You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1009 lines
31KB

  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. * @file
  23. * H.264 / AVC / MPEG-4 part10 codec.
  24. * @author Michael Niedermayer <michaelni@gmx.at>
  25. */
  26. #ifndef AVCODEC_H264_H
  27. #define AVCODEC_H264_H
  28. #include "libavutil/buffer.h"
  29. #include "libavutil/intreadwrite.h"
  30. #include "libavutil/thread.h"
  31. #include "cabac.h"
  32. #include "error_resilience.h"
  33. #include "h264_parse.h"
  34. #include "h264_sei.h"
  35. #include "h2645_parse.h"
  36. #include "h264chroma.h"
  37. #include "h264dsp.h"
  38. #include "h264pred.h"
  39. #include "h264qpel.h"
  40. #include "internal.h"
  41. #include "mpegutils.h"
  42. #include "parser.h"
  43. #include "qpeldsp.h"
  44. #include "rectangle.h"
  45. #include "videodsp.h"
  46. #define H264_MAX_PICTURE_COUNT 36
  47. #define MAX_SPS_COUNT 32
  48. #define MAX_PPS_COUNT 256
  49. #define MAX_MMCO_COUNT 66
  50. #define MAX_DELAYED_PIC_COUNT 16
  51. #define MAX_MBPAIR_SIZE (256*1024) // a tighter bound could be calculated if someone cares about a few bytes
  52. /* Compiling in interlaced support reduces the speed
  53. * of progressive decoding by about 2%. */
  54. #define ALLOW_INTERLACE
  55. #define FMO 0
  56. /**
  57. * The maximum number of slices supported by the decoder.
  58. * must be a power of 2
  59. */
  60. #define MAX_SLICES 32
  61. #ifdef ALLOW_INTERLACE
  62. #define MB_MBAFF(h) (h)->mb_mbaff
  63. #define MB_FIELD(sl) (sl)->mb_field_decoding_flag
  64. #define FRAME_MBAFF(h) (h)->mb_aff_frame
  65. #define FIELD_PICTURE(h) ((h)->picture_structure != PICT_FRAME)
  66. #define LEFT_MBS 2
  67. #define LTOP 0
  68. #define LBOT 1
  69. #define LEFT(i) (i)
  70. #else
  71. #define MB_MBAFF(h) 0
  72. #define MB_FIELD(sl) 0
  73. #define FRAME_MBAFF(h) 0
  74. #define FIELD_PICTURE(h) 0
  75. #undef IS_INTERLACED
  76. #define IS_INTERLACED(mb_type) 0
  77. #define LEFT_MBS 1
  78. #define LTOP 0
  79. #define LBOT 0
  80. #define LEFT(i) 0
  81. #endif
  82. #define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h))
  83. #ifndef CABAC
  84. #define CABAC(h) (h)->ps.pps->cabac
  85. #endif
  86. #define CHROMA(h) ((h)->ps.sps->chroma_format_idc)
  87. #define CHROMA422(h) ((h)->ps.sps->chroma_format_idc == 2)
  88. #define CHROMA444(h) ((h)->ps.sps->chroma_format_idc == 3)
  89. #define EXTENDED_SAR 255
  90. #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
  91. #define MB_TYPE_8x8DCT 0x01000000
  92. #define IS_REF0(a) ((a) & MB_TYPE_REF0)
  93. #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
  94. #define QP_MAX_NUM (51 + 6*6) // The maximum supported qp
  95. /* NAL unit types */
  96. enum {
  97. NAL_SLICE = 1,
  98. NAL_DPA = 2,
  99. NAL_DPB = 3,
  100. NAL_DPC = 4,
  101. NAL_IDR_SLICE = 5,
  102. NAL_SEI = 6,
  103. NAL_SPS = 7,
  104. NAL_PPS = 8,
  105. NAL_AUD = 9,
  106. NAL_END_SEQUENCE = 10,
  107. NAL_END_STREAM = 11,
  108. NAL_FILLER_DATA = 12,
  109. NAL_SPS_EXT = 13,
  110. NAL_AUXILIARY_SLICE = 19,
  111. };
  112. /**
  113. * Sequence parameter set
  114. */
  115. typedef struct SPS {
  116. unsigned int sps_id;
  117. int profile_idc;
  118. int level_idc;
  119. int chroma_format_idc;
  120. int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
  121. int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
  122. int poc_type; ///< pic_order_cnt_type
  123. int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
  124. int delta_pic_order_always_zero_flag;
  125. int offset_for_non_ref_pic;
  126. int offset_for_top_to_bottom_field;
  127. int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
  128. int ref_frame_count; ///< num_ref_frames
  129. int gaps_in_frame_num_allowed_flag;
  130. int mb_width; ///< pic_width_in_mbs_minus1 + 1
  131. int mb_height; ///< pic_height_in_map_units_minus1 + 1
  132. int frame_mbs_only_flag;
  133. int mb_aff; ///< mb_adaptive_frame_field_flag
  134. int direct_8x8_inference_flag;
  135. int crop; ///< frame_cropping_flag
  136. /* those 4 are already in luma samples */
  137. unsigned int crop_left; ///< frame_cropping_rect_left_offset
  138. unsigned int crop_right; ///< frame_cropping_rect_right_offset
  139. unsigned int crop_top; ///< frame_cropping_rect_top_offset
  140. unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
  141. int vui_parameters_present_flag;
  142. AVRational sar;
  143. int video_signal_type_present_flag;
  144. int full_range;
  145. int colour_description_present_flag;
  146. enum AVColorPrimaries color_primaries;
  147. enum AVColorTransferCharacteristic color_trc;
  148. enum AVColorSpace colorspace;
  149. int timing_info_present_flag;
  150. uint32_t num_units_in_tick;
  151. uint32_t time_scale;
  152. int fixed_frame_rate_flag;
  153. short offset_for_ref_frame[256]; // FIXME dyn aloc?
  154. int bitstream_restriction_flag;
  155. int num_reorder_frames;
  156. int scaling_matrix_present;
  157. uint8_t scaling_matrix4[6][16];
  158. uint8_t scaling_matrix8[6][64];
  159. int nal_hrd_parameters_present_flag;
  160. int vcl_hrd_parameters_present_flag;
  161. int pic_struct_present_flag;
  162. int time_offset_length;
  163. int cpb_cnt; ///< See H.264 E.1.2
  164. int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
  165. int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
  166. int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
  167. int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
  168. int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
  169. int residual_color_transform_flag; ///< residual_colour_transform_flag
  170. int constraint_set_flags; ///< constraint_set[0-3]_flag
  171. uint8_t data[4096];
  172. size_t data_size;
  173. } SPS;
  174. /**
  175. * Picture parameter set
  176. */
  177. typedef struct PPS {
  178. unsigned int sps_id;
  179. int cabac; ///< entropy_coding_mode_flag
  180. int pic_order_present; ///< pic_order_present_flag
  181. int slice_group_count; ///< num_slice_groups_minus1 + 1
  182. int mb_slice_group_map_type;
  183. unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
  184. int weighted_pred; ///< weighted_pred_flag
  185. int weighted_bipred_idc;
  186. int init_qp; ///< pic_init_qp_minus26 + 26
  187. int init_qs; ///< pic_init_qs_minus26 + 26
  188. int chroma_qp_index_offset[2];
  189. int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
  190. int constrained_intra_pred; ///< constrained_intra_pred_flag
  191. int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
  192. int transform_8x8_mode; ///< transform_8x8_mode_flag
  193. uint8_t scaling_matrix4[6][16];
  194. uint8_t scaling_matrix8[6][64];
  195. uint8_t chroma_qp_table[2][QP_MAX_NUM+1]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
  196. int chroma_qp_diff;
  197. uint8_t data[4096];
  198. size_t data_size;
  199. uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16];
  200. uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
  201. uint32_t(*dequant4_coeff[6])[16];
  202. uint32_t(*dequant8_coeff[6])[64];
  203. } PPS;
  204. typedef struct H264ParamSets {
  205. AVBufferRef *sps_list[MAX_SPS_COUNT];
  206. AVBufferRef *pps_list[MAX_PPS_COUNT];
  207. AVBufferRef *pps_ref;
  208. AVBufferRef *sps_ref;
  209. /* currently active parameters sets */
  210. const PPS *pps;
  211. const SPS *sps;
  212. } H264ParamSets;
  213. /**
  214. * Memory management control operation opcode.
  215. */
  216. typedef enum MMCOOpcode {
  217. MMCO_END = 0,
  218. MMCO_SHORT2UNUSED,
  219. MMCO_LONG2UNUSED,
  220. MMCO_SHORT2LONG,
  221. MMCO_SET_MAX_LONG,
  222. MMCO_RESET,
  223. MMCO_LONG,
  224. } MMCOOpcode;
  225. /**
  226. * Memory management control operation.
  227. */
  228. typedef struct MMCO {
  229. MMCOOpcode opcode;
  230. int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
  231. int long_arg; ///< index, pic_num, or num long refs depending on opcode
  232. } MMCO;
  233. typedef struct H264Picture {
  234. AVFrame *f;
  235. ThreadFrame tf;
  236. AVBufferRef *qscale_table_buf;
  237. int8_t *qscale_table;
  238. AVBufferRef *motion_val_buf[2];
  239. int16_t (*motion_val[2])[2];
  240. AVBufferRef *mb_type_buf;
  241. uint32_t *mb_type;
  242. AVBufferRef *hwaccel_priv_buf;
  243. void *hwaccel_picture_private; ///< hardware accelerator private data
  244. AVBufferRef *ref_index_buf[2];
  245. int8_t *ref_index[2];
  246. int field_poc[2]; ///< top/bottom POC
  247. int poc; ///< frame POC
  248. int frame_num; ///< frame_num (raw frame_num from slice header)
  249. int mmco_reset; /**< MMCO_RESET set this 1. Reordering code must
  250. not mix pictures before and after MMCO_RESET. */
  251. int pic_id; /**< pic_num (short -> no wrap version of pic_num,
  252. pic_num & max_pic_num; long -> long_pic_num) */
  253. int long_ref; ///< 1->long term reference 0->short term reference
  254. int ref_poc[2][2][32]; ///< POCs of the frames/fields used as reference (FIXME need per slice)
  255. int ref_count[2][2]; ///< number of entries in ref_poc (FIXME need per slice)
  256. int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF
  257. int field_picture; ///< whether or not picture was encoded in separate fields
  258. int reference;
  259. int recovered; ///< picture at IDR or recovery point + recovery count
  260. int invalid_gap;
  261. int sei_recovery_frame_cnt;
  262. int crop;
  263. int crop_left;
  264. int crop_top;
  265. } H264Picture;
  266. typedef struct H264Ref {
  267. uint8_t *data[3];
  268. int linesize[3];
  269. int reference;
  270. int poc;
  271. int pic_id;
  272. H264Picture *parent;
  273. } H264Ref;
  274. typedef struct H264SliceContext {
  275. struct H264Context *h264;
  276. GetBitContext gb;
  277. ERContext er;
  278. int slice_num;
  279. int slice_type;
  280. int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
  281. int slice_type_fixed;
  282. int qscale;
  283. int chroma_qp[2]; // QPc
  284. int qp_thresh; ///< QP threshold to skip loopfilter
  285. int last_qscale_diff;
  286. // deblock
  287. int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
  288. int slice_alpha_c0_offset;
  289. int slice_beta_offset;
  290. H264PredWeightTable pwt;
  291. int prev_mb_skipped;
  292. int next_mb_skipped;
  293. int chroma_pred_mode;
  294. int intra16x16_pred_mode;
  295. int8_t intra4x4_pred_mode_cache[5 * 8];
  296. int8_t(*intra4x4_pred_mode);
  297. int topleft_mb_xy;
  298. int top_mb_xy;
  299. int topright_mb_xy;
  300. int left_mb_xy[LEFT_MBS];
  301. int topleft_type;
  302. int top_type;
  303. int topright_type;
  304. int left_type[LEFT_MBS];
  305. const uint8_t *left_block;
  306. int topleft_partition;
  307. unsigned int topleft_samples_available;
  308. unsigned int top_samples_available;
  309. unsigned int topright_samples_available;
  310. unsigned int left_samples_available;
  311. ptrdiff_t linesize, uvlinesize;
  312. ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
  313. ptrdiff_t mb_uvlinesize;
  314. int mb_x, mb_y;
  315. int mb_xy;
  316. int resync_mb_x;
  317. int resync_mb_y;
  318. unsigned int first_mb_addr;
  319. // index of the first MB of the next slice
  320. int next_slice_idx;
  321. int mb_skip_run;
  322. int is_complex;
  323. int picture_structure;
  324. int mb_field_decoding_flag;
  325. int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
  326. int redundant_pic_count;
  327. /**
  328. * number of neighbors (top and/or left) that used 8x8 dct
  329. */
  330. int neighbor_transform_size;
  331. int direct_spatial_mv_pred;
  332. int col_parity;
  333. int col_fieldoff;
  334. int cbp;
  335. int top_cbp;
  336. int left_cbp;
  337. int dist_scale_factor[32];
  338. int dist_scale_factor_field[2][32];
  339. int map_col_to_list0[2][16 + 32];
  340. int map_col_to_list0_field[2][2][16 + 32];
  341. /**
  342. * num_ref_idx_l0/1_active_minus1 + 1
  343. */
  344. unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
  345. unsigned int list_count;
  346. H264Ref ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
  347. * Reordered version of default_ref_list
  348. * according to picture reordering in slice header */
  349. struct {
  350. uint8_t op;
  351. uint32_t val;
  352. } ref_modifications[2][32];
  353. int nb_ref_modifications[2];
  354. unsigned int pps_id;
  355. const uint8_t *intra_pcm_ptr;
  356. int16_t *dc_val_base;
  357. uint8_t *bipred_scratchpad;
  358. uint8_t *edge_emu_buffer;
  359. uint8_t (*top_borders[2])[(16 * 3) * 2];
  360. int bipred_scratchpad_allocated;
  361. int edge_emu_buffer_allocated;
  362. int top_borders_allocated[2];
  363. /**
  364. * non zero coeff count cache.
  365. * is 64 if not available.
  366. */
  367. DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
  368. /**
  369. * Motion vector cache.
  370. */
  371. DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
  372. DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
  373. DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
  374. uint8_t direct_cache[5 * 8];
  375. DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
  376. ///< as a DCT coefficient is int32_t in high depth, we need to reserve twice the space.
  377. DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2];
  378. DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
  379. ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either
  380. ///< check that i is not too large or ensure that there is some unused stuff after mb
  381. int16_t mb_padding[256 * 2];
  382. uint8_t (*mvd_table[2])[2];
  383. /**
  384. * Cabac
  385. */
  386. CABACContext cabac;
  387. uint8_t cabac_state[1024];
  388. int cabac_init_idc;
  389. MMCO mmco[MAX_MMCO_COUNT];
  390. int nb_mmco;
  391. int explicit_ref_marking;
  392. int frame_num;
  393. int poc_lsb;
  394. int delta_poc_bottom;
  395. int delta_poc[2];
  396. int curr_pic_num;
  397. int max_pic_num;
  398. } H264SliceContext;
  399. /**
  400. * H264Context
  401. */
  402. typedef struct H264Context {
  403. const AVClass *class;
  404. AVCodecContext *avctx;
  405. VideoDSPContext vdsp;
  406. H264DSPContext h264dsp;
  407. H264ChromaContext h264chroma;
  408. H264QpelContext h264qpel;
  409. H264Picture DPB[H264_MAX_PICTURE_COUNT];
  410. H264Picture *cur_pic_ptr;
  411. H264Picture cur_pic;
  412. H264Picture last_pic_for_ec;
  413. H264SliceContext *slice_ctx;
  414. int nb_slice_ctx;
  415. H2645Packet pkt;
  416. int pixel_shift; ///< 0 for 8-bit H.264, 1 for high-bit-depth H.264
  417. /* coded dimensions -- 16 * mb w/h */
  418. int width, height;
  419. int chroma_x_shift, chroma_y_shift;
  420. /**
  421. * Backup frame properties: needed, because they can be different
  422. * between returned frame and last decoded frame.
  423. **/
  424. int backup_width;
  425. int backup_height;
  426. enum AVPixelFormat backup_pix_fmt;
  427. int droppable;
  428. int coded_picture_number;
  429. int context_initialized;
  430. int flags;
  431. int workaround_bugs;
  432. /* Set when slice threading is used and at least one slice uses deblocking
  433. * mode 1 (i.e. across slice boundaries). Then we disable the loop filter
  434. * during normal MB decoding and execute it serially at the end.
  435. */
  436. int postpone_filter;
  437. int8_t(*intra4x4_pred_mode);
  438. H264PredContext hpc;
  439. uint8_t (*non_zero_count)[48];
  440. #define LIST_NOT_USED -1 // FIXME rename?
  441. #define PART_NOT_AVAILABLE -2
  442. /**
  443. * block_offset[ 0..23] for frame macroblocks
  444. * block_offset[24..47] for field macroblocks
  445. */
  446. int block_offset[2 * (16 * 3)];
  447. uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
  448. uint32_t *mb2br_xy;
  449. int b_stride; // FIXME use s->b4_stride
  450. uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
  451. // interlacing specific flags
  452. int mb_aff_frame;
  453. int picture_structure;
  454. int first_field;
  455. uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
  456. /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
  457. uint16_t *cbp_table;
  458. /* chroma_pred_mode for i4x4 or i16x16, else 0 */
  459. uint8_t *chroma_pred_mode_table;
  460. uint8_t (*mvd_table[2])[2];
  461. uint8_t *direct_table;
  462. uint8_t zigzag_scan[16];
  463. uint8_t zigzag_scan8x8[64];
  464. uint8_t zigzag_scan8x8_cavlc[64];
  465. uint8_t field_scan[16];
  466. uint8_t field_scan8x8[64];
  467. uint8_t field_scan8x8_cavlc[64];
  468. uint8_t zigzag_scan_q0[16];
  469. uint8_t zigzag_scan8x8_q0[64];
  470. uint8_t zigzag_scan8x8_cavlc_q0[64];
  471. uint8_t field_scan_q0[16];
  472. uint8_t field_scan8x8_q0[64];
  473. uint8_t field_scan8x8_cavlc_q0[64];
  474. int mb_y;
  475. int mb_height, mb_width;
  476. int mb_stride;
  477. int mb_num;
  478. // =============================================================
  479. // Things below are not used in the MB or more inner code
  480. int nal_ref_idc;
  481. int nal_unit_type;
  482. /**
  483. * Used to parse AVC variant of H.264
  484. */
  485. int is_avc; ///< this flag is != 0 if codec is avc1
  486. int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
  487. int bit_depth_luma; ///< luma bit depth from sps to detect changes
  488. int chroma_format_idc; ///< chroma format from sps to detect changes
  489. H264ParamSets ps;
  490. uint16_t *slice_table_base;
  491. H264POCContext poc;
  492. H264Ref default_ref[2];
  493. H264Picture *short_ref[32];
  494. H264Picture *long_ref[32];
  495. H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
  496. int last_pocs[MAX_DELAYED_PIC_COUNT];
  497. H264Picture *next_output_pic;
  498. int next_outputed_poc;
  499. /**
  500. * memory management control operations buffer.
  501. */
  502. MMCO mmco[MAX_MMCO_COUNT];
  503. int nb_mmco;
  504. int mmco_reset;
  505. int explicit_ref_marking;
  506. int long_ref_count; ///< number of actual long term references
  507. int short_ref_count; ///< number of actual short term references
  508. /**
  509. * @name Members for slice based multithreading
  510. * @{
  511. */
  512. /**
  513. * current slice number, used to initialize slice_num of each thread/context
  514. */
  515. int current_slice;
  516. /**
  517. * Max number of threads / contexts.
  518. * This is equal to AVCodecContext.thread_count unless
  519. * multithreaded decoding is impossible, in which case it is
  520. * reduced to 1.
  521. */
  522. int max_contexts;
  523. /**
  524. * 1 if the single thread fallback warning has already been
  525. * displayed, 0 otherwise.
  526. */
  527. int single_decode_warning;
  528. /** @} */
  529. /**
  530. * Complement sei_pic_struct
  531. * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
  532. * However, soft telecined frames may have these values.
  533. * This is used in an attempt to flag soft telecine progressive.
  534. */
  535. int prev_interlaced_frame;
  536. /**
  537. * Are the SEI recovery points looking valid.
  538. */
  539. int valid_recovery_point;
  540. /**
  541. * recovery_frame is the frame_num at which the next frame should
  542. * be fully constructed.
  543. *
  544. * Set to -1 when not expecting a recovery point.
  545. */
  546. int recovery_frame;
  547. /**
  548. * We have seen an IDR, so all the following frames in coded order are correctly
  549. * decodable.
  550. */
  551. #define FRAME_RECOVERED_IDR (1 << 0)
  552. /**
  553. * Sufficient number of frames have been decoded since a SEI recovery point,
  554. * so all the following frames in presentation order are correct.
  555. */
  556. #define FRAME_RECOVERED_SEI (1 << 1)
  557. int frame_recovered; ///< Initial frame has been completely recovered
  558. int has_recovery_point;
  559. int missing_fields;
  560. /* for frame threading, this is set to 1
  561. * after finish_setup() has been called, so we cannot modify
  562. * some context properties (which are supposed to stay constant between
  563. * slices) anymore */
  564. int setup_finished;
  565. int cur_chroma_format_idc;
  566. int cur_bit_depth_luma;
  567. int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
  568. int enable_er;
  569. H264SEIContext sei;
  570. AVBufferPool *qscale_table_pool;
  571. AVBufferPool *mb_type_pool;
  572. AVBufferPool *motion_val_pool;
  573. AVBufferPool *ref_index_pool;
  574. int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
  575. } H264Context;
  576. extern const uint16_t ff_h264_mb_sizes[4];
  577. /**
  578. * Uninit H264 param sets structure.
  579. */
  580. void ff_h264_ps_uninit(H264ParamSets *ps);
  581. /**
  582. * Decode SPS
  583. */
  584. int ff_h264_decode_seq_parameter_set(GetBitContext *gb, AVCodecContext *avctx,
  585. H264ParamSets *ps, int ignore_truncation);
  586. /**
  587. * Decode PPS
  588. */
  589. int ff_h264_decode_picture_parameter_set(GetBitContext *gb, AVCodecContext *avctx,
  590. H264ParamSets *ps, int bit_length);
  591. /**
  592. * Reconstruct bitstream slice_type.
  593. */
  594. int ff_h264_get_slice_type(const H264SliceContext *sl);
  595. /**
  596. * Allocate tables.
  597. * needs width/height
  598. */
  599. int ff_h264_alloc_tables(H264Context *h);
  600. int ff_h264_decode_ref_pic_list_reordering(const H264Context *h, H264SliceContext *sl);
  601. int ff_h264_build_ref_list(H264Context *h, H264SliceContext *sl);
  602. void ff_h264_remove_all_refs(H264Context *h);
  603. /**
  604. * Execute the reference picture marking (memory management control operations).
  605. */
  606. int ff_h264_execute_ref_pic_marking(H264Context *h);
  607. int ff_h264_decode_ref_pic_marking(const H264Context *h, H264SliceContext *sl,
  608. GetBitContext *gb);
  609. void ff_h264_hl_decode_mb(const H264Context *h, H264SliceContext *sl);
  610. int ff_h264_decode_init(AVCodecContext *avctx);
  611. void ff_h264_decode_init_vlc(void);
  612. /**
  613. * Decode a macroblock
  614. * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
  615. */
  616. int ff_h264_decode_mb_cavlc(const H264Context *h, H264SliceContext *sl);
  617. /**
  618. * Decode a CABAC coded macroblock
  619. * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
  620. */
  621. int ff_h264_decode_mb_cabac(const H264Context *h, H264SliceContext *sl);
  622. void ff_h264_init_cabac_states(const H264Context *h, H264SliceContext *sl);
  623. void ff_h264_direct_dist_scale_factor(const H264Context *const h, H264SliceContext *sl);
  624. void ff_h264_direct_ref_list_init(const H264Context *const h, H264SliceContext *sl);
  625. void ff_h264_pred_direct_motion(const H264Context *const h, H264SliceContext *sl,
  626. int *mb_type);
  627. void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
  628. uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
  629. unsigned int linesize, unsigned int uvlinesize);
  630. void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y,
  631. uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
  632. unsigned int linesize, unsigned int uvlinesize);
  633. /*
  634. * o-o o-o
  635. * / / /
  636. * o-o o-o
  637. * ,---'
  638. * o-o o-o
  639. * / / /
  640. * o-o o-o
  641. */
  642. /* Scan8 organization:
  643. * 0 1 2 3 4 5 6 7
  644. * 0 DY y y y y y
  645. * 1 y Y Y Y Y
  646. * 2 y Y Y Y Y
  647. * 3 y Y Y Y Y
  648. * 4 y Y Y Y Y
  649. * 5 DU u u u u u
  650. * 6 u U U U U
  651. * 7 u U U U U
  652. * 8 u U U U U
  653. * 9 u U U U U
  654. * 10 DV v v v v v
  655. * 11 v V V V V
  656. * 12 v V V V V
  657. * 13 v V V V V
  658. * 14 v V V V V
  659. * DY/DU/DV are for luma/chroma DC.
  660. */
  661. #define LUMA_DC_BLOCK_INDEX 48
  662. #define CHROMA_DC_BLOCK_INDEX 49
  663. // This table must be here because scan8[constant] must be known at compiletime
  664. static const uint8_t scan8[16 * 3 + 3] = {
  665. 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
  666. 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
  667. 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
  668. 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
  669. 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
  670. 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
  671. 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
  672. 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
  673. 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
  674. 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
  675. 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
  676. 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
  677. 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
  678. };
  679. static av_always_inline uint32_t pack16to32(unsigned a, unsigned b)
  680. {
  681. #if HAVE_BIGENDIAN
  682. return (b & 0xFFFF) + (a << 16);
  683. #else
  684. return (a & 0xFFFF) + (b << 16);
  685. #endif
  686. }
  687. static av_always_inline uint16_t pack8to16(unsigned a, unsigned b)
  688. {
  689. #if HAVE_BIGENDIAN
  690. return (b & 0xFF) + (a << 8);
  691. #else
  692. return (a & 0xFF) + (b << 8);
  693. #endif
  694. }
  695. /**
  696. * Get the chroma qp.
  697. */
  698. static av_always_inline int get_chroma_qp(const PPS *pps, int t, int qscale)
  699. {
  700. return pps->chroma_qp_table[t][qscale];
  701. }
  702. /**
  703. * Get the predicted intra4x4 prediction mode.
  704. */
  705. static av_always_inline int pred_intra_mode(const H264Context *h,
  706. H264SliceContext *sl, int n)
  707. {
  708. const int index8 = scan8[n];
  709. const int left = sl->intra4x4_pred_mode_cache[index8 - 1];
  710. const int top = sl->intra4x4_pred_mode_cache[index8 - 8];
  711. const int min = FFMIN(left, top);
  712. ff_tlog(h->avctx, "mode:%d %d min:%d\n", left, top, min);
  713. if (min < 0)
  714. return DC_PRED;
  715. else
  716. return min;
  717. }
  718. static av_always_inline void write_back_intra_pred_mode(const H264Context *h,
  719. H264SliceContext *sl)
  720. {
  721. int8_t *i4x4 = sl->intra4x4_pred_mode + h->mb2br_xy[sl->mb_xy];
  722. int8_t *i4x4_cache = sl->intra4x4_pred_mode_cache;
  723. AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
  724. i4x4[4] = i4x4_cache[7 + 8 * 3];
  725. i4x4[5] = i4x4_cache[7 + 8 * 2];
  726. i4x4[6] = i4x4_cache[7 + 8 * 1];
  727. }
  728. static av_always_inline void write_back_non_zero_count(const H264Context *h,
  729. H264SliceContext *sl)
  730. {
  731. const int mb_xy = sl->mb_xy;
  732. uint8_t *nnz = h->non_zero_count[mb_xy];
  733. uint8_t *nnz_cache = sl->non_zero_count_cache;
  734. AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
  735. AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
  736. AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
  737. AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
  738. AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
  739. AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
  740. AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
  741. AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
  742. if (!h->chroma_y_shift) {
  743. AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
  744. AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
  745. AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
  746. AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
  747. }
  748. }
  749. static av_always_inline void write_back_motion_list(const H264Context *h,
  750. H264SliceContext *sl,
  751. int b_stride,
  752. int b_xy, int b8_xy,
  753. int mb_type, int list)
  754. {
  755. int16_t(*mv_dst)[2] = &h->cur_pic.motion_val[list][b_xy];
  756. int16_t(*mv_src)[2] = &sl->mv_cache[list][scan8[0]];
  757. AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
  758. AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
  759. AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
  760. AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
  761. if (CABAC(h)) {
  762. uint8_t (*mvd_dst)[2] = &sl->mvd_table[list][FMO ? 8 * sl->mb_xy
  763. : h->mb2br_xy[sl->mb_xy]];
  764. uint8_t(*mvd_src)[2] = &sl->mvd_cache[list][scan8[0]];
  765. if (IS_SKIP(mb_type)) {
  766. AV_ZERO128(mvd_dst);
  767. } else {
  768. AV_COPY64(mvd_dst, mvd_src + 8 * 3);
  769. AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
  770. AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
  771. AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
  772. }
  773. }
  774. {
  775. int8_t *ref_index = &h->cur_pic.ref_index[list][b8_xy];
  776. int8_t *ref_cache = sl->ref_cache[list];
  777. ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
  778. ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
  779. ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
  780. ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
  781. }
  782. }
  783. static av_always_inline void write_back_motion(const H264Context *h,
  784. H264SliceContext *sl,
  785. int mb_type)
  786. {
  787. const int b_stride = h->b_stride;
  788. const int b_xy = 4 * sl->mb_x + 4 * sl->mb_y * h->b_stride; // try mb2b(8)_xy
  789. const int b8_xy = 4 * sl->mb_xy;
  790. if (USES_LIST(mb_type, 0)) {
  791. write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 0);
  792. } else {
  793. fill_rectangle(&h->cur_pic.ref_index[0][b8_xy],
  794. 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
  795. }
  796. if (USES_LIST(mb_type, 1))
  797. write_back_motion_list(h, sl, b_stride, b_xy, b8_xy, mb_type, 1);
  798. if (sl->slice_type_nos == AV_PICTURE_TYPE_B && CABAC(h)) {
  799. if (IS_8X8(mb_type)) {
  800. uint8_t *direct_table = &h->direct_table[4 * sl->mb_xy];
  801. direct_table[1] = sl->sub_mb_type[1] >> 1;
  802. direct_table[2] = sl->sub_mb_type[2] >> 1;
  803. direct_table[3] = sl->sub_mb_type[3] >> 1;
  804. }
  805. }
  806. }
  807. static av_always_inline int get_dct8x8_allowed(const H264Context *h, H264SliceContext *sl)
  808. {
  809. if (h->ps.sps->direct_8x8_inference_flag)
  810. return !(AV_RN64A(sl->sub_mb_type) &
  811. ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
  812. 0x0001000100010001ULL));
  813. else
  814. return !(AV_RN64A(sl->sub_mb_type) &
  815. ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
  816. 0x0001000100010001ULL));
  817. }
  818. static inline int find_start_code(const uint8_t *buf, int buf_size,
  819. int buf_index, int next_avc)
  820. {
  821. uint32_t state = -1;
  822. buf_index = avpriv_find_start_code(buf + buf_index, buf + next_avc + 1, &state) - buf - 1;
  823. return FFMIN(buf_index, buf_size);
  824. }
  825. int ff_h264_field_end(H264Context *h, H264SliceContext *sl, int in_setup);
  826. int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src);
  827. void ff_h264_unref_picture(H264Context *h, H264Picture *pic);
  828. int ff_h264_slice_context_init(H264Context *h, H264SliceContext *sl);
  829. void ff_h264_draw_horiz_band(const H264Context *h, H264SliceContext *sl, int y, int height);
  830. int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl,
  831. const H2645NAL *nal);
  832. #define SLICE_SINGLETHREAD 1
  833. #define SLICE_SKIPED 2
  834. int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count);
  835. int ff_h264_update_thread_context(AVCodecContext *dst,
  836. const AVCodecContext *src);
  837. void ff_h264_flush_change(H264Context *h);
  838. void ff_h264_free_tables(H264Context *h);
  839. void ff_h264_set_erpic(ERPicture *dst, H264Picture *src);
  840. #endif /* AVCODEC_H264_H */