2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * H.264 / AVC / MPEG4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
33 #include "mpegvideo.h"
35 #define interlaced_dct interlaced_dct_is_a_bad_name
36 #define mb_intra mb_intra_is_not_initialized_see_mb_type
38 #define LUMA_DC_BLOCK_INDEX 25
39 #define CHROMA_DC_BLOCK_INDEX 26
41 #define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
42 #define COEFF_TOKEN_VLC_BITS 8
43 #define TOTAL_ZEROS_VLC_BITS 9
44 #define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
45 #define RUN_VLC_BITS 3
46 #define RUN7_VLC_BITS 6
48 #define MAX_SPS_COUNT 32
49 #define MAX_PPS_COUNT 256
51 #define MAX_MMCO_COUNT 66
53 /* Compiling in interlaced support reduces the speed
54 * of progressive decoding by about 2%. */
55 #define ALLOW_INTERLACE
57 #ifdef ALLOW_INTERLACE
58 #define MB_MBAFF h->mb_mbaff
59 #define MB_FIELD h->mb_field_decoding_flag
60 #define FRAME_MBAFF h->mb_aff_frame
66 #define IS_INTERLACED(mb_type) 0
70 * Sequence parameter set
76 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
77 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
78 int poc_type; ///< pic_order_cnt_type
79 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
80 int delta_pic_order_always_zero_flag;
81 int offset_for_non_ref_pic;
82 int offset_for_top_to_bottom_field;
83 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
84 int ref_frame_count; ///< num_ref_frames
85 int gaps_in_frame_num_allowed_flag;
86 int mb_width; ///< frame_width_in_mbs_minus1 + 1
87 int mb_height; ///< frame_height_in_mbs_minus1 + 1
88 int frame_mbs_only_flag;
89 int mb_aff; ///<mb_adaptive_frame_field_flag
90 int direct_8x8_inference_flag;
91 int crop; ///< frame_cropping_flag
92 int crop_left; ///< frame_cropping_rect_left_offset
93 int crop_right; ///< frame_cropping_rect_right_offset
94 int crop_top; ///< frame_cropping_rect_top_offset
95 int crop_bottom; ///< frame_cropping_rect_bottom_offset
96 int vui_parameters_present_flag;
98 int timing_info_present_flag;
99 uint32_t num_units_in_tick;
101 int fixed_frame_rate_flag;
102 short offset_for_ref_frame[256]; //FIXME dyn aloc?
103 int bitstream_restriction_flag;
104 int num_reorder_frames;
105 int scaling_matrix_present;
106 uint8_t scaling_matrix4[6][16];
107 uint8_t scaling_matrix8[2][64];
111 * Picture parameter set
115 int cabac; ///< entropy_coding_mode_flag
116 int pic_order_present; ///< pic_order_present_flag
117 int slice_group_count; ///< num_slice_groups_minus1 + 1
118 int mb_slice_group_map_type;
119 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
120 int weighted_pred; ///< weighted_pred_flag
121 int weighted_bipred_idc;
122 int init_qp; ///< pic_init_qp_minus26 + 26
123 int init_qs; ///< pic_init_qs_minus26 + 26
124 int chroma_qp_index_offset[2];
125 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
126 int constrained_intra_pred; ///< constrained_intra_pred_flag
127 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
128 int transform_8x8_mode; ///< transform_8x8_mode_flag
129 uint8_t scaling_matrix4[6][16];
130 uint8_t scaling_matrix8[2][64];
131 uint8_t chroma_qp_table[2][256]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
136 * Memory management control operation opcode.
138 typedef enum MMCOOpcode{
149 * Memory management control operation.
160 typedef struct H264Context{
164 uint8_t *rbsp_buffer[2];
165 unsigned int rbsp_buffer_size[2];
168 * Used to parse AVC variant of h264
170 int is_avc; ///< this flag is != 0 if codec is avc1
171 int got_avcC; ///< flag used to parse avcC data only once
172 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
174 int chroma_qp[2]; //QPc
180 int chroma_pred_mode;
181 int intra16x16_pred_mode;
186 int8_t intra4x4_pred_mode_cache[5*8];
187 int8_t (*intra4x4_pred_mode)[8];
188 void (*pred4x4 [9+3])(uint8_t *src, uint8_t *topright, int stride);//FIXME move to dsp?
189 void (*pred8x8l [9+3])(uint8_t *src, int topleft, int topright, int stride);
190 void (*pred8x8 [4+3])(uint8_t *src, int stride);
191 void (*pred16x16[4+3])(uint8_t *src, int stride);
192 unsigned int topleft_samples_available;
193 unsigned int top_samples_available;
194 unsigned int topright_samples_available;
195 unsigned int left_samples_available;
196 uint8_t (*top_borders[2])[16+2*8];
197 uint8_t left_border[2*(17+2*9)];
200 * non zero coeff count cache.
201 * is 64 if not available.
203 DECLARE_ALIGNED_8(uint8_t, non_zero_count_cache[6*8]);
204 uint8_t (*non_zero_count)[16];
207 * Motion vector cache.
209 DECLARE_ALIGNED_8(int16_t, mv_cache[2][5*8][2]);
210 DECLARE_ALIGNED_8(int8_t, ref_cache[2][5*8]);
211 #define LIST_NOT_USED -1 //FIXME rename?
212 #define PART_NOT_AVAILABLE -2
215 * is 1 if the specific list MV&references are set to 0,0,-2.
217 int mv_cache_clean[2];
220 * number of neighbors (top and/or left) that used 8x8 dct
222 int neighbor_transform_size;
225 * block_offset[ 0..23] for frame macroblocks
226 * block_offset[24..47] for field macroblocks
228 int block_offset[2*(16+8)];
230 uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
232 int b_stride; //FIXME use s->b4_stride
235 int mb_linesize; ///< may be equal to s->linesize or s->linesize*2, for mbaff
244 int unknown_svq3_flag;
245 int next_slice_index;
247 SPS *sps_buffers[MAX_SPS_COUNT];
248 SPS sps; ///< current sps
250 PPS *pps_buffers[MAX_PPS_COUNT];
254 PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
256 uint32_t dequant4_buffer[6][52][16];
257 uint32_t dequant8_buffer[2][52][64];
258 uint32_t (*dequant4_coeff[6])[16];
259 uint32_t (*dequant8_coeff[2])[64];
260 int dequant_coeff_pps; ///< reinit tables when pps changes
263 uint8_t *slice_table_base;
264 uint8_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
266 int slice_type_fixed;
268 //interlacing specific flags
270 int mb_field_decoding_flag;
271 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
273 unsigned int sub_mb_type[4];
278 int delta_poc_bottom;
281 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
282 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
283 int frame_num_offset; ///< for POC type 2
284 int prev_frame_num_offset; ///< for POC type 2
285 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
288 * frame_num for frames or 2*frame_num for field pics.
293 * max_frame_num or 2*max_frame_num for field pics.
297 //Weighted pred stuff
299 int use_weight_chroma;
300 int luma_log2_weight_denom;
301 int chroma_log2_weight_denom;
302 int luma_weight[2][48];
303 int luma_offset[2][48];
304 int chroma_weight[2][48][2];
305 int chroma_offset[2][48][2];
306 int implicit_weight[48][48];
309 int deblocking_filter; ///< disable_deblocking_filter_idc with 1<->0
310 int slice_alpha_c0_offset;
311 int slice_beta_offset;
313 int redundant_pic_count;
315 int direct_spatial_mv_pred;
316 int dist_scale_factor[16];
317 int dist_scale_factor_field[32];
318 int map_col_to_list0[2][16];
319 int map_col_to_list0_field[2][32];
322 * num_ref_idx_l0/1_active_minus1 + 1
324 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
325 unsigned int list_count;
326 Picture *short_ref[32];
327 Picture *long_ref[32];
328 Picture default_ref_list[2][32];
329 Picture ref_list[2][48]; ///< 0..15: frame refs, 16..47: mbaff field refs
330 Picture *delayed_pic[18]; //FIXME size?
331 Picture *delayed_output_pic;
334 * memory management control operations buffer.
336 MMCO mmco[MAX_MMCO_COUNT];
339 int long_ref_count; ///< number of actual long term references
340 int short_ref_count; ///< number of actual short term references
343 GetBitContext intra_gb;
344 GetBitContext inter_gb;
345 GetBitContext *intra_gb_ptr;
346 GetBitContext *inter_gb_ptr;
348 DECLARE_ALIGNED_8(DCTELEM, mb[16*24]);
349 DCTELEM mb_padding[256]; ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either check that i is not to large or ensure that there is some unused stuff after mb
355 uint8_t cabac_state[460];
358 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
363 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
364 uint8_t *chroma_pred_mode_table;
365 int last_qscale_diff;
366 int16_t (*mvd_table[2])[2];
367 DECLARE_ALIGNED_8(int16_t, mvd_cache[2][5*8][2]);
368 uint8_t *direct_table;
369 uint8_t direct_cache[5*8];
371 uint8_t zigzag_scan[16];
372 uint8_t zigzag_scan8x8[64];
373 uint8_t zigzag_scan8x8_cavlc[64];
374 uint8_t field_scan[16];
375 uint8_t field_scan8x8[64];
376 uint8_t field_scan8x8_cavlc[64];
377 const uint8_t *zigzag_scan_q0;
378 const uint8_t *zigzag_scan8x8_q0;
379 const uint8_t *zigzag_scan8x8_cavlc_q0;
380 const uint8_t *field_scan_q0;
381 const uint8_t *field_scan8x8_q0;
382 const uint8_t *field_scan8x8_cavlc_q0;