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.

902 lines
28KB

  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 Libav.
  6. *
  7. * Libav 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. * Libav 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 Libav; 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 / MPEG4 part10 codec.
  24. * @author Michael Niedermayer <michaelni@gmx.at>
  25. */
  26. #ifndef AVCODEC_H264_H
  27. #define AVCODEC_H264_H
  28. #include "libavutil/intreadwrite.h"
  29. #include "dsputil.h"
  30. #include "cabac.h"
  31. #include "mpegvideo.h"
  32. #include "h264dsp.h"
  33. #include "h264pred.h"
  34. #include "rectangle.h"
  35. #define interlaced_dct interlaced_dct_is_a_bad_name
  36. #define mb_intra mb_intra_is_not_initialized_see_mb_type
  37. #define MAX_SPS_COUNT 32
  38. #define MAX_PPS_COUNT 256
  39. #define MAX_MMCO_COUNT 66
  40. #define MAX_DELAYED_PIC_COUNT 16
  41. /* Compiling in interlaced support reduces the speed
  42. * of progressive decoding by about 2%. */
  43. #define ALLOW_INTERLACE
  44. #define FMO 0
  45. /**
  46. * The maximum number of slices supported by the decoder.
  47. * must be a power of 2
  48. */
  49. #define MAX_SLICES 16
  50. #ifdef ALLOW_INTERLACE
  51. #define MB_MBAFF h->mb_mbaff
  52. #define MB_FIELD h->mb_field_decoding_flag
  53. #define FRAME_MBAFF h->mb_aff_frame
  54. #define FIELD_PICTURE (s->picture_structure != PICT_FRAME)
  55. #define LEFT_MBS 2
  56. #define LTOP 0
  57. #define LBOT 1
  58. #define LEFT(i) (i)
  59. #else
  60. #define MB_MBAFF 0
  61. #define MB_FIELD 0
  62. #define FRAME_MBAFF 0
  63. #define FIELD_PICTURE 0
  64. #undef IS_INTERLACED
  65. #define IS_INTERLACED(mb_type) 0
  66. #define LEFT_MBS 1
  67. #define LTOP 0
  68. #define LBOT 0
  69. #define LEFT(i) 0
  70. #endif
  71. #define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
  72. #ifndef CABAC
  73. #define CABAC h->pps.cabac
  74. #endif
  75. #define CHROMA422 (h->sps.chroma_format_idc == 2)
  76. #define CHROMA444 (h->sps.chroma_format_idc == 3)
  77. #define EXTENDED_SAR 255
  78. #define MB_TYPE_REF0 MB_TYPE_ACPRED // dirty but it fits in 16 bit
  79. #define MB_TYPE_8x8DCT 0x01000000
  80. #define IS_REF0(a) ((a) & MB_TYPE_REF0)
  81. #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
  82. #define QP_MAX_NUM (51 + 2 * 6) // The maximum supported qp
  83. /* NAL unit types */
  84. enum {
  85. NAL_SLICE = 1,
  86. NAL_DPA,
  87. NAL_DPB,
  88. NAL_DPC,
  89. NAL_IDR_SLICE,
  90. NAL_SEI,
  91. NAL_SPS,
  92. NAL_PPS,
  93. NAL_AUD,
  94. NAL_END_SEQUENCE,
  95. NAL_END_STREAM,
  96. NAL_FILLER_DATA,
  97. NAL_SPS_EXT,
  98. NAL_AUXILIARY_SLICE = 19
  99. };
  100. /**
  101. * SEI message types
  102. */
  103. typedef enum {
  104. SEI_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
  105. SEI_TYPE_PIC_TIMING = 1, ///< picture timing
  106. SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
  107. SEI_TYPE_RECOVERY_POINT = 6 ///< recovery point (frame # to decoder sync)
  108. } SEI_Type;
  109. /**
  110. * pic_struct in picture timing SEI message
  111. */
  112. typedef enum {
  113. SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
  114. SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
  115. SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
  116. SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
  117. SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
  118. SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
  119. SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
  120. SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
  121. SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
  122. } SEI_PicStructType;
  123. /**
  124. * Sequence parameter set
  125. */
  126. typedef struct SPS {
  127. int profile_idc;
  128. int level_idc;
  129. int chroma_format_idc;
  130. int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
  131. int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
  132. int poc_type; ///< pic_order_cnt_type
  133. int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
  134. int delta_pic_order_always_zero_flag;
  135. int offset_for_non_ref_pic;
  136. int offset_for_top_to_bottom_field;
  137. int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
  138. int ref_frame_count; ///< num_ref_frames
  139. int gaps_in_frame_num_allowed_flag;
  140. int mb_width; ///< pic_width_in_mbs_minus1 + 1
  141. int mb_height; ///< pic_height_in_map_units_minus1 + 1
  142. int frame_mbs_only_flag;
  143. int mb_aff; ///< mb_adaptive_frame_field_flag
  144. int direct_8x8_inference_flag;
  145. int crop; ///< frame_cropping_flag
  146. unsigned int crop_left; ///< frame_cropping_rect_left_offset
  147. unsigned int crop_right; ///< frame_cropping_rect_right_offset
  148. unsigned int crop_top; ///< frame_cropping_rect_top_offset
  149. unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
  150. int vui_parameters_present_flag;
  151. AVRational sar;
  152. int video_signal_type_present_flag;
  153. int full_range;
  154. int colour_description_present_flag;
  155. enum AVColorPrimaries color_primaries;
  156. enum AVColorTransferCharacteristic color_trc;
  157. enum AVColorSpace colorspace;
  158. int timing_info_present_flag;
  159. uint32_t num_units_in_tick;
  160. uint32_t time_scale;
  161. int fixed_frame_rate_flag;
  162. short offset_for_ref_frame[256]; // FIXME dyn aloc?
  163. int bitstream_restriction_flag;
  164. int num_reorder_frames;
  165. int scaling_matrix_present;
  166. uint8_t scaling_matrix4[6][16];
  167. uint8_t scaling_matrix8[6][64];
  168. int nal_hrd_parameters_present_flag;
  169. int vcl_hrd_parameters_present_flag;
  170. int pic_struct_present_flag;
  171. int time_offset_length;
  172. int cpb_cnt; ///< See H.264 E.1.2
  173. int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 + 1
  174. int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
  175. int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
  176. int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
  177. int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
  178. int residual_color_transform_flag; ///< residual_colour_transform_flag
  179. int constraint_set_flags; ///< constraint_set[0-3]_flag
  180. int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS
  181. } SPS;
  182. /**
  183. * Picture parameter set
  184. */
  185. typedef struct PPS {
  186. unsigned int sps_id;
  187. int cabac; ///< entropy_coding_mode_flag
  188. int pic_order_present; ///< pic_order_present_flag
  189. int slice_group_count; ///< num_slice_groups_minus1 + 1
  190. int mb_slice_group_map_type;
  191. unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
  192. int weighted_pred; ///< weighted_pred_flag
  193. int weighted_bipred_idc;
  194. int init_qp; ///< pic_init_qp_minus26 + 26
  195. int init_qs; ///< pic_init_qs_minus26 + 26
  196. int chroma_qp_index_offset[2];
  197. int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
  198. int constrained_intra_pred; ///< constrained_intra_pred_flag
  199. int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
  200. int transform_8x8_mode; ///< transform_8x8_mode_flag
  201. uint8_t scaling_matrix4[6][16];
  202. uint8_t scaling_matrix8[6][64];
  203. uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
  204. int chroma_qp_diff;
  205. } PPS;
  206. /**
  207. * Memory management control operation opcode.
  208. */
  209. typedef enum MMCOOpcode {
  210. MMCO_END = 0,
  211. MMCO_SHORT2UNUSED,
  212. MMCO_LONG2UNUSED,
  213. MMCO_SHORT2LONG,
  214. MMCO_SET_MAX_LONG,
  215. MMCO_RESET,
  216. MMCO_LONG,
  217. } MMCOOpcode;
  218. /**
  219. * Memory management control operation.
  220. */
  221. typedef struct MMCO {
  222. MMCOOpcode opcode;
  223. int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
  224. int long_arg; ///< index, pic_num, or num long refs depending on opcode
  225. } MMCO;
  226. /**
  227. * H264Context
  228. */
  229. typedef struct H264Context {
  230. MpegEncContext s;
  231. H264DSPContext h264dsp;
  232. int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
  233. int chroma_qp[2]; // QPc
  234. int qp_thresh; ///< QP threshold to skip loopfilter
  235. int prev_mb_skipped;
  236. int next_mb_skipped;
  237. // prediction stuff
  238. int chroma_pred_mode;
  239. int intra16x16_pred_mode;
  240. int topleft_mb_xy;
  241. int top_mb_xy;
  242. int topright_mb_xy;
  243. int left_mb_xy[LEFT_MBS];
  244. int topleft_type;
  245. int top_type;
  246. int topright_type;
  247. int left_type[LEFT_MBS];
  248. const uint8_t *left_block;
  249. int topleft_partition;
  250. int8_t intra4x4_pred_mode_cache[5 * 8];
  251. int8_t(*intra4x4_pred_mode);
  252. H264PredContext hpc;
  253. unsigned int topleft_samples_available;
  254. unsigned int top_samples_available;
  255. unsigned int topright_samples_available;
  256. unsigned int left_samples_available;
  257. uint8_t (*top_borders[2])[(16 * 3) * 2];
  258. /**
  259. * non zero coeff count cache.
  260. * is 64 if not available.
  261. */
  262. DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15 * 8];
  263. uint8_t (*non_zero_count)[48];
  264. /**
  265. * Motion vector cache.
  266. */
  267. DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5 * 8][2];
  268. DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];
  269. #define LIST_NOT_USED -1 // FIXME rename?
  270. #define PART_NOT_AVAILABLE -2
  271. /**
  272. * number of neighbors (top and/or left) that used 8x8 dct
  273. */
  274. int neighbor_transform_size;
  275. /**
  276. * block_offset[ 0..23] for frame macroblocks
  277. * block_offset[24..47] for field macroblocks
  278. */
  279. int block_offset[2 * (16 * 3)];
  280. uint32_t *mb2b_xy; // FIXME are these 4 a good idea?
  281. uint32_t *mb2br_xy;
  282. int b_stride; // FIXME use s->b4_stride
  283. int mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff
  284. int mb_uvlinesize;
  285. int emu_edge_width;
  286. int emu_edge_height;
  287. unsigned current_sps_id; ///< id of the current SPS
  288. SPS sps; ///< current sps
  289. /**
  290. * current pps
  291. */
  292. PPS pps; // FIXME move to Picture perhaps? (->no) do we need that?
  293. uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down?
  294. uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64];
  295. uint32_t(*dequant4_coeff[6])[16];
  296. uint32_t(*dequant8_coeff[6])[64];
  297. int slice_num;
  298. uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
  299. int slice_type;
  300. int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
  301. int slice_type_fixed;
  302. // interlacing specific flags
  303. int mb_aff_frame;
  304. int mb_field_decoding_flag;
  305. int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
  306. DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
  307. // Weighted pred stuff
  308. int use_weight;
  309. int use_weight_chroma;
  310. int luma_log2_weight_denom;
  311. int chroma_log2_weight_denom;
  312. // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
  313. int luma_weight[48][2][2];
  314. int chroma_weight[48][2][2][2];
  315. int implicit_weight[48][48][2];
  316. int direct_spatial_mv_pred;
  317. int col_parity;
  318. int col_fieldoff;
  319. int dist_scale_factor[16];
  320. int dist_scale_factor_field[2][32];
  321. int map_col_to_list0[2][16 + 32];
  322. int map_col_to_list0_field[2][2][16 + 32];
  323. /**
  324. * num_ref_idx_l0/1_active_minus1 + 1
  325. */
  326. unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
  327. unsigned int list_count;
  328. uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
  329. Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
  330. * Reordered version of default_ref_list
  331. * according to picture reordering in slice header */
  332. int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
  333. // data partitioning
  334. GetBitContext intra_gb;
  335. GetBitContext inter_gb;
  336. GetBitContext *intra_gb_ptr;
  337. GetBitContext *inter_gb_ptr;
  338. DECLARE_ALIGNED(16, DCTELEM, mb)[16 * 48 * 2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
  339. DECLARE_ALIGNED(16, DCTELEM, mb_luma_dc)[3][16 * 2];
  340. DCTELEM 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
  341. /**
  342. * Cabac
  343. */
  344. CABACContext cabac;
  345. uint8_t cabac_state[1024];
  346. /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0, 1, 2), 0x0? luma_cbp */
  347. uint16_t *cbp_table;
  348. int cbp;
  349. int top_cbp;
  350. int left_cbp;
  351. /* chroma_pred_mode for i4x4 or i16x16, else 0 */
  352. uint8_t *chroma_pred_mode_table;
  353. int last_qscale_diff;
  354. uint8_t (*mvd_table[2])[2];
  355. DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5 * 8][2];
  356. uint8_t *direct_table;
  357. uint8_t direct_cache[5 * 8];
  358. uint8_t zigzag_scan[16];
  359. uint8_t zigzag_scan8x8[64];
  360. uint8_t zigzag_scan8x8_cavlc[64];
  361. uint8_t field_scan[16];
  362. uint8_t field_scan8x8[64];
  363. uint8_t field_scan8x8_cavlc[64];
  364. const uint8_t *zigzag_scan_q0;
  365. const uint8_t *zigzag_scan8x8_q0;
  366. const uint8_t *zigzag_scan8x8_cavlc_q0;
  367. const uint8_t *field_scan_q0;
  368. const uint8_t *field_scan8x8_q0;
  369. const uint8_t *field_scan8x8_cavlc_q0;
  370. int x264_build;
  371. int mb_xy;
  372. int is_complex;
  373. // deblock
  374. int deblocking_filter; ///< disable_deblocking_filter_idc with 1 <-> 0
  375. int slice_alpha_c0_offset;
  376. int slice_beta_offset;
  377. // =============================================================
  378. // Things below are not used in the MB or more inner code
  379. int nal_ref_idc;
  380. int nal_unit_type;
  381. uint8_t *rbsp_buffer[2];
  382. unsigned int rbsp_buffer_size[2];
  383. /**
  384. * Used to parse AVC variant of h264
  385. */
  386. int is_avc; ///< this flag is != 0 if codec is avc1
  387. int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
  388. int got_first; ///< this flag is != 0 if we've parsed a frame
  389. SPS *sps_buffers[MAX_SPS_COUNT];
  390. PPS *pps_buffers[MAX_PPS_COUNT];
  391. int dequant_coeff_pps; ///< reinit tables when pps changes
  392. uint16_t *slice_table_base;
  393. // POC stuff
  394. int poc_lsb;
  395. int poc_msb;
  396. int delta_poc_bottom;
  397. int delta_poc[2];
  398. int frame_num;
  399. int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
  400. int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
  401. int frame_num_offset; ///< for POC type 2
  402. int prev_frame_num_offset; ///< for POC type 2
  403. int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
  404. /**
  405. * frame_num for frames or 2 * frame_num + 1 for field pics.
  406. */
  407. int curr_pic_num;
  408. /**
  409. * max_frame_num or 2 * max_frame_num for field pics.
  410. */
  411. int max_pic_num;
  412. int redundant_pic_count;
  413. Picture *short_ref[32];
  414. Picture *long_ref[32];
  415. Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
  416. Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
  417. int last_pocs[MAX_DELAYED_PIC_COUNT];
  418. Picture *next_output_pic;
  419. int outputed_poc;
  420. int next_outputed_poc;
  421. /**
  422. * memory management control operations buffer.
  423. */
  424. MMCO mmco[MAX_MMCO_COUNT];
  425. int mmco_index;
  426. int mmco_reset;
  427. int long_ref_count; ///< number of actual long term references
  428. int short_ref_count; ///< number of actual short term references
  429. int cabac_init_idc;
  430. /**
  431. * @name Members for slice based multithreading
  432. * @{
  433. */
  434. struct H264Context *thread_context[MAX_THREADS];
  435. /**
  436. * current slice number, used to initalize slice_num of each thread/context
  437. */
  438. int current_slice;
  439. /**
  440. * Max number of threads / contexts.
  441. * This is equal to AVCodecContext.thread_count unless
  442. * multithreaded decoding is impossible, in which case it is
  443. * reduced to 1.
  444. */
  445. int max_contexts;
  446. /**
  447. * 1 if the single thread fallback warning has already been
  448. * displayed, 0 otherwise.
  449. */
  450. int single_decode_warning;
  451. int last_slice_type;
  452. /** @} */
  453. /**
  454. * pic_struct in picture timing SEI message
  455. */
  456. SEI_PicStructType sei_pic_struct;
  457. /**
  458. * Complement sei_pic_struct
  459. * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
  460. * However, soft telecined frames may have these values.
  461. * This is used in an attempt to flag soft telecine progressive.
  462. */
  463. int prev_interlaced_frame;
  464. /**
  465. * Bit set of clock types for fields/frames in picture timing SEI message.
  466. * For each found ct_type, appropriate bit is set (e.g., bit 1 for
  467. * interlaced).
  468. */
  469. int sei_ct_type;
  470. /**
  471. * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
  472. */
  473. int sei_dpb_output_delay;
  474. /**
  475. * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
  476. */
  477. int sei_cpb_removal_delay;
  478. /**
  479. * recovery_frame_cnt from SEI message
  480. *
  481. * Set to -1 if no recovery point SEI message found or to number of frames
  482. * before playback synchronizes. Frames having recovery point are key
  483. * frames.
  484. */
  485. int sei_recovery_frame_cnt;
  486. int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
  487. int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
  488. // Timestamp stuff
  489. int sei_buffering_period_present; ///< Buffering period SEI flag
  490. int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
  491. int cur_chroma_format_idc;
  492. uint8_t *bipred_scratchpad;
  493. } H264Context;
  494. extern const uint8_t ff_h264_chroma_qp[3][QP_MAX_NUM + 1]; ///< One chroma qp table for each supported bit depth (8, 9, 10).
  495. extern const uint16_t ff_h264_mb_sizes[4];
  496. /**
  497. * Decode SEI
  498. */
  499. int ff_h264_decode_sei(H264Context *h);
  500. /**
  501. * Decode SPS
  502. */
  503. int ff_h264_decode_seq_parameter_set(H264Context *h);
  504. /**
  505. * compute profile from sps
  506. */
  507. int ff_h264_get_profile(SPS *sps);
  508. /**
  509. * Decode PPS
  510. */
  511. int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
  512. /**
  513. * Decode a network abstraction layer unit.
  514. * @param consumed is the number of bytes used as input
  515. * @param length is the length of the array
  516. * @param dst_length is the number of decoded bytes FIXME here
  517. * or a decode rbsp tailing?
  518. * @return decoded bytes, might be src+1 if no escapes
  519. */
  520. const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
  521. int *dst_length, int *consumed, int length);
  522. /**
  523. * Free any data that may have been allocated in the H264 context
  524. * like SPS, PPS etc.
  525. */
  526. av_cold void ff_h264_free_context(H264Context *h);
  527. /**
  528. * Reconstruct bitstream slice_type.
  529. */
  530. int ff_h264_get_slice_type(const H264Context *h);
  531. /**
  532. * Allocate tables.
  533. * needs width/height
  534. */
  535. int ff_h264_alloc_tables(H264Context *h);
  536. /**
  537. * Fill the default_ref_list.
  538. */
  539. int ff_h264_fill_default_ref_list(H264Context *h);
  540. int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
  541. void ff_h264_fill_mbaff_ref_list(H264Context *h);
  542. void ff_h264_remove_all_refs(H264Context *h);
  543. /**
  544. * Execute the reference picture marking (memory management control operations).
  545. */
  546. int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
  547. int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb);
  548. void ff_generate_sliding_window_mmcos(H264Context *h);
  549. /**
  550. * Check if the top & left blocks are available if needed & change the
  551. * dc mode so it only uses the available blocks.
  552. */
  553. int ff_h264_check_intra4x4_pred_mode(H264Context *h);
  554. /**
  555. * Check if the top & left blocks are available if needed & change the
  556. * dc mode so it only uses the available blocks.
  557. */
  558. int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma);
  559. void ff_h264_hl_decode_mb(H264Context *h);
  560. int ff_h264_frame_start(H264Context *h);
  561. int ff_h264_decode_extradata(H264Context *h);
  562. av_cold int ff_h264_decode_init(AVCodecContext *avctx);
  563. av_cold void ff_h264_decode_init_vlc(void);
  564. /**
  565. * Decode a macroblock
  566. * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
  567. */
  568. int ff_h264_decode_mb_cavlc(H264Context *h);
  569. /**
  570. * Decode a CABAC coded macroblock
  571. * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error
  572. */
  573. int ff_h264_decode_mb_cabac(H264Context *h);
  574. void ff_h264_init_cabac_states(H264Context *h);
  575. void ff_h264_direct_dist_scale_factor(H264Context *const h);
  576. void ff_h264_direct_ref_list_init(H264Context *const h);
  577. void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type);
  578. void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y,
  579. uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
  580. unsigned int linesize, unsigned int uvlinesize);
  581. void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
  582. uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
  583. unsigned int linesize, unsigned int uvlinesize);
  584. /**
  585. * Reset SEI values at the beginning of the frame.
  586. *
  587. * @param h H.264 context.
  588. */
  589. void ff_h264_reset_sei(H264Context *h);
  590. /*
  591. * o-o o-o
  592. * / / /
  593. * o-o o-o
  594. * ,---'
  595. * o-o o-o
  596. * / / /
  597. * o-o o-o
  598. */
  599. /* Scan8 organization:
  600. * 0 1 2 3 4 5 6 7
  601. * 0 DY y y y y y
  602. * 1 y Y Y Y Y
  603. * 2 y Y Y Y Y
  604. * 3 y Y Y Y Y
  605. * 4 y Y Y Y Y
  606. * 5 DU u u u u u
  607. * 6 u U U U U
  608. * 7 u U U U U
  609. * 8 u U U U U
  610. * 9 u U U U U
  611. * 10 DV v v v v v
  612. * 11 v V V V V
  613. * 12 v V V V V
  614. * 13 v V V V V
  615. * 14 v V V V V
  616. * DY/DU/DV are for luma/chroma DC.
  617. */
  618. #define LUMA_DC_BLOCK_INDEX 48
  619. #define CHROMA_DC_BLOCK_INDEX 49
  620. // This table must be here because scan8[constant] must be known at compiletime
  621. static const uint8_t scan8[16 * 3 + 3] = {
  622. 4 + 1 * 8, 5 + 1 * 8, 4 + 2 * 8, 5 + 2 * 8,
  623. 6 + 1 * 8, 7 + 1 * 8, 6 + 2 * 8, 7 + 2 * 8,
  624. 4 + 3 * 8, 5 + 3 * 8, 4 + 4 * 8, 5 + 4 * 8,
  625. 6 + 3 * 8, 7 + 3 * 8, 6 + 4 * 8, 7 + 4 * 8,
  626. 4 + 6 * 8, 5 + 6 * 8, 4 + 7 * 8, 5 + 7 * 8,
  627. 6 + 6 * 8, 7 + 6 * 8, 6 + 7 * 8, 7 + 7 * 8,
  628. 4 + 8 * 8, 5 + 8 * 8, 4 + 9 * 8, 5 + 9 * 8,
  629. 6 + 8 * 8, 7 + 8 * 8, 6 + 9 * 8, 7 + 9 * 8,
  630. 4 + 11 * 8, 5 + 11 * 8, 4 + 12 * 8, 5 + 12 * 8,
  631. 6 + 11 * 8, 7 + 11 * 8, 6 + 12 * 8, 7 + 12 * 8,
  632. 4 + 13 * 8, 5 + 13 * 8, 4 + 14 * 8, 5 + 14 * 8,
  633. 6 + 13 * 8, 7 + 13 * 8, 6 + 14 * 8, 7 + 14 * 8,
  634. 0 + 0 * 8, 0 + 5 * 8, 0 + 10 * 8
  635. };
  636. static av_always_inline uint32_t pack16to32(int a, int b)
  637. {
  638. #if HAVE_BIGENDIAN
  639. return (b & 0xFFFF) + (a << 16);
  640. #else
  641. return (a & 0xFFFF) + (b << 16);
  642. #endif
  643. }
  644. static av_always_inline uint16_t pack8to16(int a, int b)
  645. {
  646. #if HAVE_BIGENDIAN
  647. return (b & 0xFF) + (a << 8);
  648. #else
  649. return (a & 0xFF) + (b << 8);
  650. #endif
  651. }
  652. /**
  653. * Get the chroma qp.
  654. */
  655. static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
  656. {
  657. return h->pps.chroma_qp_table[t][qscale];
  658. }
  659. /**
  660. * Get the predicted intra4x4 prediction mode.
  661. */
  662. static av_always_inline int pred_intra_mode(H264Context *h, int n)
  663. {
  664. const int index8 = scan8[n];
  665. const int left = h->intra4x4_pred_mode_cache[index8 - 1];
  666. const int top = h->intra4x4_pred_mode_cache[index8 - 8];
  667. const int min = FFMIN(left, top);
  668. tprintf(h->s.avctx, "mode:%d %d min:%d\n", left, top, min);
  669. if (min < 0)
  670. return DC_PRED;
  671. else
  672. return min;
  673. }
  674. static av_always_inline void write_back_intra_pred_mode(H264Context *h)
  675. {
  676. int8_t *i4x4 = h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
  677. int8_t *i4x4_cache = h->intra4x4_pred_mode_cache;
  678. AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
  679. i4x4[4] = i4x4_cache[7 + 8 * 3];
  680. i4x4[5] = i4x4_cache[7 + 8 * 2];
  681. i4x4[6] = i4x4_cache[7 + 8 * 1];
  682. }
  683. static av_always_inline void write_back_non_zero_count(H264Context *h)
  684. {
  685. const int mb_xy = h->mb_xy;
  686. uint8_t *nnz = h->non_zero_count[mb_xy];
  687. uint8_t *nnz_cache = h->non_zero_count_cache;
  688. AV_COPY32(&nnz[ 0], &nnz_cache[4 + 8 * 1]);
  689. AV_COPY32(&nnz[ 4], &nnz_cache[4 + 8 * 2]);
  690. AV_COPY32(&nnz[ 8], &nnz_cache[4 + 8 * 3]);
  691. AV_COPY32(&nnz[12], &nnz_cache[4 + 8 * 4]);
  692. AV_COPY32(&nnz[16], &nnz_cache[4 + 8 * 6]);
  693. AV_COPY32(&nnz[20], &nnz_cache[4 + 8 * 7]);
  694. AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
  695. AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
  696. if (!h->s.chroma_y_shift) {
  697. AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
  698. AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
  699. AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
  700. AV_COPY32(&nnz[44], &nnz_cache[4 + 8 * 14]);
  701. }
  702. }
  703. static av_always_inline void write_back_motion_list(H264Context *h,
  704. MpegEncContext *const s,
  705. int b_stride,
  706. int b_xy, int b8_xy,
  707. int mb_type, int list)
  708. {
  709. int16_t(*mv_dst)[2] = &s->current_picture.f.motion_val[list][b_xy];
  710. int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
  711. AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
  712. AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
  713. AV_COPY128(mv_dst + 2 * b_stride, mv_src + 8 * 2);
  714. AV_COPY128(mv_dst + 3 * b_stride, mv_src + 8 * 3);
  715. if (CABAC) {
  716. uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8 * h->mb_xy
  717. : h->mb2br_xy[h->mb_xy]];
  718. uint8_t(*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
  719. if (IS_SKIP(mb_type)) {
  720. AV_ZERO128(mvd_dst);
  721. } else {
  722. AV_COPY64(mvd_dst, mvd_src + 8 * 3);
  723. AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8 * 0);
  724. AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8 * 1);
  725. AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8 * 2);
  726. }
  727. }
  728. {
  729. int8_t *ref_index = &s->current_picture.f.ref_index[list][b8_xy];
  730. int8_t *ref_cache = h->ref_cache[list];
  731. ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
  732. ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
  733. ref_index[0 + 1 * 2] = ref_cache[scan8[8]];
  734. ref_index[1 + 1 * 2] = ref_cache[scan8[12]];
  735. }
  736. }
  737. static av_always_inline void write_back_motion(H264Context *h, int mb_type)
  738. {
  739. MpegEncContext *const s = &h->s;
  740. const int b_stride = h->b_stride;
  741. const int b_xy = 4 * s->mb_x + 4 * s->mb_y * h->b_stride; // try mb2b(8)_xy
  742. const int b8_xy = 4 * h->mb_xy;
  743. if (USES_LIST(mb_type, 0)) {
  744. write_back_motion_list(h, s, b_stride, b_xy, b8_xy, mb_type, 0);
  745. } else {
  746. fill_rectangle(&s->current_picture.f.ref_index[0][b8_xy],
  747. 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
  748. }
  749. if (USES_LIST(mb_type, 1))
  750. write_back_motion_list(h, s, b_stride, b_xy, b8_xy, mb_type, 1);
  751. if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC) {
  752. if (IS_8X8(mb_type)) {
  753. uint8_t *direct_table = &h->direct_table[4 * h->mb_xy];
  754. direct_table[1] = h->sub_mb_type[1] >> 1;
  755. direct_table[2] = h->sub_mb_type[2] >> 1;
  756. direct_table[3] = h->sub_mb_type[3] >> 1;
  757. }
  758. }
  759. }
  760. static av_always_inline int get_dct8x8_allowed(H264Context *h)
  761. {
  762. if (h->sps.direct_8x8_inference_flag)
  763. return !(AV_RN64A(h->sub_mb_type) &
  764. ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) *
  765. 0x0001000100010001ULL));
  766. else
  767. return !(AV_RN64A(h->sub_mb_type) &
  768. ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8 | MB_TYPE_DIRECT2) *
  769. 0x0001000100010001ULL));
  770. }
  771. #endif /* AVCODEC_H264_H */