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.

886 lines
27KB

  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 CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
  38. #define COEFF_TOKEN_VLC_BITS 8
  39. #define TOTAL_ZEROS_VLC_BITS 9
  40. #define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
  41. #define RUN_VLC_BITS 3
  42. #define RUN7_VLC_BITS 6
  43. #define MAX_SPS_COUNT 32
  44. #define MAX_PPS_COUNT 256
  45. #define MAX_MMCO_COUNT 66
  46. #define MAX_DELAYED_PIC_COUNT 16
  47. /* Compiling in interlaced support reduces the speed
  48. * of progressive decoding by about 2%. */
  49. #define ALLOW_INTERLACE
  50. #define FMO 0
  51. /**
  52. * The maximum number of slices supported by the decoder.
  53. * must be a power of 2
  54. */
  55. #define MAX_SLICES 16
  56. #ifdef ALLOW_INTERLACE
  57. #define MB_MBAFF h->mb_mbaff
  58. #define MB_FIELD h->mb_field_decoding_flag
  59. #define FRAME_MBAFF h->mb_aff_frame
  60. #define FIELD_PICTURE (s->picture_structure != PICT_FRAME)
  61. #define LEFT_MBS 2
  62. #define LTOP 0
  63. #define LBOT 1
  64. #define LEFT(i) (i)
  65. #else
  66. #define MB_MBAFF 0
  67. #define MB_FIELD 0
  68. #define FRAME_MBAFF 0
  69. #define FIELD_PICTURE 0
  70. #undef IS_INTERLACED
  71. #define IS_INTERLACED(mb_type) 0
  72. #define LEFT_MBS 1
  73. #define LTOP 0
  74. #define LBOT 0
  75. #define LEFT(i) 0
  76. #endif
  77. #define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
  78. #ifndef CABAC
  79. #define CABAC h->pps.cabac
  80. #endif
  81. #define CHROMA444 (h->sps.chroma_format_idc == 3)
  82. #define EXTENDED_SAR 255
  83. #define MB_TYPE_REF0 MB_TYPE_ACPRED //dirty but it fits in 16 bit
  84. #define MB_TYPE_8x8DCT 0x01000000
  85. #define IS_REF0(a) ((a) & MB_TYPE_REF0)
  86. #define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
  87. /**
  88. * Value of Picture.reference when Picture is not a reference picture, but
  89. * is held for delayed output.
  90. */
  91. #define DELAYED_PIC_REF 4
  92. #define QP_MAX_NUM (51 + 2*6) // The maximum supported qp
  93. /* NAL unit types */
  94. enum {
  95. NAL_SLICE=1,
  96. NAL_DPA,
  97. NAL_DPB,
  98. NAL_DPC,
  99. NAL_IDR_SLICE,
  100. NAL_SEI,
  101. NAL_SPS,
  102. NAL_PPS,
  103. NAL_AUD,
  104. NAL_END_SEQUENCE,
  105. NAL_END_STREAM,
  106. NAL_FILLER_DATA,
  107. NAL_SPS_EXT,
  108. NAL_AUXILIARY_SLICE=19
  109. };
  110. /**
  111. * SEI message types
  112. */
  113. typedef enum {
  114. SEI_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
  115. SEI_TYPE_PIC_TIMING = 1, ///< picture timing
  116. SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
  117. SEI_TYPE_RECOVERY_POINT = 6 ///< recovery point (frame # to decoder sync)
  118. } SEI_Type;
  119. /**
  120. * pic_struct in picture timing SEI message
  121. */
  122. typedef enum {
  123. SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
  124. SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
  125. SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
  126. SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
  127. SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
  128. SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
  129. SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
  130. SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
  131. SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
  132. } SEI_PicStructType;
  133. /**
  134. * Sequence parameter set
  135. */
  136. typedef struct SPS{
  137. int profile_idc;
  138. int level_idc;
  139. int chroma_format_idc;
  140. int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
  141. int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
  142. int poc_type; ///< pic_order_cnt_type
  143. int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
  144. int delta_pic_order_always_zero_flag;
  145. int offset_for_non_ref_pic;
  146. int offset_for_top_to_bottom_field;
  147. int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
  148. int ref_frame_count; ///< num_ref_frames
  149. int gaps_in_frame_num_allowed_flag;
  150. int mb_width; ///< pic_width_in_mbs_minus1 + 1
  151. int mb_height; ///< pic_height_in_map_units_minus1 + 1
  152. int frame_mbs_only_flag;
  153. int mb_aff; ///<mb_adaptive_frame_field_flag
  154. int direct_8x8_inference_flag;
  155. int crop; ///< frame_cropping_flag
  156. unsigned int crop_left; ///< frame_cropping_rect_left_offset
  157. unsigned int crop_right; ///< frame_cropping_rect_right_offset
  158. unsigned int crop_top; ///< frame_cropping_rect_top_offset
  159. unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
  160. int vui_parameters_present_flag;
  161. AVRational sar;
  162. int video_signal_type_present_flag;
  163. int full_range;
  164. int colour_description_present_flag;
  165. enum AVColorPrimaries color_primaries;
  166. enum AVColorTransferCharacteristic color_trc;
  167. enum AVColorSpace colorspace;
  168. int timing_info_present_flag;
  169. uint32_t num_units_in_tick;
  170. uint32_t time_scale;
  171. int fixed_frame_rate_flag;
  172. short offset_for_ref_frame[256]; //FIXME dyn aloc?
  173. int bitstream_restriction_flag;
  174. int num_reorder_frames;
  175. int scaling_matrix_present;
  176. uint8_t scaling_matrix4[6][16];
  177. uint8_t scaling_matrix8[6][64];
  178. int nal_hrd_parameters_present_flag;
  179. int vcl_hrd_parameters_present_flag;
  180. int pic_struct_present_flag;
  181. int time_offset_length;
  182. int cpb_cnt; ///< See H.264 E.1.2
  183. int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 +1
  184. int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
  185. int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
  186. int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
  187. int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
  188. int residual_color_transform_flag; ///< residual_colour_transform_flag
  189. int constraint_set_flags; ///< constraint_set[0-3]_flag
  190. }SPS;
  191. /**
  192. * Picture parameter set
  193. */
  194. typedef struct PPS{
  195. unsigned int sps_id;
  196. int cabac; ///< entropy_coding_mode_flag
  197. int pic_order_present; ///< pic_order_present_flag
  198. int slice_group_count; ///< num_slice_groups_minus1 + 1
  199. int mb_slice_group_map_type;
  200. unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
  201. int weighted_pred; ///< weighted_pred_flag
  202. int weighted_bipred_idc;
  203. int init_qp; ///< pic_init_qp_minus26 + 26
  204. int init_qs; ///< pic_init_qs_minus26 + 26
  205. int chroma_qp_index_offset[2];
  206. int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
  207. int constrained_intra_pred; ///< constrained_intra_pred_flag
  208. int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
  209. int transform_8x8_mode; ///< transform_8x8_mode_flag
  210. uint8_t scaling_matrix4[6][16];
  211. uint8_t scaling_matrix8[6][64];
  212. uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
  213. int chroma_qp_diff;
  214. }PPS;
  215. /**
  216. * Memory management control operation opcode.
  217. */
  218. typedef enum MMCOOpcode{
  219. MMCO_END=0,
  220. MMCO_SHORT2UNUSED,
  221. MMCO_LONG2UNUSED,
  222. MMCO_SHORT2LONG,
  223. MMCO_SET_MAX_LONG,
  224. MMCO_RESET,
  225. MMCO_LONG,
  226. } MMCOOpcode;
  227. /**
  228. * Memory management control operation.
  229. */
  230. typedef struct MMCO{
  231. MMCOOpcode opcode;
  232. int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
  233. int long_arg; ///< index, pic_num, or num long refs depending on opcode
  234. } MMCO;
  235. /**
  236. * H264Context
  237. */
  238. typedef struct H264Context{
  239. MpegEncContext s;
  240. H264DSPContext h264dsp;
  241. int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
  242. int chroma_qp[2]; //QPc
  243. int qp_thresh; ///< QP threshold to skip loopfilter
  244. int prev_mb_skipped;
  245. int next_mb_skipped;
  246. //prediction stuff
  247. int chroma_pred_mode;
  248. int intra16x16_pred_mode;
  249. int topleft_mb_xy;
  250. int top_mb_xy;
  251. int topright_mb_xy;
  252. int left_mb_xy[LEFT_MBS];
  253. int topleft_type;
  254. int top_type;
  255. int topright_type;
  256. int left_type[LEFT_MBS];
  257. const uint8_t * left_block;
  258. int topleft_partition;
  259. int8_t intra4x4_pred_mode_cache[5*8];
  260. int8_t (*intra4x4_pred_mode);
  261. H264PredContext hpc;
  262. unsigned int topleft_samples_available;
  263. unsigned int top_samples_available;
  264. unsigned int topright_samples_available;
  265. unsigned int left_samples_available;
  266. uint8_t (*top_borders[2])[(16*3)*2];
  267. /**
  268. * non zero coeff count cache.
  269. * is 64 if not available.
  270. */
  271. DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15*8];
  272. uint8_t (*non_zero_count)[48];
  273. /**
  274. * Motion vector cache.
  275. */
  276. DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5*8][2];
  277. DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5*8];
  278. #define LIST_NOT_USED -1 //FIXME rename?
  279. #define PART_NOT_AVAILABLE -2
  280. /**
  281. * number of neighbors (top and/or left) that used 8x8 dct
  282. */
  283. int neighbor_transform_size;
  284. /**
  285. * block_offset[ 0..23] for frame macroblocks
  286. * block_offset[24..47] for field macroblocks
  287. */
  288. int block_offset[2*(16*3)];
  289. uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
  290. uint32_t *mb2br_xy;
  291. int b_stride; //FIXME use s->b4_stride
  292. int mb_linesize; ///< may be equal to s->linesize or s->linesize*2, for mbaff
  293. int mb_uvlinesize;
  294. int emu_edge_width;
  295. int emu_edge_height;
  296. SPS sps; ///< current sps
  297. /**
  298. * current pps
  299. */
  300. PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
  301. uint32_t dequant4_buffer[6][QP_MAX_NUM+1][16]; //FIXME should these be moved down?
  302. uint32_t dequant8_buffer[6][QP_MAX_NUM+1][64];
  303. uint32_t (*dequant4_coeff[6])[16];
  304. uint32_t (*dequant8_coeff[6])[64];
  305. int slice_num;
  306. uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
  307. int slice_type;
  308. int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
  309. int slice_type_fixed;
  310. //interlacing specific flags
  311. int mb_aff_frame;
  312. int mb_field_decoding_flag;
  313. int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
  314. DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
  315. //Weighted pred stuff
  316. int use_weight;
  317. int use_weight_chroma;
  318. int luma_log2_weight_denom;
  319. int chroma_log2_weight_denom;
  320. //The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss
  321. int luma_weight[48][2][2];
  322. int chroma_weight[48][2][2][2];
  323. int implicit_weight[48][48][2];
  324. int direct_spatial_mv_pred;
  325. int col_parity;
  326. int col_fieldoff;
  327. int dist_scale_factor[16];
  328. int dist_scale_factor_field[2][32];
  329. int map_col_to_list0[2][16+32];
  330. int map_col_to_list0_field[2][2][16+32];
  331. /**
  332. * num_ref_idx_l0/1_active_minus1 + 1
  333. */
  334. unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
  335. unsigned int list_count;
  336. uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
  337. Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
  338. Reordered version of default_ref_list
  339. according to picture reordering in slice header */
  340. int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
  341. //data partitioning
  342. GetBitContext intra_gb;
  343. GetBitContext inter_gb;
  344. GetBitContext *intra_gb_ptr;
  345. GetBitContext *inter_gb_ptr;
  346. 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.
  347. DECLARE_ALIGNED(16, DCTELEM, mb_luma_dc)[3][16*2];
  348. 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
  349. /**
  350. * Cabac
  351. */
  352. CABACContext cabac;
  353. uint8_t cabac_state[1024];
  354. /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
  355. uint16_t *cbp_table;
  356. int cbp;
  357. int top_cbp;
  358. int left_cbp;
  359. /* chroma_pred_mode for i4x4 or i16x16, else 0 */
  360. uint8_t *chroma_pred_mode_table;
  361. int last_qscale_diff;
  362. uint8_t (*mvd_table[2])[2];
  363. DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5*8][2];
  364. uint8_t *direct_table;
  365. uint8_t direct_cache[5*8];
  366. uint8_t zigzag_scan[16];
  367. uint8_t zigzag_scan8x8[64];
  368. uint8_t zigzag_scan8x8_cavlc[64];
  369. uint8_t field_scan[16];
  370. uint8_t field_scan8x8[64];
  371. uint8_t field_scan8x8_cavlc[64];
  372. const uint8_t *zigzag_scan_q0;
  373. const uint8_t *zigzag_scan8x8_q0;
  374. const uint8_t *zigzag_scan8x8_cavlc_q0;
  375. const uint8_t *field_scan_q0;
  376. const uint8_t *field_scan8x8_q0;
  377. const uint8_t *field_scan8x8_cavlc_q0;
  378. int x264_build;
  379. int mb_xy;
  380. int is_complex;
  381. //deblock
  382. int deblocking_filter; ///< disable_deblocking_filter_idc with 1<->0
  383. int slice_alpha_c0_offset;
  384. int slice_beta_offset;
  385. //=============================================================
  386. //Things below are not used in the MB or more inner code
  387. int nal_ref_idc;
  388. int nal_unit_type;
  389. uint8_t *rbsp_buffer[2];
  390. unsigned int rbsp_buffer_size[2];
  391. /**
  392. * Used to parse AVC variant of h264
  393. */
  394. int is_avc; ///< this flag is != 0 if codec is avc1
  395. int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
  396. int got_first; ///< this flag is != 0 if we've parsed a frame
  397. SPS *sps_buffers[MAX_SPS_COUNT];
  398. PPS *pps_buffers[MAX_PPS_COUNT];
  399. int dequant_coeff_pps; ///< reinit tables when pps changes
  400. uint16_t *slice_table_base;
  401. //POC stuff
  402. int poc_lsb;
  403. int poc_msb;
  404. int delta_poc_bottom;
  405. int delta_poc[2];
  406. int frame_num;
  407. int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
  408. int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
  409. int frame_num_offset; ///< for POC type 2
  410. int prev_frame_num_offset; ///< for POC type 2
  411. int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
  412. /**
  413. * frame_num for frames or 2*frame_num+1 for field pics.
  414. */
  415. int curr_pic_num;
  416. /**
  417. * max_frame_num or 2*max_frame_num for field pics.
  418. */
  419. int max_pic_num;
  420. int redundant_pic_count;
  421. Picture *short_ref[32];
  422. Picture *long_ref[32];
  423. Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
  424. Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
  425. Picture *next_output_pic;
  426. int outputed_poc;
  427. int next_outputed_poc;
  428. /**
  429. * memory management control operations buffer.
  430. */
  431. MMCO mmco[MAX_MMCO_COUNT];
  432. int mmco_index;
  433. int long_ref_count; ///< number of actual long term references
  434. int short_ref_count; ///< number of actual short term references
  435. int cabac_init_idc;
  436. /**
  437. * @name Members for slice based multithreading
  438. * @{
  439. */
  440. struct H264Context *thread_context[MAX_THREADS];
  441. /**
  442. * current slice number, used to initalize slice_num of each thread/context
  443. */
  444. int current_slice;
  445. /**
  446. * Max number of threads / contexts.
  447. * This is equal to AVCodecContext.thread_count unless
  448. * multithreaded decoding is impossible, in which case it is
  449. * reduced to 1.
  450. */
  451. int max_contexts;
  452. /**
  453. * 1 if the single thread fallback warning has already been
  454. * displayed, 0 otherwise.
  455. */
  456. int single_decode_warning;
  457. int last_slice_type;
  458. /** @} */
  459. /**
  460. * pic_struct in picture timing SEI message
  461. */
  462. SEI_PicStructType sei_pic_struct;
  463. /**
  464. * Complement sei_pic_struct
  465. * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
  466. * However, soft telecined frames may have these values.
  467. * This is used in an attempt to flag soft telecine progressive.
  468. */
  469. int prev_interlaced_frame;
  470. /**
  471. * Bit set of clock types for fields/frames in picture timing SEI message.
  472. * For each found ct_type, appropriate bit is set (e.g., bit 1 for
  473. * interlaced).
  474. */
  475. int sei_ct_type;
  476. /**
  477. * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
  478. */
  479. int sei_dpb_output_delay;
  480. /**
  481. * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
  482. */
  483. int sei_cpb_removal_delay;
  484. /**
  485. * recovery_frame_cnt from SEI message
  486. *
  487. * Set to -1 if no recovery point SEI message found or to number of frames
  488. * before playback synchronizes. Frames having recovery point are key
  489. * frames.
  490. */
  491. int sei_recovery_frame_cnt;
  492. int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
  493. int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
  494. // Timestamp stuff
  495. int sei_buffering_period_present; ///< Buffering period SEI flag
  496. int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
  497. }H264Context;
  498. 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).
  499. /**
  500. * Decode SEI
  501. */
  502. int ff_h264_decode_sei(H264Context *h);
  503. /**
  504. * Decode SPS
  505. */
  506. int ff_h264_decode_seq_parameter_set(H264Context *h);
  507. /**
  508. * compute profile from sps
  509. */
  510. int ff_h264_get_profile(SPS *sps);
  511. /**
  512. * Decode PPS
  513. */
  514. int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
  515. /**
  516. * Decode a network abstraction layer unit.
  517. * @param consumed is the number of bytes used as input
  518. * @param length is the length of the array
  519. * @param dst_length is the number of decoded bytes FIXME here or a decode rbsp tailing?
  520. * @return decoded bytes, might be src+1 if no escapes
  521. */
  522. const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length);
  523. /**
  524. * Free any data that may have been allocated in the H264 context 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 dc mode so it only uses the available blocks.
  551. */
  552. int ff_h264_check_intra4x4_pred_mode(H264Context *h);
  553. /**
  554. * Check if the top & left blocks are available if needed & change the dc mode so it only uses the available blocks.
  555. */
  556. int ff_h264_check_intra_pred_mode(H264Context *h, int mode);
  557. void ff_h264_hl_decode_mb(H264Context *h);
  558. int ff_h264_frame_start(H264Context *h);
  559. int ff_h264_decode_extradata(H264Context *h);
  560. av_cold int ff_h264_decode_init(AVCodecContext *avctx);
  561. av_cold int ff_h264_decode_end(AVCodecContext *avctx);
  562. av_cold void ff_h264_decode_init_vlc(void);
  563. /**
  564. * Decode a macroblock
  565. * @return 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
  566. */
  567. int ff_h264_decode_mb_cavlc(H264Context *h);
  568. /**
  569. * Decode a CABAC coded macroblock
  570. * @return 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
  571. */
  572. int ff_h264_decode_mb_cabac(H264Context *h);
  573. void ff_h264_init_cabac_states(H264Context *h);
  574. void ff_h264_direct_dist_scale_factor(H264Context * const h);
  575. void ff_h264_direct_ref_list_init(H264Context * const h);
  576. void ff_h264_pred_direct_motion(H264Context * const h, int *mb_type);
  577. void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);
  578. void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);
  579. /**
  580. * Reset SEI values at the beginning of the frame.
  581. *
  582. * @param h H.264 context.
  583. */
  584. void ff_h264_reset_sei(H264Context *h);
  585. /*
  586. o-o o-o
  587. / / /
  588. o-o o-o
  589. ,---'
  590. o-o o-o
  591. / / /
  592. o-o o-o
  593. */
  594. /* Scan8 organization:
  595. * 0 1 2 3 4 5 6 7
  596. * 0 DY y y y y y
  597. * 1 y Y Y Y Y
  598. * 2 y Y Y Y Y
  599. * 3 y Y Y Y Y
  600. * 4 y Y Y Y Y
  601. * 5 DU u u u u u
  602. * 6 u U U U U
  603. * 7 u U U U U
  604. * 8 u U U U U
  605. * 9 u U U U U
  606. * 10 DV v v v v v
  607. * 11 v V V V V
  608. * 12 v V V V V
  609. * 13 v V V V V
  610. * 14 v V V V V
  611. * DY/DU/DV are for luma/chroma DC.
  612. */
  613. #define LUMA_DC_BLOCK_INDEX 48
  614. #define CHROMA_DC_BLOCK_INDEX 49
  615. //This table must be here because scan8[constant] must be known at compiletime
  616. static const uint8_t scan8[16*3 + 3]={
  617. 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8,
  618. 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8,
  619. 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8,
  620. 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8,
  621. 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8,
  622. 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8,
  623. 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8,
  624. 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8,
  625. 4+11*8, 5+11*8, 4+12*8, 5+12*8,
  626. 6+11*8, 7+11*8, 6+12*8, 7+12*8,
  627. 4+13*8, 5+13*8, 4+14*8, 5+14*8,
  628. 6+13*8, 7+13*8, 6+14*8, 7+14*8,
  629. 0+ 0*8, 0+ 5*8, 0+10*8
  630. };
  631. static av_always_inline uint32_t pack16to32(int a, int b){
  632. #if HAVE_BIGENDIAN
  633. return (b&0xFFFF) + (a<<16);
  634. #else
  635. return (a&0xFFFF) + (b<<16);
  636. #endif
  637. }
  638. static av_always_inline uint16_t pack8to16(int a, int b){
  639. #if HAVE_BIGENDIAN
  640. return (b&0xFF) + (a<<8);
  641. #else
  642. return (a&0xFF) + (b<<8);
  643. #endif
  644. }
  645. /**
  646. * gets the chroma qp.
  647. */
  648. static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale){
  649. return h->pps.chroma_qp_table[t][qscale];
  650. }
  651. /**
  652. * gets the predicted intra4x4 prediction mode.
  653. */
  654. static av_always_inline int pred_intra_mode(H264Context *h, int n){
  655. const int index8= scan8[n];
  656. const int left= h->intra4x4_pred_mode_cache[index8 - 1];
  657. const int top = h->intra4x4_pred_mode_cache[index8 - 8];
  658. const int min= FFMIN(left, top);
  659. tprintf(h->s.avctx, "mode:%d %d min:%d\n", left ,top, min);
  660. if(min<0) return DC_PRED;
  661. else return min;
  662. }
  663. static av_always_inline void write_back_intra_pred_mode(H264Context *h){
  664. int8_t *i4x4= h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
  665. int8_t *i4x4_cache= h->intra4x4_pred_mode_cache;
  666. AV_COPY32(i4x4, i4x4_cache + 4 + 8*4);
  667. i4x4[4]= i4x4_cache[7+8*3];
  668. i4x4[5]= i4x4_cache[7+8*2];
  669. i4x4[6]= i4x4_cache[7+8*1];
  670. }
  671. static av_always_inline void write_back_non_zero_count(H264Context *h){
  672. const int mb_xy= h->mb_xy;
  673. uint8_t *nnz = h->non_zero_count[mb_xy];
  674. uint8_t *nnz_cache = h->non_zero_count_cache;
  675. AV_COPY32(&nnz[ 0], &nnz_cache[4+8* 1]);
  676. AV_COPY32(&nnz[ 4], &nnz_cache[4+8* 2]);
  677. AV_COPY32(&nnz[ 8], &nnz_cache[4+8* 3]);
  678. AV_COPY32(&nnz[12], &nnz_cache[4+8* 4]);
  679. AV_COPY32(&nnz[16], &nnz_cache[4+8* 6]);
  680. AV_COPY32(&nnz[20], &nnz_cache[4+8* 7]);
  681. AV_COPY32(&nnz[32], &nnz_cache[4+8*11]);
  682. AV_COPY32(&nnz[36], &nnz_cache[4+8*12]);
  683. if(CHROMA444){
  684. AV_COPY32(&nnz[24], &nnz_cache[4+8* 8]);
  685. AV_COPY32(&nnz[28], &nnz_cache[4+8* 9]);
  686. AV_COPY32(&nnz[40], &nnz_cache[4+8*13]);
  687. AV_COPY32(&nnz[44], &nnz_cache[4+8*14]);
  688. }
  689. }
  690. static av_always_inline void write_back_motion_list(H264Context *h, MpegEncContext * const s, int b_stride,
  691. int b_xy, int b8_xy, int mb_type, int list )
  692. {
  693. int16_t (*mv_dst)[2] = &s->current_picture.f.motion_val[list][b_xy];
  694. int16_t (*mv_src)[2] = &h->mv_cache[list][scan8[0]];
  695. AV_COPY128(mv_dst + 0*b_stride, mv_src + 8*0);
  696. AV_COPY128(mv_dst + 1*b_stride, mv_src + 8*1);
  697. AV_COPY128(mv_dst + 2*b_stride, mv_src + 8*2);
  698. AV_COPY128(mv_dst + 3*b_stride, mv_src + 8*3);
  699. if( CABAC ) {
  700. uint8_t (*mvd_dst)[2] = &h->mvd_table[list][FMO ? 8*h->mb_xy : h->mb2br_xy[h->mb_xy]];
  701. uint8_t (*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
  702. if(IS_SKIP(mb_type))
  703. AV_ZERO128(mvd_dst);
  704. else{
  705. AV_COPY64(mvd_dst, mvd_src + 8*3);
  706. AV_COPY16(mvd_dst + 3 + 3, mvd_src + 3 + 8*0);
  707. AV_COPY16(mvd_dst + 3 + 2, mvd_src + 3 + 8*1);
  708. AV_COPY16(mvd_dst + 3 + 1, mvd_src + 3 + 8*2);
  709. }
  710. }
  711. {
  712. int8_t *ref_index = &s->current_picture.f.ref_index[list][b8_xy];
  713. int8_t *ref_cache = h->ref_cache[list];
  714. ref_index[0+0*2]= ref_cache[scan8[0]];
  715. ref_index[1+0*2]= ref_cache[scan8[4]];
  716. ref_index[0+1*2]= ref_cache[scan8[8]];
  717. ref_index[1+1*2]= ref_cache[scan8[12]];
  718. }
  719. }
  720. static av_always_inline void write_back_motion(H264Context *h, int mb_type){
  721. MpegEncContext * const s = &h->s;
  722. const int b_stride = h->b_stride;
  723. const int b_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride; //try mb2b(8)_xy
  724. const int b8_xy= 4*h->mb_xy;
  725. if(USES_LIST(mb_type, 0)){
  726. write_back_motion_list(h, s, b_stride, b_xy, b8_xy, mb_type, 0);
  727. }else{
  728. fill_rectangle(&s->current_picture.f.ref_index[0][b8_xy],
  729. 2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
  730. }
  731. if(USES_LIST(mb_type, 1)){
  732. write_back_motion_list(h, s, b_stride, b_xy, b8_xy, mb_type, 1);
  733. }
  734. if(h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC){
  735. if(IS_8X8(mb_type)){
  736. uint8_t *direct_table = &h->direct_table[4*h->mb_xy];
  737. direct_table[1] = h->sub_mb_type[1]>>1;
  738. direct_table[2] = h->sub_mb_type[2]>>1;
  739. direct_table[3] = h->sub_mb_type[3]>>1;
  740. }
  741. }
  742. }
  743. static av_always_inline int get_dct8x8_allowed(H264Context *h){
  744. if(h->sps.direct_8x8_inference_flag)
  745. return !(AV_RN64A(h->sub_mb_type) & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8 )*0x0001000100010001ULL));
  746. else
  747. return !(AV_RN64A(h->sub_mb_type) & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8|MB_TYPE_DIRECT2)*0x0001000100010001ULL));
  748. }
  749. #endif /* AVCODEC_H264_H */