HNM4/HNM4A demuxer & video decoder
[libav.git] / libavcodec / h264.h
CommitLineData
26b4fe82
AJ
1/*
2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4 *
2912e87a 5 * This file is part of Libav.
26b4fe82 6 *
2912e87a 7 * Libav is free software; you can redistribute it and/or
26b4fe82
AJ
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 *
2912e87a 12 * Libav is distributed in the hope that it will be useful,
26b4fe82
AJ
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
2912e87a 18 * License along with Libav; if not, write to the Free Software
26b4fe82 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26b4fe82
AJ
20 */
21
22/**
ba87f080 23 * @file
26b4fe82
AJ
24 * H.264 / AVC / MPEG4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
98790382
SS
28#ifndef AVCODEC_H264_H
29#define AVCODEC_H264_H
26b4fe82 30
19769ece 31#include "libavutil/intreadwrite.h"
26b4fe82 32#include "cabac.h"
5f401b7b 33#include "error_resilience.h"
2c541554 34#include "get_bits.h"
26b4fe82 35#include "mpegvideo.h"
79dad2a9 36#include "h264chroma.h"
4693b031 37#include "h264dsp.h"
c92a30bb 38#include "h264pred.h"
e9d81735 39#include "h264qpel.h"
e1e94902 40#include "rectangle.h"
26b4fe82 41
be545b8a
DB
42#define MAX_SPS_COUNT 32
43#define MAX_PPS_COUNT 256
26b4fe82 44
be545b8a 45#define MAX_MMCO_COUNT 66
26b4fe82 46
be545b8a 47#define MAX_DELAYED_PIC_COUNT 16
64b9d48f 48
26b4fe82
AJ
49/* Compiling in interlaced support reduces the speed
50 * of progressive decoding by about 2%. */
51#define ALLOW_INTERLACE
52
d43c1922
MN
53#define FMO 0
54
b735aeea
MN
55/**
56 * The maximum number of slices supported by the decoder.
57 * must be a power of 2
58 */
59#define MAX_SLICES 16
60
26b4fe82 61#ifdef ALLOW_INTERLACE
82313eaa 62#define MB_MBAFF(h) h->mb_mbaff
da6be8fc 63#define MB_FIELD(h) h->mb_field_decoding_flag
7bece9b2 64#define FRAME_MBAFF(h) h->mb_aff_frame
7fa00653 65#define FIELD_PICTURE(h) (h->picture_structure != PICT_FRAME)
556f8a06 66#define LEFT_MBS 2
be545b8a
DB
67#define LTOP 0
68#define LBOT 1
69#define LEFT(i) (i)
26b4fe82 70#else
82313eaa 71#define MB_MBAFF(h) 0
da6be8fc 72#define MB_FIELD(h) 0
7bece9b2 73#define FRAME_MBAFF(h) 0
7fa00653 74#define FIELD_PICTURE(h) 0
26b4fe82
AJ
75#undef IS_INTERLACED
76#define IS_INTERLACED(mb_type) 0
556f8a06 77#define LEFT_MBS 1
be545b8a
DB
78#define LTOP 0
79#define LBOT 0
80#define LEFT(i) 0
26b4fe82 81#endif
a6931d8e 82#define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
26b4fe82 83
ddd60f28 84#ifndef CABAC
6d2b6f21 85#define CABAC(h) h->pps.cabac
ddd60f28
MN
86#endif
87
e962bd08 88#define CHROMA422(h) (h->sps.chroma_format_idc == 2)
23e85be5 89#define CHROMA444(h) (h->sps.chroma_format_idc == 3)
c90b9442 90
be545b8a 91#define EXTENDED_SAR 255
44ee6135 92
be545b8a 93#define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
ce9ff9cc
DB
94#define MB_TYPE_8x8DCT 0x01000000
95#define IS_REF0(a) ((a) & MB_TYPE_REF0)
96#define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
97
be545b8a 98#define QP_MAX_NUM (51 + 2 * 6) // The maximum supported qp
ea6f00c4 99
44ee6135
MN
100/* NAL unit types */
101enum {
be545b8a 102 NAL_SLICE = 1,
2791c608
MN
103 NAL_DPA,
104 NAL_DPB,
105 NAL_DPC,
106 NAL_IDR_SLICE,
107 NAL_SEI,
108 NAL_SPS,
109 NAL_PPS,
110 NAL_AUD,
111 NAL_END_SEQUENCE,
112 NAL_END_STREAM,
113 NAL_FILLER_DATA,
114 NAL_SPS_EXT,
61c6eef5
JG
115 NAL_AUXILIARY_SLICE = 19,
116 NAL_FF_IGNORE = 0xff0f001,
44ee6135
MN
117};
118
26b4fe82 119/**
dd02ea11
IS
120 * SEI message types
121 */
122typedef enum {
be545b8a
DB
123 SEI_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
124 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
125 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
126 SEI_TYPE_RECOVERY_POINT = 6 ///< recovery point (frame # to decoder sync)
dd02ea11
IS
127} SEI_Type;
128
129/**
b09a7c05
130 * pic_struct in picture timing SEI message
131 */
132typedef enum {
133 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
134 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
135 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
136 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
137 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
138 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
139 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
140 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
141 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
142} SEI_PicStructType;
143
144/**
26b4fe82
AJ
145 * Sequence parameter set
146 */
be545b8a 147typedef struct SPS {
26b4fe82
AJ
148 int profile_idc;
149 int level_idc;
0af6967e 150 int chroma_format_idc;
26b4fe82
AJ
151 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
152 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
153 int poc_type; ///< pic_order_cnt_type
154 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
155 int delta_pic_order_always_zero_flag;
156 int offset_for_non_ref_pic;
157 int offset_for_top_to_bottom_field;
158 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
159 int ref_frame_count; ///< num_ref_frames
160 int gaps_in_frame_num_allowed_flag;
bcc3476c
JD
161 int mb_width; ///< pic_width_in_mbs_minus1 + 1
162 int mb_height; ///< pic_height_in_map_units_minus1 + 1
26b4fe82 163 int frame_mbs_only_flag;
be545b8a 164 int mb_aff; ///< mb_adaptive_frame_field_flag
26b4fe82 165 int direct_8x8_inference_flag;
be545b8a 166 int crop; ///< frame_cropping_flag
5e83d9ac
AK
167
168 /* those 4 are already in luma samples */
59eeddf0
MN
169 unsigned int crop_left; ///< frame_cropping_rect_left_offset
170 unsigned int crop_right; ///< frame_cropping_rect_right_offset
171 unsigned int crop_top; ///< frame_cropping_rect_top_offset
172 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
26b4fe82
AJ
173 int vui_parameters_present_flag;
174 AVRational sar;
c4dffe7e
DC
175 int video_signal_type_present_flag;
176 int full_range;
177 int colour_description_present_flag;
178 enum AVColorPrimaries color_primaries;
179 enum AVColorTransferCharacteristic color_trc;
180 enum AVColorSpace colorspace;
26b4fe82
AJ
181 int timing_info_present_flag;
182 uint32_t num_units_in_tick;
183 uint32_t time_scale;
184 int fixed_frame_rate_flag;
be545b8a 185 short offset_for_ref_frame[256]; // FIXME dyn aloc?
26b4fe82
AJ
186 int bitstream_restriction_flag;
187 int num_reorder_frames;
188 int scaling_matrix_present;
189 uint8_t scaling_matrix4[6][16];
c90b9442 190 uint8_t scaling_matrix8[6][64];
b09a7c05
191 int nal_hrd_parameters_present_flag;
192 int vcl_hrd_parameters_present_flag;
193 int pic_struct_present_flag;
194 int time_offset_length;
be545b8a
DB
195 int cpb_cnt; ///< See H.264 E.1.2
196 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
197 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
198 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
199 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
200 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
201 int residual_color_transform_flag; ///< residual_colour_transform_flag
202 int constraint_set_flags; ///< constraint_set[0-3]_flag
072be3e8 203 int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
be545b8a 204} SPS;
26b4fe82
AJ
205
206/**
207 * Picture parameter set
208 */
be545b8a 209typedef struct PPS {
26b4fe82
AJ
210 unsigned int sps_id;
211 int cabac; ///< entropy_coding_mode_flag
212 int pic_order_present; ///< pic_order_present_flag
213 int slice_group_count; ///< num_slice_groups_minus1 + 1
214 int mb_slice_group_map_type;
215 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
216 int weighted_pred; ///< weighted_pred_flag
217 int weighted_bipred_idc;
218 int init_qp; ///< pic_init_qp_minus26 + 26
219 int init_qs; ///< pic_init_qs_minus26 + 26
4691a77d 220 int chroma_qp_index_offset[2];
26b4fe82 221 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
be545b8a
DB
222 int constrained_intra_pred; ///< constrained_intra_pred_flag
223 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
224 int transform_8x8_mode; ///< transform_8x8_mode_flag
26b4fe82 225 uint8_t scaling_matrix4[6][16];
c90b9442 226 uint8_t scaling_matrix8[6][64];
be545b8a 227 uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
4691a77d 228 int chroma_qp_diff;
be545b8a 229} PPS;
26b4fe82
AJ
230
231/**
232 * Memory management control operation opcode.
233 */
be545b8a
DB
234typedef enum MMCOOpcode {
235 MMCO_END = 0,
26b4fe82
AJ
236 MMCO_SHORT2UNUSED,
237 MMCO_LONG2UNUSED,
238 MMCO_SHORT2LONG,
239 MMCO_SET_MAX_LONG,
240 MMCO_RESET,
241 MMCO_LONG,
242} MMCOOpcode;
243
244/**
245 * Memory management control operation.
246 */
be545b8a 247typedef struct MMCO {
26b4fe82 248 MMCOOpcode opcode;
0d0447ea
JD
249 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
250 int long_arg; ///< index, pic_num, or num long refs depending on opcode
26b4fe82
AJ
251} MMCO;
252
253/**
254 * H264Context
255 */
be545b8a 256typedef struct H264Context {
2c541554
AK
257 AVCodecContext *avctx;
258 DSPContext dsp;
259 VideoDSPContext vdsp;
4693b031 260 H264DSPContext h264dsp;
79dad2a9 261 H264ChromaContext h264chroma;
e9d81735 262 H264QpelContext h264qpel;
2c541554
AK
263 MotionEstContext me;
264 ParseContext parse_context;
265 GetBitContext gb;
266 ERContext er;
267
268 Picture *DPB;
269 Picture *cur_pic_ptr;
270 Picture cur_pic;
2c541554 271
6e3ef511 272 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
be545b8a 273 int chroma_qp[2]; // QPc
26b4fe82 274
2e36c931
MN
275 int qp_thresh; ///< QP threshold to skip loopfilter
276
5e83d9ac 277 /* coded dimensions -- 16 * mb w/h */
2c541554 278 int width, height;
93f30547 279 ptrdiff_t linesize, uvlinesize;
2c541554
AK
280 int chroma_x_shift, chroma_y_shift;
281
282 int qscale;
283 int droppable;
284 int data_partitioning;
285 int coded_picture_number;
286 int low_delay;
287
288 int context_initialized;
289 int flags;
290 int workaround_bugs;
291
26b4fe82
AJ
292 int prev_mb_skipped;
293 int next_mb_skipped;
294
be545b8a 295 // prediction stuff
26b4fe82
AJ
296 int chroma_pred_mode;
297 int intra16x16_pred_mode;
298
c1bb66ac 299 int topleft_mb_xy;
26b4fe82 300 int top_mb_xy;
c1bb66ac 301 int topright_mb_xy;
556f8a06 302 int left_mb_xy[LEFT_MBS];
26b4fe82 303
c1bb66ac 304 int topleft_type;
99344d43 305 int top_type;
c1bb66ac 306 int topright_type;
556f8a06 307 int left_type[LEFT_MBS];
99344d43 308
be545b8a 309 const uint8_t *left_block;
c1bb66ac
MN
310 int topleft_partition;
311
be545b8a
DB
312 int8_t intra4x4_pred_mode_cache[5 * 8];
313 int8_t(*intra4x4_pred_mode);
c92a30bb 314 H264PredContext hpc;
26b4fe82
AJ
315 unsigned int topleft_samples_available;
316 unsigned int top_samples_available;
317 unsigned int topright_samples_available;
318 unsigned int left_samples_available;
be545b8a 319 uint8_t (*top_borders[2])[(16 * 3) * 2];
26b4fe82
AJ
320
321 /**
322 * non zero coeff count cache.
323 * is 64 if not available.
324 */
be545b8a 325 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
b6303e6d 326
c90b9442 327 uint8_t (*non_zero_count)[48];
26b4fe82
AJ
328
329 /**
330 * Motion vector cache.
331 */
be545b8a
DB
332 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
333 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
334#define LIST_NOT_USED -1 // FIXME rename?
26b4fe82
AJ
335#define PART_NOT_AVAILABLE -2
336
337 /**
26b4fe82
AJ
338 * number of neighbors (top and/or left) that used 8x8 dct
339 */
340 int neighbor_transform_size;
341
342 /**
343 * block_offset[ 0..23] for frame macroblocks
344 * block_offset[24..47] for field macroblocks
345 */
be545b8a 346 int block_offset[2 * (16 * 3)];
26b4fe82 347
be545b8a 348 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
d43c1922 349 uint32_t *mb2br_xy;
be545b8a 350 int b_stride; // FIXME use s->b4_stride
26b4fe82 351
93f30547
RB
352 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
353 ptrdiff_t mb_uvlinesize;
26b4fe82 354
072be3e8 355 unsigned current_sps_id; ///< id of the current SPS
26b4fe82
AJ
356 SPS sps; ///< current sps
357
26b4fe82
AJ
358 /**
359 * current pps
360 */
be545b8a 361 PPS pps; // FIXME move to Picture perhaps? (->no) do we need that?
26b4fe82 362
be545b8a
DB
363 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
364 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
365 uint32_t(*dequant4_coeff[6])[16];
366 uint32_t(*dequant8_coeff[6])[64];
26b4fe82
AJ
367
368 int slice_num;
be545b8a 369 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
26b4fe82 370 int slice_type;
be545b8a 371 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
26b4fe82
AJ
372 int slice_type_fixed;
373
be545b8a 374 // interlacing specific flags
26b4fe82
AJ
375 int mb_aff_frame;
376 int mb_field_decoding_flag;
be545b8a 377 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
2c541554
AK
378 int picture_structure;
379 int first_field;
26b4fe82 380
84dc2d8a 381 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
26b4fe82 382
be545b8a 383 // Weighted pred stuff
26b4fe82
AJ
384 int use_weight;
385 int use_weight_chroma;
386 int luma_log2_weight_denom;
387 int chroma_log2_weight_denom;
be545b8a 388 // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
3d9137c8
MN
389 int luma_weight[48][2][2];
390 int chroma_weight[48][2][2][2];
1052b76f 391 int implicit_weight[48][48][2];
26b4fe82 392
26b4fe82 393 int direct_spatial_mv_pred;
3a06e864
MN
394 int col_parity;
395 int col_fieldoff;
73ad2c2f 396 int dist_scale_factor[32];
8b1fd554 397 int dist_scale_factor_field[2][32];
be545b8a
DB
398 int map_col_to_list0[2][16 + 32];
399 int map_col_to_list0_field[2][2][16 + 32];
26b4fe82
AJ
400
401 /**
402 * num_ref_idx_l0/1_active_minus1 + 1
403 */
be545b8a 404 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
26b4fe82 405 unsigned int list_count;
be545b8a
DB
406 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
407 Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
408 * Reordered version of default_ref_list
409 * according to picture reordering in slice header */
410 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
26b4fe82 411
be545b8a 412 // data partitioning
26b4fe82
AJ
413 GetBitContext intra_gb;
414 GetBitContext inter_gb;
415 GetBitContext *intra_gb_ptr;
416 GetBitContext *inter_gb_ptr;
417
7ebfb466 418 const uint8_t *intra_pcm_ptr;
88bd7fdc
DB
419 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.
420 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
421 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
26b4fe82
AJ
422
423 /**
424 * Cabac
425 */
426 CABACContext cabac;
be545b8a 427 uint8_t cabac_state[1024];
26b4fe82 428
be545b8a
DB
429 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
430 uint16_t *cbp_table;
26b4fe82
AJ
431 int cbp;
432 int top_cbp;
433 int left_cbp;
434 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
be545b8a
DB
435 uint8_t *chroma_pred_mode_table;
436 int last_qscale_diff;
437 uint8_t (*mvd_table[2])[2];
438 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
439 uint8_t *direct_table;
440 uint8_t direct_cache[5 * 8];
26b4fe82
AJ
441
442 uint8_t zigzag_scan[16];
443 uint8_t zigzag_scan8x8[64];
444 uint8_t zigzag_scan8x8_cavlc[64];
445 uint8_t field_scan[16];
446 uint8_t field_scan8x8[64];
447 uint8_t field_scan8x8_cavlc[64];
448 const uint8_t *zigzag_scan_q0;
449 const uint8_t *zigzag_scan8x8_q0;
450 const uint8_t *zigzag_scan8x8_cavlc_q0;
451 const uint8_t *field_scan_q0;
452 const uint8_t *field_scan8x8_q0;
453 const uint8_t *field_scan8x8_cavlc_q0;
454
455 int x264_build;
afebe2f7 456
2c541554
AK
457 int mb_x, mb_y;
458 int resync_mb_x;
459 int resync_mb_y;
460 int mb_skip_run;
461 int mb_height, mb_width;
462 int mb_stride;
463 int mb_num;
bd8868e0
MN
464 int mb_xy;
465
466 int is_complex;
467
be545b8a
DB
468 // deblock
469 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
bd8868e0
MN
470 int slice_alpha_c0_offset;
471 int slice_beta_offset;
472
be545b8a
DB
473 // =============================================================
474 // Things below are not used in the MB or more inner code
bd8868e0
MN
475
476 int nal_ref_idc;
477 int nal_unit_type;
478 uint8_t *rbsp_buffer[2];
479 unsigned int rbsp_buffer_size[2];
480
481 /**
482 * Used to parse AVC variant of h264
483 */
be545b8a
DB
484 int is_avc; ///< this flag is != 0 if codec is avc1
485 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
486 int got_first; ///< this flag is != 0 if we've parsed a frame
bd8868e0 487
2c541554
AK
488 int bit_depth_luma; ///< luma bit depth from sps to detect changes
489 int chroma_format_idc; ///< chroma format from sps to detect changes
9e696d2e 490
bd8868e0
MN
491 SPS *sps_buffers[MAX_SPS_COUNT];
492 PPS *pps_buffers[MAX_PPS_COUNT];
493
be545b8a 494 int dequant_coeff_pps; ///< reinit tables when pps changes
bd8868e0
MN
495
496 uint16_t *slice_table_base;
497
be545b8a 498 // POC stuff
bd8868e0
MN
499 int poc_lsb;
500 int poc_msb;
501 int delta_poc_bottom;
502 int delta_poc[2];
503 int frame_num;
be545b8a
DB
504 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
505 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
506 int frame_num_offset; ///< for POC type 2
507 int prev_frame_num_offset; ///< for POC type 2
508 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
bd8868e0
MN
509
510 /**
be545b8a 511 * frame_num for frames or 2 * frame_num + 1 for field pics.
bd8868e0
MN
512 */
513 int curr_pic_num;
514
515 /**
be545b8a 516 * max_frame_num or 2 * max_frame_num for field pics.
bd8868e0
MN
517 */
518 int max_pic_num;
519
520 int redundant_pic_count;
521
940b8b58 522 Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
bd8868e0
MN
523 Picture *short_ref[32];
524 Picture *long_ref[32];
be545b8a 525 Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
ea2bb12e 526 int last_pocs[MAX_DELAYED_PIC_COUNT];
6a9c8594 527 Picture *next_output_pic;
bd8868e0 528 int outputed_poc;
6a9c8594 529 int next_outputed_poc;
bd8868e0
MN
530
531 /**
532 * memory management control operations buffer.
533 */
534 MMCO mmco[MAX_MMCO_COUNT];
535 int mmco_index;
adedd840 536 int mmco_reset;
bd8868e0 537
be545b8a
DB
538 int long_ref_count; ///< number of actual long term references
539 int short_ref_count; ///< number of actual short term references
bd8868e0 540
be545b8a 541 int cabac_init_idc;
bd8868e0 542
afebe2f7 543 /**
21a19b79 544 * @name Members for slice based multithreading
afebe2f7
545 * @{
546 */
547 struct H264Context *thread_context[MAX_THREADS];
548
549 /**
550 * current slice number, used to initalize slice_num of each thread/context
551 */
552 int current_slice;
553
554 /**
555 * Max number of threads / contexts.
556 * This is equal to AVCodecContext.thread_count unless
557 * multithreaded decoding is impossible, in which case it is
558 * reduced to 1.
559 */
560 int max_contexts;
561
2c541554
AK
562 int slice_context_count;
563
afebe2f7
564 /**
565 * 1 if the single thread fallback warning has already been
566 * displayed, 0 otherwise.
567 */
568 int single_decode_warning;
569
2c541554
AK
570 enum AVPictureType pict_type;
571
afebe2f7
572 int last_slice_type;
573 /** @} */
574
b09a7c05
575 /**
576 * pic_struct in picture timing SEI message
577 */
578 SEI_PicStructType sei_pic_struct;
89db0bae 579
37a558fe 580 /**
b19d493f
HY
581 * Complement sei_pic_struct
582 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
583 * However, soft telecined frames may have these values.
584 * This is used in an attempt to flag soft telecine progressive.
585 */
586 int prev_interlaced_frame;
587
588 /**
70e01da3
IS
589 * Bit set of clock types for fields/frames in picture timing SEI message.
590 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
591 * interlaced).
592 */
593 int sei_ct_type;
594
595 /**
ff594f81
IS
596 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
597 */
598 int sei_dpb_output_delay;
599
600 /**
cf6065ca
IS
601 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
602 */
603 int sei_cpb_removal_delay;
604
605 /**
37a558fe
IS
606 * recovery_frame_cnt from SEI message
607 *
608 * Set to -1 if no recovery point SEI message found or to number of frames
609 * before playback synchronizes. Frames having recovery point are key
610 * frames.
611 */
612 int sei_recovery_frame_cnt;
613
be545b8a
DB
614 int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
615 int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
2ea89d92
IS
616
617 // Timestamp stuff
be545b8a
DB
618 int sei_buffering_period_present; ///< Buffering period SEI flag
619 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
76741b0e
BC
620
621 int cur_chroma_format_idc;
a394959b 622 uint8_t *bipred_scratchpad;
2c541554
AK
623 uint8_t *edge_emu_buffer;
624 int16_t *dc_val_base;
759001c5
AK
625
626 AVBufferPool *qscale_table_pool;
627 AVBufferPool *mb_type_pool;
628 AVBufferPool *motion_val_pool;
629 AVBufferPool *ref_index_pool;
be545b8a 630} H264Context;
26b4fe82 631
be545b8a 632extern const uint8_t ff_h264_chroma_qp[3][QP_MAX_NUM + 1]; ///< One chroma qp table for each supported bit depth (8, 9, 10).
0becb078 633extern const uint16_t ff_h264_mb_sizes[4];
889fce8e 634
1790a5e9
IS
635/**
636 * Decode SEI
637 */
638int ff_h264_decode_sei(H264Context *h);
639
640/**
641 * Decode SPS
642 */
643int ff_h264_decode_seq_parameter_set(H264Context *h);
644
645/**
fe9a3fbe
JG
646 * compute profile from sps
647 */
648int ff_h264_get_profile(SPS *sps);
649
650/**
1790a5e9
IS
651 * Decode PPS
652 */
653int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
654
655/**
49bd8e4b 656 * Decode a network abstraction layer unit.
1790a5e9
IS
657 * @param consumed is the number of bytes used as input
658 * @param length is the length of the array
be545b8a
DB
659 * @param dst_length is the number of decoded bytes FIXME here
660 * or a decode rbsp tailing?
32e543f8 661 * @return decoded bytes, might be src+1 if no escapes
1790a5e9 662 */
be545b8a
DB
663const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
664 int *dst_length, int *consumed, int length);
1790a5e9
IS
665
666/**
be545b8a
DB
667 * Free any data that may have been allocated in the H264 context
668 * like SPS, PPS etc.
15861962 669 */
94ee7da0 670void ff_h264_free_context(H264Context *h);
15861962 671
75dd6938 672/**
49bd8e4b 673 * Reconstruct bitstream slice_type.
75dd6938 674 */
0dc343d4 675int ff_h264_get_slice_type(const H264Context *h);
75dd6938 676
903d58f6 677/**
49bd8e4b 678 * Allocate tables.
903d58f6
MN
679 * needs width/height
680 */
681int ff_h264_alloc_tables(H264Context *h);
682
683/**
49bd8e4b 684 * Fill the default_ref_list.
ea6f00c4
MN
685 */
686int ff_h264_fill_default_ref_list(H264Context *h);
687
688int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
689void ff_h264_fill_mbaff_ref_list(H264Context *h);
690void ff_h264_remove_all_refs(H264Context *h);
691
692/**
49bd8e4b 693 * Execute the reference picture marking (memory management control operations).
ea6f00c4
MN
694 */
695int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
696
bad446e2
RB
697int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
698 int first_slice);
ea6f00c4 699
ea382767 700int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
733f5990 701
ea6f00c4 702/**
be545b8a
DB
703 * Check if the top & left blocks are available if needed & change the
704 * dc mode so it only uses the available blocks.
903d58f6 705 */
2bedc0e8
MN
706int ff_h264_check_intra4x4_pred_mode(H264Context *h);
707
708/**
be545b8a
DB
709 * Check if the top & left blocks are available if needed & change the
710 * dc mode so it only uses the available blocks.
2bedc0e8 711 */
45b7bd7c 712int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
903d58f6 713
903d58f6 714void ff_h264_hl_decode_mb(H264Context *h);
05e95319 715int ff_h264_decode_extradata(H264Context *h);
94ee7da0
DB
716int ff_h264_decode_init(AVCodecContext *avctx);
717void ff_h264_decode_init_vlc(void);
e1e94902
MN
718
719/**
49bd8e4b 720 * Decode a macroblock
be545b8a 721 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
e1e94902
MN
722 */
723int ff_h264_decode_mb_cavlc(H264Context *h);
903d58f6 724
cc51b282 725/**
49bd8e4b 726 * Decode a CABAC coded macroblock
be545b8a 727 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
cc51b282
MN
728 */
729int ff_h264_decode_mb_cabac(H264Context *h);
730
731void ff_h264_init_cabac_states(H264Context *h);
732
be545b8a
DB
733void ff_h264_direct_dist_scale_factor(H264Context *const h);
734void ff_h264_direct_ref_list_init(H264Context *const h);
735void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type);
943f69a6 736
be545b8a
DB
737void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y,
738 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
739 unsigned int linesize, unsigned int uvlinesize);
740void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
741 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
742 unsigned int linesize, unsigned int uvlinesize);
082cf971 743
9c095463
MN
744/**
745 * Reset SEI values at the beginning of the frame.
746 *
747 * @param h H.264 context.
748 */
749void ff_h264_reset_sei(H264Context *h);
750
903d58f6 751/*
be545b8a
DB
752 * o-o o-o
753 * / / /
754 * o-o o-o
755 * ,---'
756 * o-o o-o
757 * / / /
758 * o-o o-o
759 */
5657d140
JGG
760
761/* Scan8 organization:
c90b9442
JGG
762 * 0 1 2 3 4 5 6 7
763 * 0 DY y y y y y
764 * 1 y Y Y Y Y
765 * 2 y Y Y Y Y
766 * 3 y Y Y Y Y
767 * 4 y Y Y Y Y
768 * 5 DU u u u u u
769 * 6 u U U U U
770 * 7 u U U U U
771 * 8 u U U U U
772 * 9 u U U U U
773 * 10 DV v v v v v
774 * 11 v V V V V
775 * 12 v V V V V
776 * 13 v V V V V
777 * 14 v V V V V
5657d140
JGG
778 * DY/DU/DV are for luma/chroma DC.
779 */
780
c90b9442
JGG
781#define LUMA_DC_BLOCK_INDEX 48
782#define CHROMA_DC_BLOCK_INDEX 49
783
be545b8a
DB
784// This table must be here because scan8[constant] must be known at compiletime
785static const uint8_t scan8[16 * 3 + 3] = {
786 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
787 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
788 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
789 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
790 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
791 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
792 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
793 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
794 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
795 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
796 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
797 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
798 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
903d58f6
MN
799};
800
be545b8a
DB
801static av_always_inline uint32_t pack16to32(int a, int b)
802{
903d58f6 803#if HAVE_BIGENDIAN
be545b8a 804 return (b & 0xFFFF) + (a << 16);
903d58f6 805#else
be545b8a 806 return (a & 0xFFFF) + (b << 16);
903d58f6
MN
807#endif
808}
809
be545b8a
DB
810static av_always_inline uint16_t pack8to16(int a, int b)
811{
b5bd0700 812#if HAVE_BIGENDIAN
be545b8a 813 return (b & 0xFF) + (a << 8);
b5bd0700 814#else
be545b8a 815 return (a & 0xFF) + (b << 8);
b5bd0700
MN
816#endif
817}
818
903d58f6 819/**
58c42af7 820 * Get the chroma qp.
082cf971 821 */
be545b8a
DB
822static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
823{
082cf971
MN
824 return h->pps.chroma_qp_table[t][qscale];
825}
826
44a5e7b6 827/**
58c42af7 828 * Get the predicted intra4x4 prediction mode.
e1e94902 829 */
be545b8a
DB
830static av_always_inline int pred_intra_mode(H264Context *h, int n)
831{
832 const int index8 = scan8[n];
833 const int left = h->intra4x4_pred_mode_cache[index8 - 1];
834 const int top = h->intra4x4_pred_mode_cache[index8 - 8];
835 const int min = FFMIN(left, top);
e1e94902 836
2c541554 837 tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min);
e1e94902 838
be545b8a
DB
839 if (min < 0)
840 return DC_PRED;
841 else
842 return min;
e1e94902
MN
843}
844
be545b8a
DB
845static av_always_inline void write_back_intra_pred_mode(H264Context *h)
846{
847 int8_t *i4x4 = h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
848 int8_t *i4x4_cache = h->intra4x4_pred_mode_cache;
3b7ebeb4 849
be545b8a
DB
850 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
851 i4x4[4] = i4x4_cache[7 + 8 * 3];
852 i4x4[5] = i4x4_cache[7 + 8 * 2];
853 i4x4[6] = i4x4_cache[7 + 8 * 1];
3b7ebeb4 854}
e1e94902 855
be545b8a
DB
856static av_always_inline void write_back_non_zero_count(H264Context *h)
857{
858 const int mb_xy = h->mb_xy;
859 uint8_t *nnz = h->non_zero_count[mb_xy];
3b7ebeb4
JGG
860 uint8_t *nnz_cache = h->non_zero_count_cache;
861
be545b8a
DB
862 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
863 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
864 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
865 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
866 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
867 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
868 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
869 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
870
2c541554 871 if (!h->chroma_y_shift) {
be545b8a
DB
872 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
873 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
874 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
875 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
3b7ebeb4
JGG
876 }
877}
878
be545b8a 879static av_always_inline void write_back_motion_list(H264Context *h,
be545b8a
DB
880 int b_stride,
881 int b_xy, int b8_xy,
882 int mb_type, int list)
3b7ebeb4 883{
759001c5 884 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
be545b8a
DB
885 int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
886 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
887 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
888 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
889 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
6d2b6f21 890 if (CABAC(h)) {
be545b8a
DB
891 uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8 * h->mb_xy
892 : h->mb2br_xy[h->mb_xy]];
893 uint8_t(*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
894 if (IS_SKIP(mb_type)) {
3b7ebeb4 895 AV_ZERO128(mvd_dst);
be545b8a
DB
896 } else {
897 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
898 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
899 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
900 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
3b7ebeb4
JGG
901 }
902 }
903
904 {
759001c5 905 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
3b7ebeb4 906 int8_t *ref_cache = h->ref_cache[list];
be545b8a
DB
907 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
908 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
909 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
910 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
c90b9442 911 }
e1e94902
MN
912}
913
be545b8a
DB
914static av_always_inline void write_back_motion(H264Context *h, int mb_type)
915{
be545b8a 916 const int b_stride = h->b_stride;
2c541554 917 const int b_xy = 4 * h->mb_x + 4 * h->mb_y * h->b_stride; // try mb2b(8)_xy
be545b8a 918 const int b8_xy = 4 * h->mb_xy;
e1e94902 919
be545b8a 920 if (USES_LIST(mb_type, 0)) {
2c541554 921 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 0);
be545b8a 922 } else {
759001c5 923 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
657ccb5a 924 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
3b7ebeb4 925 }
be545b8a 926 if (USES_LIST(mb_type, 1))
2c541554 927 write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 1);
e1e94902 928
6d2b6f21 929 if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
be545b8a
DB
930 if (IS_8X8(mb_type)) {
931 uint8_t *direct_table = &h->direct_table[4 * h->mb_xy];
932 direct_table[1] = h->sub_mb_type[1] >> 1;
933 direct_table[2] = h->sub_mb_type[2] >> 1;
934 direct_table[3] = h->sub_mb_type[3] >> 1;
e1e94902
MN
935 }
936 }
937}
938
be545b8a
DB
939static av_always_inline int get_dct8x8_allowed(H264Context *h)
940{
941 if (h->sps.direct_8x8_inference_flag)
942 return !(AV_RN64A(h->sub_mb_type) &
943 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
944 0x0001000100010001ULL));
e1e94902 945 else
be545b8a
DB
946 return !(AV_RN64A(h->sub_mb_type) &
947 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
948 0x0001000100010001ULL));
e1e94902
MN
949}
950
2c541554 951void ff_h264_draw_horiz_band(H264Context *h, int y, int height);
3f1a7ceb 952int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
4baba6c8
YN
953int ff_pred_weight_table(H264Context *h);
954int ff_set_ref_count(H264Context *h);
2c541554 955
98790382 956#endif /* AVCODEC_H264_H */