Remove unused variable. Seems i forgot to commit this.
[libav.git] / libavcodec / h264.h
CommitLineData
26b4fe82
AJ
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
26b4fe82
AJ
20 */
21
22/**
bad5537e 23 * @file libavcodec/h264.h
26b4fe82
AJ
24 * H.264 / AVC / MPEG4 part10 codec.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
98790382
SS
28#ifndef AVCODEC_H264_H
29#define AVCODEC_H264_H
26b4fe82 30
19769ece 31#include "libavutil/intreadwrite.h"
26b4fe82
AJ
32#include "dsputil.h"
33#include "cabac.h"
34#include "mpegvideo.h"
c92a30bb 35#include "h264pred.h"
e1e94902 36#include "rectangle.h"
26b4fe82
AJ
37
38#define interlaced_dct interlaced_dct_is_a_bad_name
90b5b51e 39#define mb_intra mb_intra_is_not_initialized_see_mb_type
26b4fe82
AJ
40
41#define LUMA_DC_BLOCK_INDEX 25
42#define CHROMA_DC_BLOCK_INDEX 26
43
44#define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
45#define COEFF_TOKEN_VLC_BITS 8
46#define TOTAL_ZEROS_VLC_BITS 9
47#define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
48#define RUN_VLC_BITS 3
49#define RUN7_VLC_BITS 6
50
51#define MAX_SPS_COUNT 32
52#define MAX_PPS_COUNT 256
53
54#define MAX_MMCO_COUNT 66
55
64b9d48f
AS
56#define MAX_DELAYED_PIC_COUNT 16
57
26b4fe82
AJ
58/* Compiling in interlaced support reduces the speed
59 * of progressive decoding by about 2%. */
60#define ALLOW_INTERLACE
61
0af6967e
MN
62#define ALLOW_NOCHROMA
63
b735aeea
MN
64/**
65 * The maximum number of slices supported by the decoder.
66 * must be a power of 2
67 */
68#define MAX_SLICES 16
69
26b4fe82
AJ
70#ifdef ALLOW_INTERLACE
71#define MB_MBAFF h->mb_mbaff
72#define MB_FIELD h->mb_field_decoding_flag
73#define FRAME_MBAFF h->mb_aff_frame
d6c52130 74#define FIELD_PICTURE (s->picture_structure != PICT_FRAME)
26b4fe82
AJ
75#else
76#define MB_MBAFF 0
77#define MB_FIELD 0
78#define FRAME_MBAFF 0
bbb3edb8 79#define FIELD_PICTURE 0
26b4fe82
AJ
80#undef IS_INTERLACED
81#define IS_INTERLACED(mb_type) 0
82#endif
f3e53d9f 83#define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
26b4fe82 84
0af6967e
MN
85#ifdef ALLOW_NOCHROMA
86#define CHROMA h->sps.chroma_format_idc
87#else
88#define CHROMA 1
89#endif
90
ddd60f28
MN
91#ifndef CABAC
92#define CABAC h->pps.cabac
93#endif
94
44ee6135
MN
95#define EXTENDED_SAR 255
96
ce9ff9cc
DB
97#define MB_TYPE_REF0 MB_TYPE_ACPRED //dirty but it fits in 16 bit
98#define MB_TYPE_8x8DCT 0x01000000
99#define IS_REF0(a) ((a) & MB_TYPE_REF0)
100#define IS_8x8DCT(a) ((a) & MB_TYPE_8x8DCT)
101
ea6f00c4
MN
102/**
103 * Value of Picture.reference when Picture is not a reference picture, but
104 * is held for delayed output.
105 */
106#define DELAYED_PIC_REF 4
107
108
44ee6135
MN
109/* NAL unit types */
110enum {
2791c608
MN
111 NAL_SLICE=1,
112 NAL_DPA,
113 NAL_DPB,
114 NAL_DPC,
115 NAL_IDR_SLICE,
116 NAL_SEI,
117 NAL_SPS,
118 NAL_PPS,
119 NAL_AUD,
120 NAL_END_SEQUENCE,
121 NAL_END_STREAM,
122 NAL_FILLER_DATA,
123 NAL_SPS_EXT,
124 NAL_AUXILIARY_SLICE=19
44ee6135
MN
125};
126
26b4fe82 127/**
dd02ea11
IS
128 * SEI message types
129 */
130typedef enum {
2ea89d92 131 SEI_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1)
dd02ea11
IS
132 SEI_TYPE_PIC_TIMING = 1, ///< picture timing
133 SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
134 SEI_TYPE_RECOVERY_POINT = 6 ///< recovery point (frame # to decoder sync)
135} SEI_Type;
136
137/**
b09a7c05
138 * pic_struct in picture timing SEI message
139 */
140typedef enum {
141 SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame
142 SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field
143 SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field
144 SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order
145 SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order
146 SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order
147 SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order
148 SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling
149 SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling
150} SEI_PicStructType;
151
152/**
26b4fe82
AJ
153 * Sequence parameter set
154 */
155typedef struct SPS{
156
157 int profile_idc;
158 int level_idc;
0af6967e 159 int chroma_format_idc;
26b4fe82
AJ
160 int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
161 int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
162 int poc_type; ///< pic_order_cnt_type
163 int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
164 int delta_pic_order_always_zero_flag;
165 int offset_for_non_ref_pic;
166 int offset_for_top_to_bottom_field;
167 int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
168 int ref_frame_count; ///< num_ref_frames
169 int gaps_in_frame_num_allowed_flag;
bcc3476c
JD
170 int mb_width; ///< pic_width_in_mbs_minus1 + 1
171 int mb_height; ///< pic_height_in_map_units_minus1 + 1
26b4fe82
AJ
172 int frame_mbs_only_flag;
173 int mb_aff; ///<mb_adaptive_frame_field_flag
174 int direct_8x8_inference_flag;
175 int crop; ///< frame_cropping_flag
59eeddf0
MN
176 unsigned int crop_left; ///< frame_cropping_rect_left_offset
177 unsigned int crop_right; ///< frame_cropping_rect_right_offset
178 unsigned int crop_top; ///< frame_cropping_rect_top_offset
179 unsigned int crop_bottom; ///< frame_cropping_rect_bottom_offset
26b4fe82
AJ
180 int vui_parameters_present_flag;
181 AVRational sar;
c4dffe7e
DC
182 int video_signal_type_present_flag;
183 int full_range;
184 int colour_description_present_flag;
185 enum AVColorPrimaries color_primaries;
186 enum AVColorTransferCharacteristic color_trc;
187 enum AVColorSpace colorspace;
26b4fe82
AJ
188 int timing_info_present_flag;
189 uint32_t num_units_in_tick;
190 uint32_t time_scale;
191 int fixed_frame_rate_flag;
192 short offset_for_ref_frame[256]; //FIXME dyn aloc?
193 int bitstream_restriction_flag;
194 int num_reorder_frames;
195 int scaling_matrix_present;
196 uint8_t scaling_matrix4[6][16];
197 uint8_t scaling_matrix8[2][64];
b09a7c05
198 int nal_hrd_parameters_present_flag;
199 int vcl_hrd_parameters_present_flag;
200 int pic_struct_present_flag;
201 int time_offset_length;
24a414e0 202 int cpb_cnt; ///< See H.264 E.1.2
82e38c29 203 int initial_cpb_removal_delay_length; ///< initial_cpb_removal_delay_length_minus1 +1
b09a7c05
204 int cpb_removal_delay_length; ///< cpb_removal_delay_length_minus1 + 1
205 int dpb_output_delay_length; ///< dpb_output_delay_length_minus1 + 1
cb99c652
GB
206 int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
207 int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8
208 int residual_color_transform_flag; ///< residual_colour_transform_flag
26b4fe82
AJ
209}SPS;
210
211/**
212 * Picture parameter set
213 */
214typedef struct PPS{
215 unsigned int sps_id;
216 int cabac; ///< entropy_coding_mode_flag
217 int pic_order_present; ///< pic_order_present_flag
218 int slice_group_count; ///< num_slice_groups_minus1 + 1
219 int mb_slice_group_map_type;
220 unsigned int ref_count[2]; ///< num_ref_idx_l0/1_active_minus1 + 1
221 int weighted_pred; ///< weighted_pred_flag
222 int weighted_bipred_idc;
223 int init_qp; ///< pic_init_qp_minus26 + 26
224 int init_qs; ///< pic_init_qs_minus26 + 26
4691a77d 225 int chroma_qp_index_offset[2];
26b4fe82
AJ
226 int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
227 int constrained_intra_pred; ///< constrained_intra_pred_flag
228 int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
229 int transform_8x8_mode; ///< transform_8x8_mode_flag
230 uint8_t scaling_matrix4[6][16];
231 uint8_t scaling_matrix8[2][64];
5a78bfbd 232 uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
4691a77d 233 int chroma_qp_diff;
26b4fe82
AJ
234}PPS;
235
236/**
237 * Memory management control operation opcode.
238 */
239typedef enum MMCOOpcode{
240 MMCO_END=0,
241 MMCO_SHORT2UNUSED,
242 MMCO_LONG2UNUSED,
243 MMCO_SHORT2LONG,
244 MMCO_SET_MAX_LONG,
245 MMCO_RESET,
246 MMCO_LONG,
247} MMCOOpcode;
248
249/**
250 * Memory management control operation.
251 */
252typedef struct MMCO{
253 MMCOOpcode opcode;
0d0447ea
JD
254 int short_pic_num; ///< pic_num without wrapping (pic_num & max_pic_num)
255 int long_arg; ///< index, pic_num, or num long refs depending on opcode
26b4fe82
AJ
256} MMCO;
257
258/**
259 * H264Context
260 */
261typedef struct H264Context{
262 MpegEncContext s;
263 int nal_ref_idc;
264 int nal_unit_type;
24456882
265 uint8_t *rbsp_buffer[2];
266 unsigned int rbsp_buffer_size[2];
26b4fe82
AJ
267
268 /**
269 * Used to parse AVC variant of h264
270 */
271 int is_avc; ///< this flag is != 0 if codec is avc1
26b4fe82
AJ
272 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
273
4691a77d 274 int chroma_qp[2]; //QPc
26b4fe82 275
2e36c931
MN
276 int qp_thresh; ///< QP threshold to skip loopfilter
277
26b4fe82
AJ
278 int prev_mb_skipped;
279 int next_mb_skipped;
280
281 //prediction stuff
282 int chroma_pred_mode;
283 int intra16x16_pred_mode;
284
c1bb66ac 285 int topleft_mb_xy;
26b4fe82 286 int top_mb_xy;
c1bb66ac 287 int topright_mb_xy;
26b4fe82
AJ
288 int left_mb_xy[2];
289
c1bb66ac 290 int topleft_type;
99344d43 291 int top_type;
c1bb66ac 292 int topright_type;
99344d43
MN
293 int left_type[2];
294
c1bb66ac
MN
295 const uint8_t * left_block;
296 int topleft_partition;
297
26b4fe82
AJ
298 int8_t intra4x4_pred_mode_cache[5*8];
299 int8_t (*intra4x4_pred_mode)[8];
c92a30bb 300 H264PredContext hpc;
26b4fe82
AJ
301 unsigned int topleft_samples_available;
302 unsigned int top_samples_available;
303 unsigned int topright_samples_available;
304 unsigned int left_samples_available;
305 uint8_t (*top_borders[2])[16+2*8];
306 uint8_t left_border[2*(17+2*9)];
307
308 /**
309 * non zero coeff count cache.
310 * is 64 if not available.
311 */
c6727809 312 DECLARE_ALIGNED_8(uint8_t, non_zero_count_cache)[6*8];
b6303e6d
MN
313
314 /*
315 .UU.YYYY
316 .UU.YYYY
317 .vv.YYYY
318 .VV.YYYY
319 */
c988f975 320 uint8_t (*non_zero_count)[32];
26b4fe82
AJ
321
322 /**
323 * Motion vector cache.
324 */
0b69d625 325 DECLARE_ALIGNED_16(int16_t, mv_cache)[2][5*8][2];
c6727809 326 DECLARE_ALIGNED_8(int8_t, ref_cache)[2][5*8];
26b4fe82
AJ
327#define LIST_NOT_USED -1 //FIXME rename?
328#define PART_NOT_AVAILABLE -2
329
330 /**
331 * is 1 if the specific list MV&references are set to 0,0,-2.
332 */
333 int mv_cache_clean[2];
334
335 /**
336 * number of neighbors (top and/or left) that used 8x8 dct
337 */
338 int neighbor_transform_size;
339
340 /**
341 * block_offset[ 0..23] for frame macroblocks
342 * block_offset[24..47] for field macroblocks
343 */
344 int block_offset[2*(16+8)];
345
346 uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
347 uint32_t *mb2b8_xy;
348 int b_stride; //FIXME use s->b4_stride
349 int b8_stride;
350
351 int mb_linesize; ///< may be equal to s->linesize or s->linesize*2, for mbaff
352 int mb_uvlinesize;
353
354 int emu_edge_width;
355 int emu_edge_height;
356
357 int halfpel_flag;
358 int thirdpel_flag;
359
360 int unknown_svq3_flag;
361 int next_slice_index;
362
7978debd 363 SPS *sps_buffers[MAX_SPS_COUNT];
26b4fe82
AJ
364 SPS sps; ///< current sps
365
7978debd 366 PPS *pps_buffers[MAX_PPS_COUNT];
26b4fe82
AJ
367 /**
368 * current pps
369 */
370 PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
371
372 uint32_t dequant4_buffer[6][52][16];
373 uint32_t dequant8_buffer[2][52][64];
374 uint32_t (*dequant4_coeff[6])[16];
375 uint32_t (*dequant8_coeff[2])[64];
376 int dequant_coeff_pps; ///< reinit tables when pps changes
377
378 int slice_num;
b735aeea
MN
379 uint16_t *slice_table_base;
380 uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
26b4fe82 381 int slice_type;
e3e6f18f 382 int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
26b4fe82
AJ
383 int slice_type_fixed;
384
385 //interlacing specific flags
386 int mb_aff_frame;
387 int mb_field_decoding_flag;
388 int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
389
c6727809 390 DECLARE_ALIGNED_8(uint16_t, sub_mb_type)[4];
26b4fe82
AJ
391
392 //POC stuff
393 int poc_lsb;
394 int poc_msb;
395 int delta_poc_bottom;
396 int delta_poc[2];
397 int frame_num;
398 int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0
399 int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0
400 int frame_num_offset; ///< for POC type 2
401 int prev_frame_num_offset; ///< for POC type 2
402 int prev_frame_num; ///< frame_num of the last pic for POC type 1/2
403
404 /**
f57e2af6 405 * frame_num for frames or 2*frame_num+1 for field pics.
26b4fe82
AJ
406 */
407 int curr_pic_num;
408
409 /**
410 * max_frame_num or 2*max_frame_num for field pics.
411 */
412 int max_pic_num;
413
414 //Weighted pred stuff
415 int use_weight;
416 int use_weight_chroma;
417 int luma_log2_weight_denom;
418 int chroma_log2_weight_denom;
419 int luma_weight[2][48];
420 int luma_offset[2][48];
421 int chroma_weight[2][48][2];
422 int chroma_offset[2][48][2];
423 int implicit_weight[48][48];
424
425 //deblock
426 int deblocking_filter; ///< disable_deblocking_filter_idc with 1<->0
427 int slice_alpha_c0_offset;
428 int slice_beta_offset;
429
430 int redundant_pic_count;
431
432 int direct_spatial_mv_pred;
3a06e864
MN
433 int col_parity;
434 int col_fieldoff;
26b4fe82 435 int dist_scale_factor[16];
8b1fd554 436 int dist_scale_factor_field[2][32];
f4d3382d
MN
437 int map_col_to_list0[2][16+32];
438 int map_col_to_list0_field[2][2][16+32];
26b4fe82
AJ
439
440 /**
441 * num_ref_idx_l0/1_active_minus1 + 1
442 */
443 unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
444 unsigned int list_count;
c988f975 445 uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
26b4fe82
AJ
446 Picture *short_ref[32];
447 Picture *long_ref[32];
bcc3476c
JD
448 Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
449 Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs.
450 Reordered version of default_ref_list
451 according to picture reordering in slice header */
b735aeea 452 int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1
64b9d48f 453 Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
df8a7dff 454 int outputed_poc;
26b4fe82
AJ
455
456 /**
457 * memory management control operations buffer.
458 */
459 MMCO mmco[MAX_MMCO_COUNT];
460 int mmco_index;
461
462 int long_ref_count; ///< number of actual long term references
463 int short_ref_count; ///< number of actual short term references
464
465 //data partitioning
466 GetBitContext intra_gb;
467 GetBitContext inter_gb;
468 GetBitContext *intra_gb_ptr;
469 GetBitContext *inter_gb_ptr;
470
c6727809 471 DECLARE_ALIGNED_16(DCTELEM, mb)[16*24];
2cab6401 472 DCTELEM mb_padding[256]; ///< 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
26b4fe82
AJ
473
474 /**
475 * Cabac
476 */
477 CABACContext cabac;
478 uint8_t cabac_state[460];
479 int cabac_init_idc;
480
481 /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
482 uint16_t *cbp_table;
483 int cbp;
484 int top_cbp;
485 int left_cbp;
486 /* chroma_pred_mode for i4x4 or i16x16, else 0 */
487 uint8_t *chroma_pred_mode_table;
488 int last_qscale_diff;
489 int16_t (*mvd_table[2])[2];
0b69d625 490 DECLARE_ALIGNED_16(int16_t, mvd_cache)[2][5*8][2];
26b4fe82
AJ
491 uint8_t *direct_table;
492 uint8_t direct_cache[5*8];
493
494 uint8_t zigzag_scan[16];
495 uint8_t zigzag_scan8x8[64];
496 uint8_t zigzag_scan8x8_cavlc[64];
497 uint8_t field_scan[16];
498 uint8_t field_scan8x8[64];
499 uint8_t field_scan8x8_cavlc[64];
500 const uint8_t *zigzag_scan_q0;
501 const uint8_t *zigzag_scan8x8_q0;
502 const uint8_t *zigzag_scan8x8_cavlc_q0;
503 const uint8_t *field_scan_q0;
504 const uint8_t *field_scan8x8_q0;
505 const uint8_t *field_scan8x8_cavlc_q0;
506
507 int x264_build;
afebe2f7
508
509 /**
510 * @defgroup multithreading Members for slice based multithreading
511 * @{
512 */
513 struct H264Context *thread_context[MAX_THREADS];
514
515 /**
516 * current slice number, used to initalize slice_num of each thread/context
517 */
518 int current_slice;
519
520 /**
521 * Max number of threads / contexts.
522 * This is equal to AVCodecContext.thread_count unless
523 * multithreaded decoding is impossible, in which case it is
524 * reduced to 1.
525 */
526 int max_contexts;
527
528 /**
529 * 1 if the single thread fallback warning has already been
530 * displayed, 0 otherwise.
531 */
532 int single_decode_warning;
533
534 int last_slice_type;
535 /** @} */
536
64514ee8
AS
537 int mb_xy;
538
1fdcdf8d 539 uint32_t svq3_watermark_key;
b09a7c05
540
541 /**
542 * pic_struct in picture timing SEI message
543 */
544 SEI_PicStructType sei_pic_struct;
89db0bae 545
37a558fe 546 /**
b19d493f
HY
547 * Complement sei_pic_struct
548 * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames.
549 * However, soft telecined frames may have these values.
550 * This is used in an attempt to flag soft telecine progressive.
551 */
552 int prev_interlaced_frame;
553
554 /**
70e01da3
IS
555 * Bit set of clock types for fields/frames in picture timing SEI message.
556 * For each found ct_type, appropriate bit is set (e.g., bit 1 for
557 * interlaced).
558 */
559 int sei_ct_type;
560
561 /**
ff594f81
IS
562 * dpb_output_delay in picture timing SEI message, see H.264 C.2.2
563 */
564 int sei_dpb_output_delay;
565
566 /**
cf6065ca
IS
567 * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2
568 */
569 int sei_cpb_removal_delay;
570
571 /**
37a558fe
IS
572 * recovery_frame_cnt from SEI message
573 *
574 * Set to -1 if no recovery point SEI message found or to number of frames
575 * before playback synchronizes. Frames having recovery point are key
576 * frames.
577 */
578 int sei_recovery_frame_cnt;
579
89db0bae 580 int is_complex;
cb99c652
GB
581
582 int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
583 int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
2ea89d92
IS
584
585 // Timestamp stuff
586 int sei_buffering_period_present; ///< Buffering period SEI flag
587 int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
26b4fe82
AJ
588}H264Context;
589
889fce8e
MN
590
591extern const uint8_t ff_h264_chroma_qp[52];
592
881b5b80
RC
593void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);
594
595void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
889fce8e 596
1790a5e9
IS
597/**
598 * Decode SEI
599 */
600int ff_h264_decode_sei(H264Context *h);
601
602/**
603 * Decode SPS
604 */
605int ff_h264_decode_seq_parameter_set(H264Context *h);
606
607/**
608 * Decode PPS
609 */
610int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
611
612/**
613 * Decodes a network abstraction layer unit.
614 * @param consumed is the number of bytes used as input
615 * @param length is the length of the array
616 * @param dst_length is the number of decoded bytes FIXME here or a decode rbsp tailing?
617 * @returns decoded bytes, might be src+1 if no escapes
618 */
619const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length);
620
621/**
622 * identifies the exact end of the bitstream
623 * @return the length of the trailing, or 0 if damaged
624 */
625int ff_h264_decode_rbsp_trailing(H264Context *h, const uint8_t *src);
626
15861962
RD
627/**
628 * frees any data that may have been allocated in the H264 context like SPS, PPS etc.
629 */
cbf1eae9 630av_cold void ff_h264_free_context(H264Context *h);
15861962 631
75dd6938
LA
632/**
633 * reconstructs bitstream slice_type.
634 */
0dc343d4 635int ff_h264_get_slice_type(const H264Context *h);
75dd6938 636
903d58f6
MN
637/**
638 * allocates tables.
639 * needs width/height
640 */
641int ff_h264_alloc_tables(H264Context *h);
642
643/**
ea6f00c4
MN
644 * fills the default_ref_list.
645 */
646int ff_h264_fill_default_ref_list(H264Context *h);
647
648int ff_h264_decode_ref_pic_list_reordering(H264Context *h);
649void ff_h264_fill_mbaff_ref_list(H264Context *h);
650void ff_h264_remove_all_refs(H264Context *h);
651
652/**
653 * Executes the reference picture marking (memory management control operations).
654 */
655int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
656
657int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb);
658
659
660/**
903d58f6
MN
661 * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
662 */
2bedc0e8
MN
663int ff_h264_check_intra4x4_pred_mode(H264Context *h);
664
665/**
666 * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
667 */
903d58f6
MN
668int ff_h264_check_intra_pred_mode(H264Context *h, int mode);
669
670void ff_h264_write_back_intra_pred_mode(H264Context *h);
671void ff_h264_hl_decode_mb(H264Context *h);
672int ff_h264_frame_start(H264Context *h);
673av_cold int ff_h264_decode_init(AVCodecContext *avctx);
674av_cold int ff_h264_decode_end(AVCodecContext *avctx);
e1e94902
MN
675av_cold void ff_h264_decode_init_vlc(void);
676
677/**
678 * decodes a macroblock
679 * @returns 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
680 */
681int ff_h264_decode_mb_cavlc(H264Context *h);
903d58f6 682
cc51b282
MN
683/**
684 * decodes a CABAC coded macroblock
685 * @returns 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
686 */
687int ff_h264_decode_mb_cabac(H264Context *h);
688
689void ff_h264_init_cabac_states(H264Context *h);
690
943f69a6
MN
691void ff_h264_direct_dist_scale_factor(H264Context * const h);
692void ff_h264_direct_ref_list_init(H264Context * const h);
693void ff_h264_pred_direct_motion(H264Context * const h, int *mb_type);
694
082cf971
MN
695void 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);
696void 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);
697
9c095463
MN
698/**
699 * Reset SEI values at the beginning of the frame.
700 *
701 * @param h H.264 context.
702 */
703void ff_h264_reset_sei(H264Context *h);
704
705
903d58f6
MN
706/*
707o-o o-o
708 / / /
709o-o o-o
710 ,---'
711o-o o-o
712 / / /
713o-o o-o
714*/
715//This table must be here because scan8[constant] must be known at compiletime
716static const uint8_t scan8[16 + 2*4]={
717 4+1*8, 5+1*8, 4+2*8, 5+2*8,
718 6+1*8, 7+1*8, 6+2*8, 7+2*8,
719 4+3*8, 5+3*8, 4+4*8, 5+4*8,
720 6+3*8, 7+3*8, 6+4*8, 7+4*8,
721 1+1*8, 2+1*8,
722 1+2*8, 2+2*8,
723 1+4*8, 2+4*8,
724 1+5*8, 2+5*8,
725};
726
727static av_always_inline uint32_t pack16to32(int a, int b){
728#if HAVE_BIGENDIAN
729 return (b&0xFFFF) + (a<<16);
730#else
731 return (a&0xFFFF) + (b<<16);
732#endif
733}
734
735/**
082cf971
MN
736 * gets the chroma qp.
737 */
738static inline int get_chroma_qp(H264Context *h, int t, int qscale){
739 return h->pps.chroma_qp_table[t][qscale];
740}
741
e1e94902
MN
742static inline void pred_pskip_motion(H264Context * const h, int * const mx, int * const my);
743
c1bb66ac 744static void fill_decode_neighbors(H264Context *h, int mb_type){
e1e94902
MN
745 MpegEncContext * const s = &h->s;
746 const int mb_xy= h->mb_xy;
747 int topleft_xy, top_xy, topright_xy, left_xy[2];
b6303e6d
MN
748 static const uint8_t left_block_options[4][16]={
749 {0,1,2,3,7,10,8,11,7+0*8, 7+1*8, 7+2*8, 7+3*8, 2+0*8, 2+3*8, 2+1*8, 2+2*8},
750 {2,2,3,3,8,11,8,11,7+2*8, 7+2*8, 7+3*8, 7+3*8, 2+1*8, 2+2*8, 2+1*8, 2+2*8},
751 {0,0,1,1,7,10,7,10,7+0*8, 7+0*8, 7+1*8, 7+1*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8},
752 {0,2,0,2,7,10,7,10,7+0*8, 7+2*8, 7+0*8, 7+2*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8}
e1e94902
MN
753 };
754
c1bb66ac
MN
755 h->topleft_partition= -1;
756
806ac67b 757 top_xy = mb_xy - (s->mb_stride << MB_FIELD);
e1e94902 758
e1e94902
MN
759 /* Wow, what a mess, why didn't they simplify the interlacing & intra
760 * stuff, I can't imagine that these complex rules are worth it. */
761
762 topleft_xy = top_xy - 1;
763 topright_xy= top_xy + 1;
764 left_xy[1] = left_xy[0] = mb_xy-1;
c1bb66ac 765 h->left_block = left_block_options[0];
e1e94902 766 if(FRAME_MBAFF){
806ac67b 767 const int left_mb_field_flag = IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]);
e1e94902 768 const int curr_mb_field_flag = IS_INTERLACED(mb_type);
806ac67b
MN
769 if(s->mb_y&1){
770 if (left_mb_field_flag != curr_mb_field_flag) {
771 left_xy[1] = left_xy[0] = mb_xy - s->mb_stride - 1;
772 if (curr_mb_field_flag) {
773 left_xy[1] += s->mb_stride;
c1bb66ac 774 h->left_block = left_block_options[3];
806ac67b
MN
775 } else {
776 topleft_xy += s->mb_stride;
777 // take top left mv from the middle of the mb, as opposed to all other modes which use the bottom right partition
c1bb66ac
MN
778 h->topleft_partition = 0;
779 h->left_block = left_block_options[1];
806ac67b
MN
780 }
781 }
782 }else{
783 if(curr_mb_field_flag){
784 topleft_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy - 1]>>7)&1)-1);
785 topright_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy + 1]>>7)&1)-1);
786 top_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy ]>>7)&1)-1);
787 }
788 if (left_mb_field_flag != curr_mb_field_flag) {
789 left_xy[1] = left_xy[0] = mb_xy - 1;
790 if (curr_mb_field_flag) {
791 left_xy[1] += s->mb_stride;
c1bb66ac 792 h->left_block = left_block_options[3];
806ac67b 793 } else {
c1bb66ac 794 h->left_block = left_block_options[2];
806ac67b 795 }
e1e94902
MN
796 }
797 }
798 }
799
c1bb66ac
MN
800 h->topleft_mb_xy = topleft_xy;
801 h->top_mb_xy = top_xy;
802 h->topright_mb_xy= topright_xy;
e1e94902
MN
803 h->left_mb_xy[0] = left_xy[0];
804 h->left_mb_xy[1] = left_xy[1];
c1bb66ac
MN
805 //FIXME do we need all in the context?
806 h->topleft_type = h->slice_table[topleft_xy ] == h->slice_num ? s->current_picture.mb_type[topleft_xy] : 0;
807 h->top_type = h->slice_table[top_xy ] == h->slice_num ? s->current_picture.mb_type[top_xy] : 0;
808 h->topright_type= h->slice_table[topright_xy] == h->slice_num ? s->current_picture.mb_type[topright_xy]: 0;
809 h->left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
810 h->left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
811}
812
813static void fill_decode_caches(H264Context *h, int mb_type){
814 MpegEncContext * const s = &h->s;
c1bb66ac
MN
815 int topleft_xy, top_xy, topright_xy, left_xy[2];
816 int topleft_type, top_type, topright_type, left_type[2];
817 const uint8_t * left_block= h->left_block;
818 int i;
819
820 topleft_xy = h->topleft_mb_xy ;
821 top_xy = h->top_mb_xy ;
822 topright_xy = h->topright_mb_xy;
823 left_xy[0] = h->left_mb_xy[0] ;
824 left_xy[1] = h->left_mb_xy[1] ;
825 topleft_type = h->topleft_type ;
826 top_type = h->top_type ;
827 topright_type= h->topright_type ;
828 left_type[0] = h->left_type[0] ;
829 left_type[1] = h->left_type[1] ;
e1e94902 830
2e4362af 831 if(!IS_SKIP(mb_type)){
7a93858a
MN
832 if(IS_INTRA(mb_type)){
833 int type_mask= h->pps.constrained_intra_pred ? IS_INTRA(-1) : -1;
834 h->topleft_samples_available=
835 h->top_samples_available=
836 h->left_samples_available= 0xFFFF;
837 h->topright_samples_available= 0xEEEA;
838
839 if(!(top_type & type_mask)){
840 h->topleft_samples_available= 0xB3FF;
841 h->top_samples_available= 0x33FF;
842 h->topright_samples_available= 0x26EA;
843 }
844 if(IS_INTERLACED(mb_type) != IS_INTERLACED(left_type[0])){
845 if(IS_INTERLACED(mb_type)){
846 if(!(left_type[0] & type_mask)){
847 h->topleft_samples_available&= 0xDFFF;
848 h->left_samples_available&= 0x5FFF;
849 }
850 if(!(left_type[1] & type_mask)){
851 h->topleft_samples_available&= 0xFF5F;
852 h->left_samples_available&= 0xFF5F;
853 }
854 }else{
855 int left_typei = h->slice_table[left_xy[0] + s->mb_stride ] == h->slice_num
856 ? s->current_picture.mb_type[left_xy[0] + s->mb_stride] : 0;
857 assert(left_xy[0] == left_xy[1]);
858 if(!((left_typei & type_mask) && (left_type[0] & type_mask))){
859 h->topleft_samples_available&= 0xDF5F;
860 h->left_samples_available&= 0x5F5F;
861 }
e1e94902
MN
862 }
863 }else{
7a93858a 864 if(!(left_type[0] & type_mask)){
e1e94902
MN
865 h->topleft_samples_available&= 0xDF5F;
866 h->left_samples_available&= 0x5F5F;
867 }
868 }
e1e94902 869
7a93858a
MN
870 if(!(topleft_type & type_mask))
871 h->topleft_samples_available&= 0x7FFF;
e1e94902 872
7a93858a
MN
873 if(!(topright_type & type_mask))
874 h->topright_samples_available&= 0xFBFF;
e1e94902 875
7a93858a
MN
876 if(IS_INTRA4x4(mb_type)){
877 if(IS_INTRA4x4(top_type)){
878 h->intra4x4_pred_mode_cache[4+8*0]= h->intra4x4_pred_mode[top_xy][4];
879 h->intra4x4_pred_mode_cache[5+8*0]= h->intra4x4_pred_mode[top_xy][5];
880 h->intra4x4_pred_mode_cache[6+8*0]= h->intra4x4_pred_mode[top_xy][6];
881 h->intra4x4_pred_mode_cache[7+8*0]= h->intra4x4_pred_mode[top_xy][3];
e1e94902
MN
882 }else{
883 int pred;
7a93858a 884 if(!(top_type & type_mask))
e1e94902
MN
885 pred= -1;
886 else{
887 pred= 2;
888 }
7a93858a
MN
889 h->intra4x4_pred_mode_cache[4+8*0]=
890 h->intra4x4_pred_mode_cache[5+8*0]=
891 h->intra4x4_pred_mode_cache[6+8*0]=
892 h->intra4x4_pred_mode_cache[7+8*0]= pred;
893 }
894 for(i=0; i<2; i++){
895 if(IS_INTRA4x4(left_type[i])){
896 h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[0+2*i]];
897 h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[1+2*i]];
898 }else{
899 int pred;
900 if(!(left_type[i] & type_mask))
901 pred= -1;
902 else{
903 pred= 2;
904 }
905 h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]=
906 h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= pred;
907 }
e1e94902
MN
908 }
909 }
910 }
e1e94902
MN
911
912
913/*
9140 . T T. T T T T
9151 L . .L . . . .
9162 L . .L . . . .
9173 . T TL . . . .
9184 L . .L . . . .
9195 L . .. . . . .
920*/
921//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec)
922 if(top_type){
19769ece 923 AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][4+3*8]);
7a93858a
MN
924 h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][1+1*8];
925 h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][2+1*8];
e1e94902 926
7a93858a
MN
927 h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][1+2*8];
928 h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][2+2*8];
599fe45b 929 }else {
7a93858a
MN
930 h->non_zero_count_cache[1+8*0]=
931 h->non_zero_count_cache[2+8*0]=
e1e94902 932
7a93858a
MN
933 h->non_zero_count_cache[1+8*3]=
934 h->non_zero_count_cache[2+8*3]=
19769ece 935 AV_WN32A(&h->non_zero_count_cache[4+8*0], CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040);
e1e94902
MN
936 }
937
938 for (i=0; i<2; i++) {
939 if(left_type[i]){
b6303e6d
MN
940 h->non_zero_count_cache[3+8*1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]];
941 h->non_zero_count_cache[3+8*2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]];
7a93858a
MN
942 h->non_zero_count_cache[0+8*1 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]];
943 h->non_zero_count_cache[0+8*4 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]];
599fe45b 944 }else{
7a93858a
MN
945 h->non_zero_count_cache[3+8*1 + 2*8*i]=
946 h->non_zero_count_cache[3+8*2 + 2*8*i]=
947 h->non_zero_count_cache[0+8*1 + 8*i]=
948 h->non_zero_count_cache[0+8*4 + 8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64;
e1e94902
MN
949 }
950 }
951
599fe45b 952 if( CABAC ) {
e1e94902
MN
953 // top_cbp
954 if(top_type) {
955 h->top_cbp = h->cbp_table[top_xy];
956 } else if(IS_INTRA(mb_type)) {
3d2c3ef4 957 h->top_cbp = 0x1CF;
e1e94902 958 } else {
3d2c3ef4 959 h->top_cbp = 0x00F;
e1e94902
MN
960 }
961 // left_cbp
962 if (left_type[0]) {
963 h->left_cbp = h->cbp_table[left_xy[0]] & 0x1f0;
964 } else if(IS_INTRA(mb_type)) {
3d2c3ef4 965 h->left_cbp = 0x1CF;
e1e94902 966 } else {
3d2c3ef4 967 h->left_cbp = 0x00F;
e1e94902
MN
968 }
969 if (left_type[0]) {
970 h->left_cbp |= ((h->cbp_table[left_xy[0]]>>((left_block[0]&(~1))+1))&0x1) << 1;
971 }
972 if (left_type[1]) {
973 h->left_cbp |= ((h->cbp_table[left_xy[1]]>>((left_block[2]&(~1))+1))&0x1) << 3;
974 }
975 }
2e4362af 976 }
e1e94902
MN
977
978#if 1
da452aca 979 if(IS_INTER(mb_type) || (IS_DIRECT(mb_type) && h->direct_spatial_mv_pred)){
e1e94902
MN
980 int list;
981 for(list=0; list<h->list_count; list++){
5ca43c25 982 if(!USES_LIST(mb_type, list)){
e1e94902
MN
983 /*if(!h->mv_cache_clean[list]){
984 memset(h->mv_cache [list], 0, 8*5*2*sizeof(int16_t)); //FIXME clean only input? clean at all?
985 memset(h->ref_cache[list], PART_NOT_AVAILABLE, 8*5*sizeof(int8_t));
986 h->mv_cache_clean[list]= 1;
987 }*/
988 continue;
989 }
da452aca
MN
990 assert(!(IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred));
991
e1e94902
MN
992 h->mv_cache_clean[list]= 0;
993
994 if(USES_LIST(top_type, list)){
995 const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
996 const int b8_xy= h->mb2b8_xy[top_xy] + h->b8_stride;
0b69d625 997 AV_COPY128(h->mv_cache[list][scan8[0] + 0 - 1*8], s->current_picture.motion_val[list][b_xy + 0]);
7a93858a
MN
998 h->ref_cache[list][scan8[0] + 0 - 1*8]=
999 h->ref_cache[list][scan8[0] + 1 - 1*8]= s->current_picture.ref_index[list][b8_xy + 0];
1000 h->ref_cache[list][scan8[0] + 2 - 1*8]=
1001 h->ref_cache[list][scan8[0] + 3 - 1*8]= s->current_picture.ref_index[list][b8_xy + 1];
e1e94902 1002 }else{
0b69d625 1003 AV_ZERO128(h->mv_cache[list][scan8[0] + 0 - 1*8]);
19769ece 1004 AV_WN32A(&h->ref_cache[list][scan8[0] + 0 - 1*8], ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101);
e1e94902
MN
1005 }
1006
1007 for(i=0; i<2; i++){
1008 int cache_idx = scan8[0] - 1 + i*2*8;
1009 if(USES_LIST(left_type[i], list)){
1010 const int b_xy= h->mb2b_xy[left_xy[i]] + 3;
1011 const int b8_xy= h->mb2b8_xy[left_xy[i]] + 1;
19769ece
MR
1012 AV_COPY32(h->mv_cache[list][cache_idx ], s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0+i*2]]);
1013 AV_COPY32(h->mv_cache[list][cache_idx+8], s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[1+i*2]]);
7a93858a
MN
1014 h->ref_cache[list][cache_idx ]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[0+i*2]>>1)];
1015 h->ref_cache[list][cache_idx+8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[1+i*2]>>1)];
e1e94902 1016 }else{
19769ece
MR
1017 AV_ZERO32(h->mv_cache [list][cache_idx ]);
1018 AV_ZERO32(h->mv_cache [list][cache_idx+8]);
e1e94902 1019 h->ref_cache[list][cache_idx ]=
aebf3123 1020 h->ref_cache[list][cache_idx+8]= (left_type[i]) ? LIST_NOT_USED : PART_NOT_AVAILABLE;
e1e94902
MN
1021 }
1022 }
1023
e1e94902 1024 if(USES_LIST(topleft_type, list)){
c1bb66ac
MN
1025 const int b_xy = h->mb2b_xy [topleft_xy] + 3 + h->b_stride + (h->topleft_partition & 2*h->b_stride);
1026 const int b8_xy= h->mb2b8_xy[topleft_xy] + 1 + (h->topleft_partition & h->b8_stride);
19769ece 1027 AV_COPY32(h->mv_cache[list][scan8[0] - 1 - 1*8], s->current_picture.motion_val[list][b_xy]);
e1e94902
MN
1028 h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
1029 }else{
19769ece 1030 AV_ZERO32(h->mv_cache[list][scan8[0] - 1 - 1*8]);
e1e94902
MN
1031 h->ref_cache[list][scan8[0] - 1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
1032 }
1033
1034 if(USES_LIST(topright_type, list)){
1035 const int b_xy= h->mb2b_xy[topright_xy] + 3*h->b_stride;
1036 const int b8_xy= h->mb2b8_xy[topright_xy] + h->b8_stride;
19769ece 1037 AV_COPY32(h->mv_cache[list][scan8[0] + 4 - 1*8], s->current_picture.motion_val[list][b_xy]);
e1e94902
MN
1038 h->ref_cache[list][scan8[0] + 4 - 1*8]= s->current_picture.ref_index[list][b8_xy];
1039 }else{
19769ece 1040 AV_ZERO32(h->mv_cache [list][scan8[0] + 4 - 1*8]);
e1e94902
MN
1041 h->ref_cache[list][scan8[0] + 4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
1042 }
1043
bb770c5b 1044 if((mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2)) && !FRAME_MBAFF)
e1e94902
MN
1045 continue;
1046
e2b28acf 1047 if(!(mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2))) {
e1e94902
MN
1048 h->ref_cache[list][scan8[5 ]+1] =
1049 h->ref_cache[list][scan8[7 ]+1] =
1050 h->ref_cache[list][scan8[13]+1] = //FIXME remove past 3 (init somewhere else)
1051 h->ref_cache[list][scan8[4 ]] =
1052 h->ref_cache[list][scan8[12]] = PART_NOT_AVAILABLE;
19769ece
MR
1053 AV_ZERO32(h->mv_cache [list][scan8[5 ]+1]);
1054 AV_ZERO32(h->mv_cache [list][scan8[7 ]+1]);
1055 AV_ZERO32(h->mv_cache [list][scan8[13]+1]); //FIXME remove past 3 (init somewhere else)
1056 AV_ZERO32(h->mv_cache [list][scan8[4 ]]);
1057 AV_ZERO32(h->mv_cache [list][scan8[12]]);
e1e94902 1058
cb9285a2 1059 if( CABAC ) {
e1e94902
MN
1060 /* XXX beurk, Load mvd */
1061 if(USES_LIST(top_type, list)){
1062 const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
0b69d625 1063 AV_COPY128(h->mvd_cache[list][scan8[0] + 0 - 1*8], h->mvd_table[list][b_xy + 0]);
e1e94902 1064 }else{
0b69d625 1065 AV_ZERO128(h->mvd_cache[list][scan8[0] + 0 - 1*8]);
e1e94902
MN
1066 }
1067 if(USES_LIST(left_type[0], list)){
1068 const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
19769ece
MR
1069 AV_COPY32(h->mvd_cache[list][scan8[0] - 1 + 0*8], h->mvd_table[list][b_xy + h->b_stride*left_block[0]]);
1070 AV_COPY32(h->mvd_cache[list][scan8[0] - 1 + 1*8], h->mvd_table[list][b_xy + h->b_stride*left_block[1]]);
e1e94902 1071 }else{
19769ece
MR
1072 AV_ZERO32(h->mvd_cache [list][scan8[0] - 1 + 0*8]);
1073 AV_ZERO32(h->mvd_cache [list][scan8[0] - 1 + 1*8]);
e1e94902
MN
1074 }
1075 if(USES_LIST(left_type[1], list)){
1076 const int b_xy= h->mb2b_xy[left_xy[1]] + 3;
19769ece
MR
1077 AV_COPY32(h->mvd_cache[list][scan8[0] - 1 + 2*8], h->mvd_table[list][b_xy + h->b_stride*left_block[2]]);
1078 AV_COPY32(h->mvd_cache[list][scan8[0] - 1 + 3*8], h->mvd_table[list][b_xy + h->b_stride*left_block[3]]);
e1e94902 1079 }else{
19769ece
MR
1080 AV_ZERO32(h->mvd_cache [list][scan8[0] - 1 + 2*8]);
1081 AV_ZERO32(h->mvd_cache [list][scan8[0] - 1 + 3*8]);
e1e94902 1082 }
19769ece
MR
1083 AV_ZERO32(h->mvd_cache [list][scan8[5 ]+1]);
1084 AV_ZERO32(h->mvd_cache [list][scan8[7 ]+1]);
1085 AV_ZERO32(h->mvd_cache [list][scan8[13]+1]); //FIXME remove past 3 (init somewhere else)
1086 AV_ZERO32(h->mvd_cache [list][scan8[4 ]]);
1087 AV_ZERO32(h->mvd_cache [list][scan8[12]]);
e1e94902
MN
1088
1089 if(h->slice_type_nos == FF_B_TYPE){
2dc380ca 1090 fill_rectangle(&h->direct_cache[scan8[0]], 4, 4, 8, MB_TYPE_16x16>>1, 1);
e1e94902
MN
1091
1092 if(IS_DIRECT(top_type)){
19769ece 1093 AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_DIRECT2>>1));
e1e94902
MN
1094 }else if(IS_8X8(top_type)){
1095 int b8_xy = h->mb2b8_xy[top_xy] + h->b8_stride;
1096 h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy];
1097 h->direct_cache[scan8[0] + 2 - 1*8]= h->direct_table[b8_xy + 1];
1098 }else{
19769ece 1099 AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_16x16>>1));
e1e94902
MN
1100 }
1101
1102 if(IS_DIRECT(left_type[0]))
2dc380ca 1103 h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_DIRECT2>>1;
e1e94902
MN
1104 else if(IS_8X8(left_type[0]))
1105 h->direct_cache[scan8[0] - 1 + 0*8]= h->direct_table[h->mb2b8_xy[left_xy[0]] + 1 + h->b8_stride*(left_block[0]>>1)];
1106 else
2dc380ca 1107 h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_16x16>>1;
e1e94902
MN
1108
1109 if(IS_DIRECT(left_type[1]))
2dc380ca 1110 h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_DIRECT2>>1;
e1e94902
MN
1111 else if(IS_8X8(left_type[1]))
1112 h->direct_cache[scan8[0] - 1 + 2*8]= h->direct_table[h->mb2b8_xy[left_xy[1]] + 1 + h->b8_stride*(left_block[2]>>1)];
1113 else
2dc380ca 1114 h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_16x16>>1;
e1e94902
MN
1115 }
1116 }
cb9285a2 1117 }
e1e94902
MN
1118 if(FRAME_MBAFF){
1119#define MAP_MVS\
1120 MAP_F2F(scan8[0] - 1 - 1*8, topleft_type)\
1121 MAP_F2F(scan8[0] + 0 - 1*8, top_type)\
1122 MAP_F2F(scan8[0] + 1 - 1*8, top_type)\
1123 MAP_F2F(scan8[0] + 2 - 1*8, top_type)\
1124 MAP_F2F(scan8[0] + 3 - 1*8, top_type)\
1125 MAP_F2F(scan8[0] + 4 - 1*8, topright_type)\
1126 MAP_F2F(scan8[0] - 1 + 0*8, left_type[0])\
1127 MAP_F2F(scan8[0] - 1 + 1*8, left_type[0])\
1128 MAP_F2F(scan8[0] - 1 + 2*8, left_type[1])\
1129 MAP_F2F(scan8[0] - 1 + 3*8, left_type[1])
1130 if(MB_FIELD){
1131#define MAP_F2F(idx, mb_type)\
1132 if(!IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
1133 h->ref_cache[list][idx] <<= 1;\
1134 h->mv_cache[list][idx][1] /= 2;\
1135 h->mvd_cache[list][idx][1] /= 2;\
1136 }
1137 MAP_MVS
1138#undef MAP_F2F
1139 }else{
1140#define MAP_F2F(idx, mb_type)\
1141 if(IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
1142 h->ref_cache[list][idx] >>= 1;\
1143 h->mv_cache[list][idx][1] <<= 1;\
1144 h->mvd_cache[list][idx][1] <<= 1;\
1145 }
1146 MAP_MVS
1147#undef MAP_F2F
1148 }
1149 }
1150 }
1151 }
1152#endif
1153
7a93858a 1154 h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
f432b43b
MN
1155}
1156
44a5e7b6
MN
1157/**
1158 *
1159 * @returns non zero if the loop filter can be skiped
1160 */
1161static int fill_filter_caches(H264Context *h, int mb_type){
599fe45b
MN
1162 MpegEncContext * const s = &h->s;
1163 const int mb_xy= h->mb_xy;
1164 int top_xy, left_xy[2];
1165 int top_type, left_type[2];
599fe45b
MN
1166
1167 top_xy = mb_xy - (s->mb_stride << MB_FIELD);
1168
1169 //FIXME deblocking could skip the intra and nnz parts.
1170
1171 /* Wow, what a mess, why didn't they simplify the interlacing & intra
1172 * stuff, I can't imagine that these complex rules are worth it. */
1173
1174 left_xy[1] = left_xy[0] = mb_xy-1;
1175 if(FRAME_MBAFF){
1176 const int left_mb_field_flag = IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]);
1177 const int curr_mb_field_flag = IS_INTERLACED(mb_type);
1178 if(s->mb_y&1){
1179 if (left_mb_field_flag != curr_mb_field_flag) {
8652e44a 1180 left_xy[0] -= s->mb_stride;
599fe45b
MN
1181 }
1182 }else{
1183 if(curr_mb_field_flag){
1184 top_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy ]>>7)&1)-1);
1185 }
1186 if (left_mb_field_flag != curr_mb_field_flag) {
8652e44a 1187 left_xy[1] += s->mb_stride;
599fe45b
MN
1188 }
1189 }
1190 }
1191
1192 h->top_mb_xy = top_xy;
1193 h->left_mb_xy[0] = left_xy[0];
1194 h->left_mb_xy[1] = left_xy[1];
1195 {
1196 //for sufficiently low qp, filtering wouldn't do anything
1197 //this is a conservative estimate: could also check beta_offset and more accurate chroma_qp
1198 int qp_thresh = h->qp_thresh; //FIXME strictly we should store qp_thresh for each mb of a slice
1199 int qp = s->current_picture.qscale_table[mb_xy];
1200 if(qp <= qp_thresh
1201 && (left_xy[0]<0 || ((qp + s->current_picture.qscale_table[left_xy[0]] + 1)>>1) <= qp_thresh)
1202 && (top_xy < 0 || ((qp + s->current_picture.qscale_table[top_xy ] + 1)>>1) <= qp_thresh)){
1203 if(!FRAME_MBAFF)
1204 return 1;
1205 if( (left_xy[0]< 0 || ((qp + s->current_picture.qscale_table[left_xy[1] ] + 1)>>1) <= qp_thresh)
1206 && (top_xy < s->mb_stride || ((qp + s->current_picture.qscale_table[top_xy -s->mb_stride] + 1)>>1) <= qp_thresh))
1207 return 1;
1208 }
1209 }
1210
1211 if(h->deblocking_filter == 2){
1212 h->top_type = top_type = h->slice_table[top_xy ] == h->slice_num ? s->current_picture.mb_type[top_xy] : 0;
1213 h->left_type[0]= left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
1214 h->left_type[1]= left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
1215 }else{
1216 h->top_type = top_type = h->slice_table[top_xy ] < 0xFFFF ? s->current_picture.mb_type[top_xy] : 0;
1217 h->left_type[0]= left_type[0] = h->slice_table[left_xy[0] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[0]] : 0;
1218 h->left_type[1]= left_type[1] = h->slice_table[left_xy[1] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[1]] : 0;
1219 }
1220 if(IS_INTRA(mb_type))
1221 return 0;
1222
1223 AV_COPY64(&h->non_zero_count_cache[0+8*1], &h->non_zero_count[mb_xy][ 0]);
1224 AV_COPY64(&h->non_zero_count_cache[0+8*2], &h->non_zero_count[mb_xy][ 8]);
19769ece
MR
1225 AV_COPY32(&h->non_zero_count_cache[0+8*5], &h->non_zero_count[mb_xy][16]);
1226 AV_COPY32(&h->non_zero_count_cache[4+8*3], &h->non_zero_count[mb_xy][20]);
599fe45b
MN
1227 AV_COPY64(&h->non_zero_count_cache[0+8*4], &h->non_zero_count[mb_xy][24]);
1228
1229 h->cbp= h->cbp_table[mb_xy];
1230
1231 {
1232 int list;
1233 for(list=0; list<h->list_count; list++){
1234 int8_t *ref;
1235 int y, b_stride;
1236 int16_t (*mv_dst)[2];
1237 int16_t (*mv_src)[2];
1238
1239 if(!USES_LIST(mb_type, list)){
1240 fill_rectangle( h->mv_cache[list][scan8[0]], 4, 4, 8, pack16to32(0,0), 4);
19769ece
MR
1241 AV_WN32A(&h->ref_cache[list][scan8[ 0]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
1242 AV_WN32A(&h->ref_cache[list][scan8[ 2]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
1243 AV_WN32A(&h->ref_cache[list][scan8[ 8]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
1244 AV_WN32A(&h->ref_cache[list][scan8[10]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
599fe45b
MN
1245 continue;
1246 }
1247
1248 ref = &s->current_picture.ref_index[list][h->mb2b8_xy[mb_xy]];
1249 {
056c5021 1250 int (*ref2frm)[64] = h->ref2frm[ h->slice_num&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
19769ece
MR
1251 AV_WN32A(&h->ref_cache[list][scan8[ 0]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
1252 AV_WN32A(&h->ref_cache[list][scan8[ 2]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
599fe45b 1253 ref += h->b8_stride;
19769ece
MR
1254 AV_WN32A(&h->ref_cache[list][scan8[ 8]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
1255 AV_WN32A(&h->ref_cache[list][scan8[10]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
599fe45b
MN
1256 }
1257
1258 b_stride = h->b_stride;
1259 mv_dst = &h->mv_cache[list][scan8[0]];
1260 mv_src = &s->current_picture.motion_val[list][4*s->mb_x + 4*s->mb_y*b_stride];
1261 for(y=0; y<4; y++){
1262 AV_COPY128(mv_dst + 8*y, mv_src + y*b_stride);
1263 }
1264
1265 }
1266 }
1267
1268
1269/*
12700 . T T. T T T T
12711 L . .L . . . .
12722 L . .L . . . .
12733 . T TL . . . .
12744 L . .L . . . .
12755 L . .. . . . .
1276*/
1277//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec)
1278 if(top_type){
19769ece 1279 AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][4+3*8]);
599fe45b
MN
1280 }
1281
1282 if(left_type[0]){
1283 h->non_zero_count_cache[3+8*1]= h->non_zero_count[left_xy[0]][7+0*8];
1284 h->non_zero_count_cache[3+8*2]= h->non_zero_count[left_xy[0]][7+1*8];
1285 h->non_zero_count_cache[3+8*3]= h->non_zero_count[left_xy[0]][7+2*8];
1286 h->non_zero_count_cache[3+8*4]= h->non_zero_count[left_xy[0]][7+3*8];
1287 }
1288
1289 // CAVLC 8x8dct requires NNZ values for residual decoding that differ from what the loop filter needs
1290 if(!CABAC && h->pps.transform_8x8_mode){
1291 if(IS_8x8DCT(top_type)){
1292 h->non_zero_count_cache[4+8*0]=
1293 h->non_zero_count_cache[5+8*0]= h->cbp_table[top_xy] & 4;
1294 h->non_zero_count_cache[6+8*0]=
1295 h->non_zero_count_cache[7+8*0]= h->cbp_table[top_xy] & 8;
1296 }
1297 if(IS_8x8DCT(left_type[0])){
1298 h->non_zero_count_cache[3+8*1]=
1299 h->non_zero_count_cache[3+8*2]= h->cbp_table[left_xy[0]]&2; //FIXME check MBAFF
1300 }
1301 if(IS_8x8DCT(left_type[1])){
1302 h->non_zero_count_cache[3+8*3]=
1303 h->non_zero_count_cache[3+8*4]= h->cbp_table[left_xy[1]]&8; //FIXME check MBAFF
1304 }
1305
1306 if(IS_8x8DCT(mb_type)){
1307 h->non_zero_count_cache[scan8[0 ]]= h->non_zero_count_cache[scan8[1 ]]=
1308 h->non_zero_count_cache[scan8[2 ]]= h->non_zero_count_cache[scan8[3 ]]= h->cbp & 1;
1309
1310 h->non_zero_count_cache[scan8[0+ 4]]= h->non_zero_count_cache[scan8[1+ 4]]=
1311 h->non_zero_count_cache[scan8[2+ 4]]= h->non_zero_count_cache[scan8[3+ 4]]= h->cbp & 2;
1312
1313 h->non_zero_count_cache[scan8[0+ 8]]= h->non_zero_count_cache[scan8[1+ 8]]=
1314 h->non_zero_count_cache[scan8[2+ 8]]= h->non_zero_count_cache[scan8[3+ 8]]= h->cbp & 4;
1315
1316 h->non_zero_count_cache[scan8[0+12]]= h->non_zero_count_cache[scan8[1+12]]=
1317 h->non_zero_count_cache[scan8[2+12]]= h->non_zero_count_cache[scan8[3+12]]= h->cbp & 8;
1318 }
1319 }
1320
1321 if(IS_INTER(mb_type) || IS_DIRECT(mb_type)){
1322 int list;
1323 for(list=0; list<h->list_count; list++){
1324 if(USES_LIST(top_type, list)){
1325 const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
1326 const int b8_xy= h->mb2b8_xy[top_xy] + h->b8_stride;
056c5021 1327 int (*ref2frm)[64] = h->ref2frm[ h->slice_table[top_xy]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
599fe45b
MN
1328 AV_COPY128(h->mv_cache[list][scan8[0] + 0 - 1*8], s->current_picture.motion_val[list][b_xy + 0]);
1329 h->ref_cache[list][scan8[0] + 0 - 1*8]=
1330 h->ref_cache[list][scan8[0] + 1 - 1*8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 0]];
1331 h->ref_cache[list][scan8[0] + 2 - 1*8]=
1332 h->ref_cache[list][scan8[0] + 3 - 1*8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 1]];
1333 }else{
1334 AV_ZERO128(h->mv_cache[list][scan8[0] + 0 - 1*8]);
19769ece 1335 AV_WN32A(&h->ref_cache[list][scan8[0] + 0 - 1*8], ((LIST_NOT_USED)&0xFF)*0x01010101u);
599fe45b
MN
1336 }
1337
1338 if(!IS_INTERLACED(mb_type^left_type[0])){
1339 if(USES_LIST(left_type[0], list)){
1340 const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
1341 const int b8_xy= h->mb2b8_xy[left_xy[0]] + 1;
056c5021 1342 int (*ref2frm)[64] = h->ref2frm[ h->slice_table[left_xy[0]]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
19769ece
MR
1343 AV_COPY32(h->mv_cache[list][scan8[0] - 1 + 0 ], s->current_picture.motion_val[list][b_xy + h->b_stride*0]);
1344 AV_COPY32(h->mv_cache[list][scan8[0] - 1 + 8 ], s->current_picture.motion_val[list][b_xy + h->b_stride*1]);
1345 AV_COPY32(h->mv_cache[list][scan8[0] - 1 +16 ], s->current_picture.motion_val[list][b_xy + h->b_stride*2]);
1346 AV_COPY32(h->mv_cache[list][scan8[0] - 1 +24 ], s->current_picture.motion_val[list][b_xy + h->b_stride*3]);
599fe45b
MN
1347 h->ref_cache[list][scan8[0] - 1 + 0 ]=
1348 h->ref_cache[list][scan8[0] - 1 + 8 ]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + h->b8_stride*0]];
1349 h->ref_cache[list][scan8[0] - 1 +16 ]=
1350 h->ref_cache[list][scan8[0] - 1 +24 ]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + h->b8_stride*1]];
1351 }else{
19769ece
MR
1352 AV_ZERO32(h->mv_cache [list][scan8[0] - 1 + 0 ]);
1353 AV_ZERO32(h->mv_cache [list][scan8[0] - 1 + 8 ]);
1354 AV_ZERO32(h->mv_cache [list][scan8[0] - 1 +16 ]);
1355 AV_ZERO32(h->mv_cache [list][scan8[0] - 1 +24 ]);
599fe45b
MN
1356 h->ref_cache[list][scan8[0] - 1 + 0 ]=
1357 h->ref_cache[list][scan8[0] - 1 + 8 ]=
1358 h->ref_cache[list][scan8[0] - 1 + 16 ]=
1359 h->ref_cache[list][scan8[0] - 1 + 24 ]= LIST_NOT_USED;
1360 }
1361 }
1362 }
1363 }
1364
1365 return 0;
f432b43b
MN
1366}
1367
e1e94902
MN
1368/**
1369 * gets the predicted intra4x4 prediction mode.
1370 */
1371static inline int pred_intra_mode(H264Context *h, int n){
1372 const int index8= scan8[n];
1373 const int left= h->intra4x4_pred_mode_cache[index8 - 1];
1374 const int top = h->intra4x4_pred_mode_cache[index8 - 8];
1375 const int min= FFMIN(left, top);
1376
1377 tprintf(h->s.avctx, "mode:%d %d min:%d\n", left ,top, min);
1378
1379 if(min<0) return DC_PRED;
1380 else return min;
1381}
1382
1383static inline void write_back_non_zero_count(H264Context *h){
1384 const int mb_xy= h->mb_xy;
1385
0b69d625
AS
1386 AV_COPY64(&h->non_zero_count[mb_xy][ 0], &h->non_zero_count_cache[0+8*1]);
1387 AV_COPY64(&h->non_zero_count[mb_xy][ 8], &h->non_zero_count_cache[0+8*2]);
19769ece
MR
1388 AV_COPY32(&h->non_zero_count[mb_xy][16], &h->non_zero_count_cache[0+8*5]);
1389 AV_COPY32(&h->non_zero_count[mb_xy][20], &h->non_zero_count_cache[4+8*3]);
0b69d625 1390 AV_COPY64(&h->non_zero_count[mb_xy][24], &h->non_zero_count_cache[0+8*4]);
e1e94902
MN
1391}
1392
1393static inline void write_back_motion(H264Context *h, int mb_type){
1394 MpegEncContext * const s = &h->s;
1395 const int b_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
1396 const int b8_xy= 2*s->mb_x + 2*s->mb_y*h->b8_stride;
1397 int list;
1398
1399 if(!USES_LIST(mb_type, 0))
1400 fill_rectangle(&s->current_picture.ref_index[0][b8_xy], 2, 2, h->b8_stride, (uint8_t)LIST_NOT_USED, 1);
1401
1402 for(list=0; list<h->list_count; list++){
0b69d625
AS
1403 int y, b_stride;
1404 int16_t (*mv_dst)[2];
1405 int16_t (*mv_src)[2];
1406
e1e94902
MN
1407 if(!USES_LIST(mb_type, list))
1408 continue;
1409
0b69d625
AS
1410 b_stride = h->b_stride;
1411 mv_dst = &s->current_picture.motion_val[list][b_xy];
1412 mv_src = &h->mv_cache[list][scan8[0]];
e1e94902 1413 for(y=0; y<4; y++){
0b69d625 1414 AV_COPY128(mv_dst + y*b_stride, mv_src + 8*y);
e1e94902 1415 }
ddd60f28 1416 if( CABAC ) {
0b69d625
AS
1417 int16_t (*mvd_dst)[2] = &h->mvd_table[list][b_xy];
1418 int16_t (*mvd_src)[2] = &h->mvd_cache[list][scan8[0]];
e1e94902 1419 if(IS_SKIP(mb_type))
0b69d625 1420 fill_rectangle(mvd_dst, 4, 4, h->b_stride, 0, 4);
e1e94902
MN
1421 else
1422 for(y=0; y<4; y++){
0b69d625 1423 AV_COPY128(mvd_dst + y*b_stride, mvd_src + 8*y);
e1e94902
MN
1424 }
1425 }
1426
1427 {
1428 int8_t *ref_index = &s->current_picture.ref_index[list][b8_xy];
1429 ref_index[0+0*h->b8_stride]= h->ref_cache[list][scan8[0]];
1430 ref_index[1+0*h->b8_stride]= h->ref_cache[list][scan8[4]];
1431 ref_index[0+1*h->b8_stride]= h->ref_cache[list][scan8[8]];
1432 ref_index[1+1*h->b8_stride]= h->ref_cache[list][scan8[12]];
1433 }
1434 }
1435
ddd60f28 1436 if(h->slice_type_nos == FF_B_TYPE && CABAC){
e1e94902
MN
1437 if(IS_8X8(mb_type)){
1438 uint8_t *direct_table = &h->direct_table[b8_xy];
2dc380ca
MN
1439 direct_table[1+0*h->b8_stride] = h->sub_mb_type[1]>>1;
1440 direct_table[0+1*h->b8_stride] = h->sub_mb_type[2]>>1;
1441 direct_table[1+1*h->b8_stride] = h->sub_mb_type[3]>>1;
e1e94902
MN
1442 }
1443 }
1444}
1445
1446static inline int get_dct8x8_allowed(H264Context *h){
1447 if(h->sps.direct_8x8_inference_flag)
19769ece 1448 return !(AV_RN64A(h->sub_mb_type) & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8 )*0x0001000100010001ULL));
e1e94902 1449 else
19769ece 1450 return !(AV_RN64A(h->sub_mb_type) & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8|MB_TYPE_DIRECT2)*0x0001000100010001ULL));
e1e94902
MN
1451}
1452
e1e94902
MN
1453/**
1454 * decodes a P_SKIP or B_SKIP macroblock
1455 */
1456static void decode_mb_skip(H264Context *h){
1457 MpegEncContext * const s = &h->s;
1458 const int mb_xy= h->mb_xy;
1459 int mb_type=0;
1460
c988f975 1461 memset(h->non_zero_count[mb_xy], 0, 32);
e1e94902
MN
1462 memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui
1463
1464 if(MB_FIELD)
1465 mb_type|= MB_TYPE_INTERLACED;
1466
1467 if( h->slice_type_nos == FF_B_TYPE )
1468 {
1469 // just for fill_caches. pred_direct_motion will set the real mb_type
5ca43c25 1470 mb_type|= MB_TYPE_L0L1|MB_TYPE_DIRECT2|MB_TYPE_SKIP;
c1bb66ac
MN
1471 if(h->direct_spatial_mv_pred){
1472 fill_decode_neighbors(h, mb_type);
f432b43b 1473 fill_decode_caches(h, mb_type); //FIXME check what is needed and what not ...
c1bb66ac 1474 }
e1e94902
MN
1475 ff_h264_pred_direct_motion(h, &mb_type);
1476 mb_type|= MB_TYPE_SKIP;
1477 }
1478 else
1479 {
1480 int mx, my;
1481 mb_type|= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P1L0|MB_TYPE_SKIP;
1482
c1bb66ac 1483 fill_decode_neighbors(h, mb_type);
f432b43b 1484 fill_decode_caches(h, mb_type); //FIXME check what is needed and what not ...
e1e94902
MN
1485 pred_pskip_motion(h, &mx, &my);
1486 fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, 0, 1);
1487 fill_rectangle( h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mx,my), 4);
1488 }
1489
1490 write_back_motion(h, mb_type);
1491 s->current_picture.mb_type[mb_xy]= mb_type;
1492 s->current_picture.qscale_table[mb_xy]= s->qscale;
1493 h->slice_table[ mb_xy ]= h->slice_num;
1494 h->prev_mb_skipped= 1;
1495}
1496
7c2de274
MN
1497#include "h264_mvpred.h" //For pred_pskip_motion()
1498
98790382 1499#endif /* AVCODEC_H264_H */