMU_BUFFER_STRIDE * 2];
365 /* The extended size between the new edge emu buffer is abused by SAO */
366 DECLARE_ALIGNED(32, uint8_t, edge_emu_buffer2)[(MAX_PB_SIZE + 7) * EDGE_EMU_BUFFER_STRIDE * 2];
367 DECLARE_ALIGNED(32, int16_t, tmp)[MAX_PB_SIZE * MAX_PB_SIZE];
368
369 int ct_depth;
370 CodingUnit cu;
371 PredictionUnit pu;
372 NeighbourAvailable na;
373
374 #define BOUNDARY_LEFT_SLICE (1 << 0)
375 #define BOUNDARY_LEFT_TILE (1 << 1)
376 #define BOUNDARY_UPPER_SLICE (1 << 2)
377 #define BOUNDARY_UPPER_TILE (1 << 3)
378 /* properties of the boundary of the current CTB for the purposes
379 * of the deblocking filter */
380 int boundary_flags;
381 } HEVCLocalContext;
382
383 typedef struct HEVCContext {
384 const AVClass *c; // needed by private avoptions
385 AVCodecContext *avctx;
386
387 struct HEVCContext *sList[MAX_NB_THREADS];
388
389 HEVCLocalContext *HEVClcList[MAX_NB_THREADS];
390 HEVCLocalContext *HEVClc;
391
392 uint8_t threads_type;
393 uint8_t threads_number;
394
395 int width;
396 int height;
397
398 uint8_t *cabac_state;
399
400 /** 1 if the independent slice segment header was successfully parsed */
401 uint8_t slice_initialized;
402
403 AVFrame *frame;
404 AVFrame *output_frame;
405 uint8_t *sao_pixel_buffer_h[3];
406 uint8_t *sao_pixel_buffer_v[3];
407
408 HEVCParamSets ps;
409 HEVCSEI sei;
410 struct AVMD5 *md5_ctx;
411
412 AVBufferPool *tab_mvf_pool;
413 AVBufferPool *rpl_tab_pool;
414
415 ///< candidate references for the current frame
416 RefPicList rps[5];
417
418 SliceHeader sh;
419 SAOParams *sao;
420 DBParams *deblock;
421 enum HEVCNALUnitType nal_unit_type;
422 int temporal_id; ///< temporal_id_plus1 - 1
423 HEVCFrame *ref;
424 HEVCFrame DPB[32];
425 int poc;
426 int pocTid0;
427 int slice_idx; ///< number of the slice being currently decoded
428 int eos; ///< current packet contains an EOS/EOB NAL
429 int last_eos; ///< last packet contains an EOS/EOB NAL
430 int max_ra;
431 int bs_width;
432 int bs_height;
433 int overlap;
434
435 int is_decoded;
436 int no_rasl_output_flag;
437
438 HEVCPredContext hpc;
439 HEVCDSPContext hevcdsp;
440 VideoDSPContext vdsp;
441 BswapDSPContext bdsp;
442 int8_t *qp_y_tab;
443 uint8_t *horizontal_bs;
444 uint8_t *vertical_bs;
445
446 int32_t *tab_slice_address;
447
448 // CU
449 uint8_t *skip_flag;
450 uint8_t *tab_ct_depth;
451 // PU
452 uint8_t *tab_ipm;
453
454 uint8_t *cbf_luma; // cbf_luma of colocated TU
455 uint8_t *is_pcm;
456
457 // CTB-level flags affecting loop filter operation
458 uint8_t *filter_slice_edges;
459
460 /** used on BE to byteswap the lines for checksumming */
461 uint8_t *checksum_buf;
462 int checksum_buf_size;
463
464 /**
465 * Sequence counters for decoded and output frames, so that old
466 * frames are output first after a POC reset
467 */
468 uint16_t seq_decode;
469 uint16_t seq_output;
470
471 int enable_parallel_tiles;
472 atomic_int wpp_err;
473
474 const uint8_t *data;
475
476 H2645Packet pkt;
477 // type of the first VCL NAL of the current frame
478 enum HEVCNALUnitType first_nal_type;
479
480 uint8_t context_initialized;
481 int is_nalff; ///< this flag is != 0 if bitstream is encapsulated
482 ///< as a format defined in 14496-15
483 int apply_defdispwin;
484
485 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
486 int nuh_layer_id;
487 } HEVCContext;
488
489 /**
490 * Mark all frames in DPB as unused for reference.
491 */
492 void ff_hevc_clear_refs(HEVCContext *s);
493
494 /**
495 * Drop all frames currently in DPB.
496 */
497 void ff_hevc_flush_dpb(HEVCContext *s);
498
499 RefPicList *ff_hevc_get_ref_list(HEVCContext *s, HEVCFrame *frame,
500 int x0, int y0);
501
502 /**
503 * Construct the reference picture sets for the current frame.
504 */
505 int ff_hevc_frame_rps(HEVCContext *s);
506
507 /**
508 * Construct the reference picture list(s) for the current slice.
509 */
510 int ff_hevc_slice_rpl(HEVCContext *s);
511
512 void ff_hevc_save_states(HEVCContext *s, int ctb_addr_ts);
513 int ff_hevc_cabac_init(HEVCContext *s, int ctb_addr_ts);
514 int ff_hevc_sao_merge_flag_decode(HEVCContext *s);
515 int ff_hevc_sao_type_idx_decode(HEVCContext *s);
516 int ff_hevc_sao_band_position_decode(HEVCContext *s);
517 int ff_hevc_sao_offset_abs_decode(HEVCContext *s);
518 int ff_hevc_sao_offset_sign_decode(HEVCContext *s);
519 int ff_hevc_sao_eo_class_decode(HEVCContext *s);
520 int ff_hevc_end_of_slice_flag_decode(HEVCContext *s);
521 int ff_hevc_cu_transquant_bypass_flag_decode(HEVCContext *s);
522 int ff_hevc_skip_flag_decode(HEVCContext *s, int x0, int y0,
523 int x_cb, int y_cb);
524 int ff_hevc_pred_mode_decode(HEVCContext *s);
525 int ff_hevc_split_coding_unit_flag_decode(HEVCContext *s, int ct_depth,
526 int x0, int y0);
527 int ff_hevc_part_mode_decode(HEVCContext *s, int log2_cb_size);
528 int ff_hevc_pcm_flag_decode(HEVCContext *s);
529 int ff_hevc_prev_intra_luma_pred_flag_decode(HEVCContext *s);
530 int ff_hevc_mpm_idx_decode(HEVCContext *s);
531 int ff_hevc_rem_intra_luma_pred_mode_decode(HEVCContext *s);
532 int ff_hevc_intra_chroma_pred_mode_decode(HEVCContext *s);
533 int ff_hevc_merge_idx_decode(HEVCContext *s);
534 int ff_hevc_merge_flag_decode(HEVCContext *s);
535 int ff_hevc_inter_pred_idc_decode(HEVCContext *s, int nPbW, int nPbH);
536 int ff_hevc_ref_idx_lx_decode(HEVCContext *s, int num_ref_idx_lx);
537 int ff_hevc_mvp_lx_flag_decode(HEVCContext *s);
538 int ff_hevc_no_residual_syntax_flag_decode(HEVCContext *s);
539 int ff_hevc_split_transform_flag_decode(HEVCContext *s, int log2_trafo_size);
540 int ff_hevc_cbf_cb_cr_decode(HEVCContext *s, int trafo_depth);
541 int ff_hevc_cbf_luma_decode(HEVCContext *s, int trafo_depth);
542 int ff_hevc_log2_res_scale_abs(HEVCContext *s, int idx);
543 int ff_hevc_res_scale_sign_flag(HEVCContext *s, int idx);
544
545 /**
546 * Get the number of candidate references for the current frame.
547 */
548 int ff_hevc_frame_nb_refs(const HEVCContext *s);
549
550 int ff_hevc_set_new_ref(HEVCContext *s, AVFrame **frame, int poc);
551
ff_hevc_nal_is_nonref(enum HEVCNALUnitType type)552 static av_always_inline int ff_hevc_nal_is_nonref(enum HEVCNALUnitType type)
553 {
554 switch (type) {
555 case HEVC_NAL_TRAIL_N:
556 case HEVC_NAL_TSA_N:
557 case HEVC_NAL_STSA_N:
558 case HEVC_NAL_RADL_N:
559 case HEVC_NAL_RASL_N:
560 case HEVC_NAL_VCL_N10:
561 case HEVC_NAL_VCL_N12:
562 case HEVC_NAL_VCL_N14:
563 return 1;
564 default: break;
565 }
566 return 0;
567 }
568
569 /**
570 * Find next frame in output order and put a reference to it in frame.
571 * @return 1 if a frame was output, 0 otherwise
572 */
573 int ff_hevc_output_frame(HEVCContext *s, AVFrame *frame, int flush);
574
575 void ff_hevc_bump_frame(HEVCContext *s);
576
577 void ff_hevc_unref_frame(HEVCContext *s, HEVCFrame *frame, int flags);
578
579 void ff_hevc_set_neighbour_available(HEVCContext *s, int x0, int y0,
580 int nPbW, int nPbH);
581 void ff_hevc_luma_mv_merge_mode(HEVCContext *s, int x0, int y0,
582 int nPbW, int nPbH, int log2_cb_size,
583 int part_idx, int merge_idx, MvField *mv);
584 void ff_hevc_luma_mv_mvp_mode(HEVCContext *s, int x0, int y0,
585 int nPbW, int nPbH, int log2_cb_size,
586 int part_idx, int merge_idx,
587 MvField *mv, int mvp_lx_flag, int LX);
588 void ff_hevc_set_qPy(HEVCContext *s, int xBase, int yBase,
589 int log2_cb_size);
590 void ff_hevc_deblocking_boundary_strengths(HEVCContext *s, int x0, int y0,
591 int log2_trafo_size);
592 int ff_hevc_cu_qp_delta_sign_flag(HEVCContext *s);
593 int ff_hevc_cu_qp_delta_abs(HEVCContext *s);
594 int ff_hevc_cu_chroma_qp_offset_flag(HEVCContext *s);
595 int ff_hevc_cu_chroma_qp_offset_idx(HEVCContext *s);
596 void ff_hevc_hls_filter(HEVCContext *s, int x, int y, int ctb_size);
597 void ff_hevc_hls_filters(HEVCContext *s, int x_ctb, int y_ctb, int ctb_size);
598 void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0,
599 int log2_trafo_size, enum ScanType scan_idx,
600 int c_idx);
601
602 void ff_hevc_hls_mvd_coding(HEVCContext *s, int x0, int y0, int log2_cb_size);
603
604 extern const uint8_t ff_hevc_qpel_extra_before[4];
605 extern const uint8_t ff_hevc_qpel_extra_after[4];
606 extern const uint8_t ff_hevc_qpel_extra[4];
607
608 #endif /* AVCODEC_HEVCDEC_H */
609