2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
5 * This file is part of FFmpeg.
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.
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.
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
24 * H.264 / AVC / MPEG4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
28 #ifndef AVCODEC_H264_H
29 #define AVCODEC_H264_H
31 #include "libavutil/buffer.h"
32 #include "libavutil/intreadwrite.h"
33 #include "libavutil/thread.h"
35 #include "error_resilience.h"
37 #include "h264_parse.h"
38 #include "h2645_parse.h"
39 #include "h264chroma.h"
44 #include "mpegutils.h"
47 #include "rectangle.h"
50 #define H264_MAX_PICTURE_COUNT 36
51 #define H264_MAX_THREADS 32
53 #define MAX_SPS_COUNT 32
54 #define MAX_PPS_COUNT 256
56 #define MAX_MMCO_COUNT 66
58 #define MAX_DELAYED_PIC_COUNT 16
60 #define MAX_MBPAIR_SIZE (256*1024) // a tighter bound could be calculated if someone cares about a few bytes
62 /* Compiling in interlaced support reduces the speed
63 * of progressive decoding by about 2%. */
64 #define ALLOW_INTERLACE
69 * The maximum number of slices supported by the decoder.
70 * must be a power of 2
74 #ifdef ALLOW_INTERLACE
75 #define MB_MBAFF(h) (h)->mb_mbaff
76 #define MB_FIELD(sl) (sl)->mb_field_decoding_flag
77 #define FRAME_MBAFF(h) (h)->mb_aff_frame
78 #define FIELD_PICTURE(h) ((h)->picture_structure != PICT_FRAME)
85 #define MB_FIELD(sl) 0
86 #define FRAME_MBAFF(h) 0
87 #define FIELD_PICTURE(h) 0
89 #define IS_INTERLACED(mb_type) 0
95 #define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
98 #define CABAC(h) (h)->ps.pps->cabac
101 #define CHROMA(h) ((h)->ps.sps->chroma_format_idc)
102 #define CHROMA422(h) ((h)->ps.sps->chroma_format_idc == 2)
103 #define CHROMA444(h) ((h)->ps.sps->chroma_format_idc == 3)
105 #define EXTENDED_SAR 255
107 #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
108 #define MB_TYPE_8x8DCT 0x01000000
109 #define IS_REF0(a) ((a) & MB_TYPE_REF0)
110 #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
112 #define QP_MAX_NUM (51 + 6*6) // The maximum supported qp
125 NAL_END_SEQUENCE = 10,
127 NAL_FILLER_DATA = 12,
129 NAL_AUXILIARY_SLICE = 19,
130 NAL_FF_IGNORE = 0xff0f001,
137 SEI_TYPE_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
138 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
139 SEI_TYPE_USER_DATA_REGISTERED = 4, ///< registered user data as specified by Rec. ITU-T T.35
140 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
141 SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
142 SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
143 SEI_TYPE_DISPLAY_ORIENTATION = 47, ///< display orientation
144 SEI_TYPE_GREEN_METADATA = 56 ///< GreenMPEG information
148 * pic_struct in picture timing SEI message
151 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
152 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
153 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
154 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
155 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
156 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
157 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
158 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
159 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
163 * frame_packing_arrangement types
166 SEI_FPA_TYPE_CHECKERBOARD = 0,
167 SEI_FPA_TYPE_INTERLEAVE_COLUMN = 1,
168 SEI_FPA_TYPE_INTERLEAVE_ROW = 2,
169 SEI_FPA_TYPE_SIDE_BY_SIDE = 3,
170 SEI_FPA_TYPE_TOP_BOTTOM = 4,
171 SEI_FPA_TYPE_INTERLEAVE_TEMPORAL = 5,
176 * Sequence parameter set
182 int chroma_format_idc;
183 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
184 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
185 int poc_type; ///< pic_order_cnt_type
186 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
187 int delta_pic_order_always_zero_flag;
188 int offset_for_non_ref_pic;
189 int offset_for_top_to_bottom_field;
190 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
191 int ref_frame_count; ///< num_ref_frames
192 int gaps_in_frame_num_allowed_flag;
193 int mb_width; ///< pic_width_in_mbs_minus1 + 1
194 int mb_height; ///< pic_height_in_map_units_minus1 + 1
195 int frame_mbs_only_flag;
196 int mb_aff; ///< mb_adaptive_frame_field_flag
197 int direct_8x8_inference_flag;
198 int crop; ///< frame_cropping_flag
200 /* those 4 are already in luma samples */
201 unsigned int crop_left; ///< frame_cropping_rect_left_offset
202 unsigned int crop_right; ///< frame_cropping_rect_right_offset
203 unsigned int crop_top; ///< frame_cropping_rect_top_offset
204 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
205 int vui_parameters_present_flag;
207 int video_signal_type_present_flag;
209 int colour_description_present_flag;
210 enum AVColorPrimaries color_primaries;
211 enum AVColorTransferCharacteristic color_trc;
212 enum AVColorSpace colorspace;
213 int timing_info_present_flag;
214 uint32_t num_units_in_tick;
216 int fixed_frame_rate_flag;
217 short offset_for_ref_frame[256]; // FIXME dyn aloc?
218 int bitstream_restriction_flag;
219 int num_reorder_frames;
220 int scaling_matrix_present;
221 uint8_t scaling_matrix4[6][16];
222 uint8_t scaling_matrix8[6][64];
223 int nal_hrd_parameters_present_flag;
224 int vcl_hrd_parameters_present_flag;
225 int pic_struct_present_flag;
226 int time_offset_length;
227 int cpb_cnt; ///< See H.264 E.1.2
228 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
229 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
230 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
231 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
232 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
233 int residual_color_transform_flag; ///< residual_colour_transform_flag
234 int constraint_set_flags; ///< constraint_set[0-3]_flag
240 * Picture parameter set
244 int cabac; ///< entropy_coding_mode_flag
245 int pic_order_present; ///< pic_order_present_flag
246 int slice_group_count; ///< num_slice_groups_minus1 + 1
247 int mb_slice_group_map_type;
248 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
249 int weighted_pred; ///< weighted_pred_flag
250 int weighted_bipred_idc;
251 int init_qp; ///< pic_init_qp_minus26 + 26
252 int init_qs; ///< pic_init_qs_minus26 + 26
253 int chroma_qp_index_offset[2];
254 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
255 int constrained_intra_pred; ///< constrained_intra_pred_flag
256 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
257 int transform_8x8_mode; ///< transform_8x8_mode_flag
258 uint8_t scaling_matrix4[6][16];
259 uint8_t scaling_matrix8[6][64];
260 uint8_t chroma_qp_table[2][QP_MAX_NUM+1]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
265 uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16];
266 uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
267 uint32_t(*dequant4_coeff[6])[16];
268 uint32_t(*dequant8_coeff[6])[64];
271 typedef struct H264ParamSets {
272 AVBufferRef *sps_list[MAX_SPS_COUNT];
273 AVBufferRef *pps_list[MAX_PPS_COUNT];
275 AVBufferRef *pps_ref;
276 AVBufferRef *sps_ref;
277 /* currently active parameters sets */
279 // FIXME this should properly be const
284 * Frame Packing Arrangement Type
287 int frame_packing_arrangement_id;
288 int frame_packing_arrangement_cancel_flag; ///< is previous arrangement canceled, -1 if never received
289 SEI_FpaType frame_packing_arrangement_type;
290 int frame_packing_arrangement_repetition_period;
291 int content_interpretation_type;
292 int quincunx_sampling_flag;
296 * Green MetaData Information Type
298 typedef struct GreenMetaData {
299 uint8_t green_metadata_type;
301 uint16_t num_seconds;
302 uint16_t num_pictures;
303 uint8_t percent_non_zero_macroblocks;
304 uint8_t percent_intra_coded_macroblocks;
305 uint8_t percent_six_tap_filtering;
306 uint8_t percent_alpha_point_deblocking_instance;
307 uint8_t xsd_metric_type;
308 uint16_t xsd_metric_value;
312 * Memory management control operation opcode.
314 typedef enum MMCOOpcode {
325 * Memory management control operation.
327 typedef struct MMCO {
329 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
330 int long_arg; ///< index, pic_num, or num long refs depending on opcode
333 typedef struct H264Picture {
337 AVBufferRef *qscale_table_buf;
338 int8_t *qscale_table;
340 AVBufferRef *motion_val_buf[2];
341 int16_t (*motion_val[2])[2];
343 AVBufferRef *mb_type_buf;
346 AVBufferRef *hwaccel_priv_buf;
347 void *hwaccel_picture_private; ///< hardware accelerator private data
349 AVBufferRef *ref_index_buf[2];
350 int8_t *ref_index[2];
352 int field_poc[2]; ///< top/bottom POC
353 int poc; ///< frame POC
354 int frame_num; ///< frame_num (raw frame_num from slice header)
355 int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
356 not mix pictures before and after MMCO_RESET. */
357 int pic_id; /**< pic_num (short -> no wrap version of pic_num,
358 pic_num & max_pic_num; long -> long_pic_num) */
359 int long_ref; ///< 1->long term reference 0->short term reference
360 int ref_poc[2][2][32]; ///< POCs of the frames/fields used as reference (FIXME need per slice)
361 int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
362 int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
363 int field_picture; ///< whether or not picture was encoded in separate fields
366 int recovered; ///< picture at IDR or recovery point + recovery count
368 int sei_recovery_frame_cnt;
375 typedef struct H264Ref {
386 typedef struct H264SliceContext {
387 struct H264Context *h264;
393 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
394 int slice_type_fixed;
397 int chroma_qp[2]; // QPc
398 int qp_thresh; ///< QP threshold to skip loopfilter
399 int last_qscale_diff;
402 int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
403 int slice_alpha_c0_offset;
404 int slice_beta_offset;
406 H264PredWeightTable pwt;
411 int chroma_pred_mode;
412 int intra16x16_pred_mode;
414 int8_t intra4x4_pred_mode_cache[5 * 8];
415 int8_t(*intra4x4_pred_mode);
420 int left_mb_xy[LEFT_MBS];
425 int left_type[LEFT_MBS];
427 const uint8_t *left_block;
428 int topleft_partition;
430 unsigned int topleft_samples_available;
431 unsigned int top_samples_available;
432 unsigned int topright_samples_available;
433 unsigned int left_samples_available;
435 ptrdiff_t linesize, uvlinesize;
436 ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
437 ptrdiff_t mb_uvlinesize;
443 // index of the first MB of the next slice
448 int mb_field_decoding_flag;
449 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
451 int redundant_pic_count;
454 * number of neighbors (top and/or left) that used 8x8 dct
456 int neighbor_transform_size;
458 int direct_spatial_mv_pred;
466 int dist_scale_factor[32];
467 int dist_scale_factor_field[2][32];
468 int map_col_to_list0[2][16 + 32];
469 int map_col_to_list0_field[2][2][16 + 32];
472 * num_ref_idx_l0/1_active_minus1 + 1
474 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
475 unsigned int list_count;
476 H264Ref ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
477 * Reordered version of default_ref_list
478 * according to picture reordering in slice header */
479 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
481 const uint8_t *intra_pcm_ptr;
482 int16_t *dc_val_base;
484 uint8_t *bipred_scratchpad;
485 uint8_t *edge_emu_buffer;
486 uint8_t (*top_borders[2])[(16 * 3) * 2];
487 int bipred_scratchpad_allocated;
488 int edge_emu_buffer_allocated;
489 int top_borders_allocated[2];
492 * non zero coeff count cache.
493 * is 64 if not available.
495 DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
498 * Motion vector cache.
500 DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
501 DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
502 DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
503 uint8_t direct_cache[5 * 8];
505 DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
507 ///< as a dct coefficient is int32_t in high depth, we need to reserve twice the space.
508 DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2];
509 DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
510 ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either
511 ///< check that i is not too large or ensure that there is some unused stuff after mb
512 int16_t mb_padding[256 * 2];
514 uint8_t (*mvd_table[2])[2];
520 uint8_t cabac_state[1024];
523 // rbsp buffer used for this slice
524 uint8_t *rbsp_buffer;
525 unsigned int rbsp_buffer_size;
531 typedef struct H264Context {
532 const AVClass *class;
533 AVCodecContext *avctx;
534 VideoDSPContext vdsp;
535 H264DSPContext h264dsp;
536 H264ChromaContext h264chroma;
537 H264QpelContext h264qpel;
540 H264Picture DPB[H264_MAX_PICTURE_COUNT];
541 H264Picture *cur_pic_ptr;
543 H264Picture last_pic_for_ec;
545 H264SliceContext *slice_ctx;
550 int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
552 /* coded dimensions -- 16 * mb w/h */
554 int chroma_x_shift, chroma_y_shift;
557 * Backup frame properties: needed, because they can be different
558 * between returned frame and last decoded frame.
562 enum AVPixelFormat backup_pix_fmt;
565 int coded_picture_number;
568 int context_initialized;
572 int8_t(*intra4x4_pred_mode);
575 uint8_t (*non_zero_count)[48];
577 #define LIST_NOT_USED -1 // FIXME rename?
578 #define PART_NOT_AVAILABLE -2
581 * block_offset[ 0..23] for frame macroblocks
582 * block_offset[24..47] for field macroblocks
584 int block_offset[2 * (16 * 3)];
586 uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
588 int b_stride; // FIXME use s->b4_stride
591 unsigned current_sps_id; ///< id of the current SPS
593 int au_pps_id; ///< pps_id of current access unit
595 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
597 // interlacing specific flags
599 int picture_structure;
602 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
604 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
607 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
608 uint8_t *chroma_pred_mode_table;
609 uint8_t (*mvd_table[2])[2];
610 uint8_t *direct_table;
612 uint8_t zigzag_scan[16];
613 uint8_t zigzag_scan8x8[64];
614 uint8_t zigzag_scan8x8_cavlc[64];
615 uint8_t field_scan[16];
616 uint8_t field_scan8x8[64];
617 uint8_t field_scan8x8_cavlc[64];
618 uint8_t zigzag_scan_q0[16];
619 uint8_t zigzag_scan8x8_q0[64];
620 uint8_t zigzag_scan8x8_cavlc_q0[64];
621 uint8_t field_scan_q0[16];
622 uint8_t field_scan8x8_q0[64];
623 uint8_t field_scan8x8_cavlc_q0[64];
628 int mb_height, mb_width;
632 // =============================================================
633 // Things below are not used in the MB or more inner code
639 * Used to parse AVC variant of h264
641 int is_avc; ///< this flag is != 0 if codec is avc1
642 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
644 int bit_depth_luma; ///< luma bit depth from sps to detect changes
645 int chroma_format_idc; ///< chroma format from sps to detect changes
649 uint16_t *slice_table_base;
654 int delta_poc_bottom;
657 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
658 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
659 int frame_num_offset; ///< for POC type 2
660 int prev_frame_num_offset; ///< for POC type 2
661 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
664 * frame_num for frames or 2 * frame_num + 1 for field pics.
669 * max_frame_num or 2 * max_frame_num for field pics.
673 H264Ref default_ref[2];
674 H264Picture *short_ref[32];
675 H264Picture *long_ref[32];
676 H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
677 int last_pocs[MAX_DELAYED_PIC_COUNT];
678 H264Picture *next_output_pic;
679 int next_outputed_poc;
682 * memory management control operations buffer.
684 MMCO mmco[MAX_MMCO_COUNT];
688 int long_ref_count; ///< number of actual long term references
689 int short_ref_count; ///< number of actual short term references
692 * @name Members for slice based multithreading
696 * current slice number, used to initialize slice_num of each thread/context
701 * Max number of threads / contexts.
702 * This is equal to AVCodecContext.thread_count unless
703 * multithreaded decoding is impossible, in which case it is
708 int slice_context_count;
711 * 1 if the single thread fallback warning has already been
712 * displayed, 0 otherwise.
714 int single_decode_warning;
716 enum AVPictureType pict_type;
721 * pic_struct in picture timing SEI message
723 SEI_PicStructType sei_pic_struct;
726 * Complement sei_pic_struct
727 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
728 * However, soft telecined frames may have these values.
729 * This is used in an attempt to flag soft telecine progressive.
731 int prev_interlaced_frame;
734 * frame_packing_arrangment SEI message
736 int sei_frame_packing_present;
737 int frame_packing_arrangement_type;
738 int content_interpretation_type;
739 int quincunx_subsampling;
742 * display orientation SEI message
744 int sei_display_orientation_present;
745 int sei_anticlockwise_rotation;
746 int sei_hflip, sei_vflip;
749 * User data registered by Rec. ITU-T T.35 SEI
751 int sei_reguserdata_afd_present;
752 uint8_t active_format_description;
753 int a53_caption_size;
754 uint8_t *a53_caption;
757 * Bit set of clock types for fields/frames in picture timing SEI message.
758 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
764 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
766 int sei_dpb_output_delay;
769 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
771 int sei_cpb_removal_delay;
774 * recovery_frame_cnt from SEI message
776 * Set to -1 if no recovery point SEI message found or to number of frames
777 * before playback synchronizes. Frames having recovery point are key
780 int sei_recovery_frame_cnt;
783 * Are the SEI recovery points looking valid.
785 int valid_recovery_point;
790 * recovery_frame is the frame_num at which the next frame should
791 * be fully constructed.
793 * Set to -1 when not expecting a recovery point.
798 * We have seen an IDR, so all the following frames in coded order are correctly
801 #define FRAME_RECOVERED_IDR (1 << 0)
803 * Sufficient number of frames have been decoded since a SEI recovery point,
804 * so all the following frames in presentation order are correct.
806 #define FRAME_RECOVERED_SEI (1 << 1)
808 int frame_recovered; ///< Initial frame has been completely recovered
810 int has_recovery_point;
814 /* for frame threading, this is set to 1
815 * after finish_setup() has been called, so we cannot modify
816 * some context properties (which are supposed to stay constant between
821 int sei_buffering_period_present; ///< Buffering period SEI flag
822 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
824 int cur_chroma_format_idc;
825 int cur_bit_depth_luma;
826 int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
828 uint8_t parse_history[6];
829 int parse_history_count;
834 AVBufferPool *qscale_table_pool;
835 AVBufferPool *mb_type_pool;
836 AVBufferPool *motion_val_pool;
837 AVBufferPool *ref_index_pool;
839 /* Motion Estimation */
840 qpel_mc_func (*qpel_put)[16];
841 qpel_mc_func (*qpel_avg)[16];
844 GreenMetaData sei_green_metadata;
848 extern const uint16_t ff_h264_mb_sizes[4];
853 int ff_h264_decode_sei(H264Context *h);
858 int ff_h264_decode_seq_parameter_set(GetBitContext *gb, AVCodecContext *avctx,
859 H264ParamSets *ps, int ignore_truncation);
862 * compute profile from sps
864 int ff_h264_get_profile(const SPS *sps);
869 int ff_h264_decode_picture_parameter_set(GetBitContext *gb, AVCodecContext *avctx,
870 H264ParamSets *ps, int bit_length);
873 * Free any data that may have been allocated in the H264 context
876 void ff_h264_free_context(H264Context *h);
879 * Reconstruct bitstream slice_type.
881 int ff_h264_get_slice_type(const H264SliceContext *sl);
887 int ff_h264_alloc_tables(H264Context *h);
889 int ff_h264_decode_ref_pic_list_reordering(H264Context *h, H264SliceContext *sl);
890 void ff_h264_fill_mbaff_ref_list(H264Context *h, H264SliceContext *sl);
891 void ff_h264_remove_all_refs(H264Context *h);
894 * Execute the reference picture marking (memory management control operations).
896 int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
898 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
901 int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice);
903 void ff_h264_hl_decode_mb(const H264Context *h, H264SliceContext *sl);
904 int ff_h264_decode_extradata(H264Context *h, const uint8_t *buf, int size);
905 int ff_h264_decode_init(AVCodecContext *avctx);
906 void ff_h264_decode_init_vlc(void);
909 * Decode a macroblock
910 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
912 int ff_h264_decode_mb_cavlc(const H264Context *h, H264SliceContext *sl);
915 * Decode a CABAC coded macroblock
916 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
918 int ff_h264_decode_mb_cabac(const H264Context *h, H264SliceContext *sl);
920 void ff_h264_init_cabac_states(const H264Context *h, H264SliceContext *sl);
922 void ff_h264_direct_dist_scale_factor(const H264Context *const h, H264SliceContext *sl);
923 void ff_h264_direct_ref_list_init(const H264Context *const h, H264SliceContext *sl);
924 void ff_h264_pred_direct_motion(const H264Context *const h, H264SliceContext *sl,
927 void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
928 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
929 unsigned int linesize, unsigned int uvlinesize);
930 void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
931 uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
932 unsigned int linesize, unsigned int uvlinesize);
935 * Reset SEI values at the beginning of the frame.
937 * @param h H.264 context.
939 void ff_h264_reset_sei(H264Context *h);
942 * Get stereo_mode string from the h264 frame_packing_arrangement
943 * @param h H.264 context.
945 const char* ff_h264_sei_stereo_mode(H264Context *h);
957 /* Scan8 organization:
974 * DY/DU/DV are for luma/chroma DC.
977 #define LUMA_DC_BLOCK_INDEX 48
978 #define CHROMA_DC_BLOCK_INDEX 49
980 // This table must be here because scan8[constant] must be known at compiletime
981 static const uint8_t scan8[16 * 3 + 3] = {
982 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
983 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
984 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
985 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
986 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
987 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
988 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
989 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
990 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
991 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
992 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
993 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
994 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
997 static av_always_inline uint32_t pack16to32(unsigned a, unsigned b)
1000 return (b & 0xFFFF) + (a << 16);
1002 return (a & 0xFFFF) + (b << 16);
1006 static av_always_inline uint16_t pack8to16(unsigned a, unsigned b)
1009 return (b & 0xFF) + (a << 8);
1011 return (a & 0xFF) + (b << 8);
1016 * Get the chroma qp.
1018 static av_always_inline int get_chroma_qp(const H264Context *h, int t, int qscale)
1020 return h->ps.pps->chroma_qp_table[t][qscale];
1024 * Get the predicted intra4x4 prediction mode.
1026 static av_always_inline int pred_intra_mode(const H264Context *h,
1027 H264SliceContext *sl, int n)
1029 const int index8 = scan8[n];
1030 const int left = sl->intra4x4_pred_mode_cache[index8 - 1];
1031 const int top = sl->intra4x4_pred_mode_cache[index8 - 8];
1032 const int min = FFMIN(left, top);
1034 ff_tlog(h->avctx, "mode:%d %d min:%d\n", left, top, min);
1042 static av_always_inline void write_back_intra_pred_mode(const H264Context *h,
1043 H264SliceContext *sl)
1045 int8_t *i4x4 = sl->intra4x4_pred_mode + h->mb2br_xy[sl->mb_xy];
1046 int8_t *i4x4_cache = sl->intra4x4_pred_mode_cache;
1048 AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
1049 i4x4[4] = i4x4_cache[7 + 8 * 3];
1050 i4x4[5] = i4x4_cache[7 + 8 * 2];
1051 i4x4[6] = i4x4_cache[7 + 8 * 1];
1054 static av_always_inline void write_back_non_zero_count(const H264Context *h,
1055 H264SliceContext *sl)
1057 const int mb_xy = sl->mb_xy;
1058 uint8_t *nnz = h->non_zero_count[mb_xy];
1059 uint8_t *nnz_cache = sl->non_zero_count_cache;
1061 AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
1062 AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
1063 AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
1064 AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
1065 AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
1066 AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
1067 AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
1068 AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
1070 if (!h->chroma_y_shift) {
1071 AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
1072 AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
1073 AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
1074 AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
1078 static av_always_inline void write_back_motion_list(const H264Context *h,
1079 H264SliceContext *sl,
1081 int b_xy, int b8_xy,
1082 int mb_type, int list)
1084 int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
1085 int16_t(*mv_src)[2] = &sl->mv_cache[list][scan8[0]];
1086 AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
1087 AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
1088 AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
1089 AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
1091 uint8_t (*mvd_dst)[2] = &sl->mvd_table[list][FMO ? 8 * sl->mb_xy
1092 : h->mb2br_xy[sl->mb_xy]];
1093 uint8_t(*mvd_src)[2] = &sl->mvd_cache[list][scan8[0]];
1094 if (IS_SKIP(mb_type)) {
1095 AV_ZERO128(mvd_dst);
1097 AV_COPY64(mvd_dst, mvd_src + 8 * 3);
1098 AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
1099 AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
1100 AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
1105 int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
1106 int8_t *ref_cache = sl->ref_cache[list];
1107 ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
1108 ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
1109 ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
1110 ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
1114 static av_always_inline void write_back_motion(const H264Context *h,
1115 H264SliceContext *sl,
1118 const int b_stride = h->b_stride;
1119 const int b_xy = 4 * sl->mb_x + 4 * sl->mb_y * h->b_stride; // try mb2b(8)_xy
1120 const int b8_xy = 4 * sl->mb_xy;
1122 if (USES_LIST(mb_type, 0)) {
1123 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 0);
1125 fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
1126 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
1128 if (USES_LIST(mb_type, 1))
1129 write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 1);
1131 if (sl->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
1132 if (IS_8X8(mb_type)) {
1133 uint8_t *direct_table = &h->direct_table[4 * sl->mb_xy];
1134 direct_table[1] = sl->sub_mb_type[1] >> 1;
1135 direct_table[2] = sl->sub_mb_type[2] >> 1;
1136 direct_table[3] = sl->sub_mb_type[3] >> 1;
1141 static av_always_inline int get_dct8x8_allowed(const H264Context *h, H264SliceContext *sl)
1143 if (h->ps.sps->direct_8x8_inference_flag)
1144 return !(AV_RN64A(sl->sub_mb_type) &
1145 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
1146 0x0001000100010001ULL));
1148 return !(AV_RN64A(sl->sub_mb_type) &
1149 ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
1150 0x0001000100010001ULL));
1153 static inline int find_start_code(const uint8_t *buf, int buf_size,
1154 int buf_index, int next_avc)
1156 uint32_t state = -1;
1158 buf_index = avpriv_find_start_code(buf + buf_index, buf + next_avc + 1, &state) - buf - 1;
1160 return FFMIN(buf_index, buf_size);
1163 static inline int get_avc_nalsize(H264Context *h, const uint8_t *buf,
1164 int buf_size, int *buf_index)
1168 if (*buf_index >= buf_size - h->nal_length_size) {
1169 // the end of the buffer is reached, refill it.
1170 return AVERROR(EAGAIN);
1173 for (i = 0; i < h->nal_length_size; i++)
1174 nalsize = ((unsigned)nalsize << 8) | buf[(*buf_index)++];
1175 if (nalsize <= 0 || nalsize > buf_size - *buf_index) {
1176 av_log(h->avctx, AV_LOG_ERROR,
1177 "AVC: nal size %d\n", nalsize);
1178 return AVERROR_INVALIDDATA;
1183 int ff_h264_field_end(H264Context *h, H264SliceContext *sl, int in_setup);
1185 int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
1186 void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
1188 int ff_h264_slice_context_init(H264Context *h, H264SliceContext *sl);
1190 void ff_h264_draw_horiz_band(const H264Context *h, H264SliceContext *sl, int y, int height);
1191 int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc);
1193 int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl);
1194 #define SLICE_SINGLETHREAD 1
1195 #define SLICE_SKIPED 2
1197 int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count);
1198 int ff_h264_update_thread_context(AVCodecContext *dst,
1199 const AVCodecContext *src);
1201 void ff_h264_flush_change(H264Context *h);
1203 void ff_h264_free_tables(H264Context *h);
1205 void ff_h264_set_erpic(ERPicture *dst, H264Picture *src);
1207 #endif /* AVCODEC_H264_H */