1 /*****************************************************************************
2 * common.h: h264 encoder
3 *****************************************************************************
4 * Copyright (C) 2003 Laurent Aimar
5 * $Id: common.h,v 1.1 2004/06/03 19:27:06 fenrir Exp $
7 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program; if not, write to the Free Software
21 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
22 *****************************************************************************/
27 /****************************************************************************
29 ****************************************************************************/
30 #define X264_MIN(a,b) ( (a)<(b) ? (a) : (b) )
31 #define X264_MAX(a,b) ( (a)>(b) ? (a) : (b) )
32 #define X264_MIN3(a,b,c) X264_MIN((a),X264_MIN((b),(c)))
33 #define X264_MAX3(a,b,c) X264_MAX((a),X264_MAX((b),(c)))
34 #define X264_MIN4(a,b,c,d) X264_MIN((a),X264_MIN3((b),(c),(d)))
35 #define X264_MAX4(a,b,c,d) X264_MAX((a),X264_MAX3((b),(c),(d)))
36 #define XCHG(type,a,b) { type t = a; a = b; b = t; }
37 #define FIX8(f) ((int)(f*(1<<8)+.5))
39 #define CHECKED_MALLOC( var, size )\
41 var = x264_malloc( size );\
44 x264_log( h, X264_LOG_ERROR, "malloc failed\n" );\
49 #define X264_BFRAME_MAX 16
50 #define X264_THREAD_MAX 128
51 #define X264_SLICE_MAX 4
52 #define X264_NAL_MAX (4 + X264_SLICE_MAX)
54 // number of pixels (per thread) in progress at any given time.
55 // 16 for the macroblock in progress + 3 for deblocking + 3 for motion compensation filter + 2 for extra safety
56 #define X264_THREAD_HEIGHT 24
58 /****************************************************************************
60 ****************************************************************************/
78 /****************************************************************************
80 ****************************************************************************/
81 /* x264_malloc : will do or emulate a memalign
82 * you have to use x264_free for buffers allocated with x264_malloc */
83 void *x264_malloc( int );
84 void *x264_realloc( void *p, int i_size );
85 void x264_free( void * );
87 /* x264_slurp_file: malloc space for the whole file and read it */
88 char *x264_slurp_file( const char *filename );
90 /* mdate: return the current date in microsecond */
91 int64_t x264_mdate( void );
93 /* x264_param2string: return a (malloced) string containing most of
94 * the encoding options */
95 char *x264_param2string( x264_param_t *p, int b_res );
98 void x264_log( x264_t *h, int i_level, const char *psz_fmt, ... );
100 void x264_reduce_fraction( int *n, int *d );
102 static inline uint8_t x264_clip_uint8( int x )
104 return x&(~255) ? (-x)>>31 : x;
107 static inline int x264_clip3( int v, int i_min, int i_max )
109 return ( (v < i_min) ? i_min : (v > i_max) ? i_max : v );
112 static inline double x264_clip3f( double v, double f_min, double f_max )
114 return ( (v < f_min) ? f_min : (v > f_max) ? f_max : v );
117 static inline int x264_median( int a, int b, int c )
123 max = b; /* no need to do 'b > max' (more consuming than always doing affectation) */
130 return a + b + c - min - max;
134 /****************************************************************************
136 ****************************************************************************/
146 static const char slice_type_to_char[] = { 'P', 'B', 'I', 'S', 'S' };
165 int i_idr_pic_id; /* -1 if nal_type != 5 */
168 int i_delta_poc_bottom;
171 int i_redundant_pic_cnt;
173 int b_direct_spatial_mv_pred;
175 int b_num_ref_idx_override;
176 int i_num_ref_idx_l0_active;
177 int i_num_ref_idx_l1_active;
179 int b_ref_pic_list_reordering_l0;
180 int b_ref_pic_list_reordering_l1;
184 } ref_pic_list_order[2][16];
186 int i_cabac_init_idc;
193 /* deblocking filter */
194 int i_disable_deblocking_filter_idc;
195 int i_alpha_c0_offset;
198 } x264_slice_header_t;
202 #define X264_SCAN8_SIZE (6*8)
203 #define X264_SCAN8_0 (4+1*8)
205 static const int x264_scan8[16+2*4] =
208 4+1*8, 5+1*8, 4+2*8, 5+2*8,
209 6+1*8, 7+1*8, 6+2*8, 7+2*8,
210 4+3*8, 5+3*8, 4+4*8, 5+4*8,
211 6+3*8, 7+3*8, 6+4*8, 7+4*8,
231 typedef struct x264_ratecontrol_t x264_ratecontrol_t;
235 /* encoder parameters */
238 x264_t *thread[X264_THREAD_MAX];
239 x264_pthread_t thread_handle;
241 int i_thread_phase; /* which thread to use for the next frame */
243 /* bitstream output */
247 x264_nal_t nal[X264_NAL_MAX];
248 int i_bitstream; /* size of p_bitstream */
249 uint8_t *p_bitstream; /* will hold data for all nal */
254 /* frame number/poc */
257 int i_frame_offset; /* decoding only */
258 int i_frame_num; /* decoding only */
259 int i_poc_msb; /* decoding only */
260 int i_poc_lsb; /* decoding only */
261 int i_poc; /* decoding only */
263 int i_thread_num; /* threads only */
264 int i_nal_type; /* threads only */
265 int i_nal_ref_idc; /* threads only */
267 /* We use only one SPS and one PPS */
268 x264_sps_t sps_array[1];
270 x264_pps_t pps_array[1];
274 /* quantization matrix for decoding, [cqm][qp%6][coef_y][coef_x] */
275 int (*dequant4_mf[4])[4][4]; /* [4][6][4][4] */
276 int (*dequant8_mf[2])[8][8]; /* [2][6][8][8] */
277 /* quantization matrix for trellis, [cqm][qp][coef] */
278 int (*unquant4_mf[4])[16]; /* [4][52][16] */
279 int (*unquant8_mf[2])[64]; /* [2][52][64] */
280 /* quantization matrix for deadzone */
281 uint16_t (*quant4_mf[4])[16]; /* [4][52][16] */
282 uint16_t (*quant8_mf[2])[64]; /* [2][52][64] */
283 uint16_t (*quant4_bias[4])[16]; /* [4][52][16] */
284 uint16_t (*quant8_bias[2])[64]; /* [2][52][64] */
286 uint32_t nr_residual_sum[2][64];
287 uint32_t nr_offset[2][64];
288 uint32_t nr_count[2];
291 x264_slice_header_t sh;
298 /* Frames to be encoded (whose types have been decided) */
299 x264_frame_t *current[X264_BFRAME_MAX+3];
300 /* Temporary buffer (frames types not yet decided) */
301 x264_frame_t *next[X264_BFRAME_MAX+3];
303 x264_frame_t *unused[X264_BFRAME_MAX + X264_THREAD_MAX*2 + 16+4];
304 /* For adaptive B decision */
305 x264_frame_t *last_nonb;
307 /* frames used for reference + sentinels */
308 x264_frame_t *reference[16+2];
310 int i_last_idr; /* Frame number of the last IDR */
312 int i_input; /* Number of input frames already accepted */
314 int i_max_dpb; /* Number of frames allocated in the decoded picture buffer */
317 int i_delay; /* Number of frames buffered for B reordering */
318 int b_have_lowres; /* Whether 1/2 resolution luma planes are being used */
321 /* current frame being encoded */
324 /* frame being reconstructed */
327 /* references lists */
329 x264_frame_t *fref0[16+3]; /* ref list 0 */
331 x264_frame_t *fref1[16+3]; /* ref list 1 */
332 int b_ref_reorder[2];
336 /* Current MB DCT coeffs */
339 DECLARE_ALIGNED_16( int16_t luma16x16_dc[16] );
340 DECLARE_ALIGNED_16( int16_t chroma_dc[2][4] );
341 // FIXME share memory?
342 DECLARE_ALIGNED_16( int16_t luma8x8[4][64] );
343 DECLARE_ALIGNED_16( int16_t luma4x4[16+8][16] );
346 /* MB table and cache for current frame/mb */
349 int i_mb_count; /* number of mbs in a frame */
363 /* Search parameters */
368 int b_noise_reduction;
372 /* Allowed qpel MV range to stay within the picture + emulated edge pixels */
375 /* Subpel MV range for motion search.
376 * same mv_min/max but includes levels' i_mv_range. */
379 /* Fullpel MV range for motion search */
383 /* neighboring MBs */
384 unsigned int i_neighbour;
385 unsigned int i_neighbour8[4]; /* neighbours of each 8x8 or 4x4 block that are available */
386 unsigned int i_neighbour4[16]; /* at the time the block is coded */
389 int i_mb_type_topleft;
390 int i_mb_type_topright;
395 int8_t *type; /* mb type */
396 int8_t *qp; /* mb qp */
397 int16_t *cbp; /* mb cbp: 0x0?: luma, 0x?0: chroma, 0x100: luma dc, 0x0200 and 0x0400: chroma dc (all set for PCM)*/
398 int8_t (*intra4x4_pred_mode)[7]; /* intra4x4 pred mode. for non I4x4 set to I_PRED_4x4_DC(2) */
399 uint8_t (*non_zero_count)[16+4+4]; /* nzc. for I_PCM set to 16 */
400 int8_t *chroma_pred_mode; /* chroma_pred_mode. cabac only. for non intra I_PRED_CHROMA_DC(0) */
401 int16_t (*mv[2])[2]; /* mb mv. set to 0 for intra mb */
402 int16_t (*mvd[2])[2]; /* mb mv difference with predict. set to 0 if intra. cabac only */
403 int8_t *ref[2]; /* mb ref. set to -1 if non used (intra or Lx only) */
404 int16_t (*mvr[2][32])[2]; /* 16x16 mv for each possible ref */
405 int8_t *skipbp; /* block pattern for SKIP or DIRECT (sub)mbs. B-frames + cabac only */
406 int8_t *mb_transform_size; /* transform_size_8x8_flag of each mb */
407 uint8_t *intra_border_backup[2][3]; /* bottom pixels of the previous mb row, used for intra prediction after the framebuffer has been deblocked */
408 uint8_t (*nnz_backup)[16]; /* when using cavlc + 8x8dct, the deblocker uses a modified nnz */
413 int i_sub_partition[4];
419 int i_intra16x16_pred_mode;
420 int i_chroma_pred_mode;
422 /* skip flags for i4x4 and i8x8
423 * 0 = encode as normal.
424 * 1 (non-RD only) = the DCT is still in h->dct, restore fdec and skip reconstruction.
425 * 2 (RD only) = the DCT has since been overwritten by RD; restore that too. */
430 /* space for p_fenc and p_fdec */
431 #define FENC_STRIDE 16
432 #define FDEC_STRIDE 32
433 DECLARE_ALIGNED_16( uint8_t fenc_buf[24*FENC_STRIDE] );
434 DECLARE_ALIGNED_16( uint8_t fdec_buf[27*FDEC_STRIDE] );
436 /* i4x4 and i8x8 backup data, for skipping the encode stage when possible */
437 DECLARE_ALIGNED_16( uint8_t i4x4_fdec_buf[16*16] );
438 DECLARE_ALIGNED_16( uint8_t i8x8_fdec_buf[16*16] );
439 DECLARE_ALIGNED_16( int16_t i8x8_dct_buf[3][64] );
440 DECLARE_ALIGNED_16( int16_t i4x4_dct_buf[15][16] );
442 /* pointer over mb of the frame to be compressed */
445 /* pointer over mb of the frame to be reconstructed */
448 /* pointer over mb of the references */
450 uint8_t *p_fref[2][32][4+2]; /* last: lN, lH, lV, lHV, cU, cV */
451 uint16_t *p_integral[2][16];
460 /* real intra4x4_pred_mode if I_4X4 or I_8X8, I_PRED_4x4_DC if mb available, -1 if not */
461 int8_t intra4x4_pred_mode[X264_SCAN8_SIZE];
463 /* i_non_zero_count if available else 0x80 */
464 uint8_t non_zero_count[X264_SCAN8_SIZE];
466 /* -1 if unused, -2 if unavailable */
467 DECLARE_ALIGNED_4( int8_t ref[2][X264_SCAN8_SIZE] );
469 /* 0 if not available */
470 DECLARE_ALIGNED_16( int16_t mv[2][X264_SCAN8_SIZE][2] );
471 DECLARE_ALIGNED_4( int16_t mvd[2][X264_SCAN8_SIZE][2] );
473 /* 1 if SKIP or DIRECT. set only for B-frames + CABAC */
474 DECLARE_ALIGNED_4( int8_t skip[X264_SCAN8_SIZE] );
476 DECLARE_ALIGNED_16( int16_t direct_mv[2][X264_SCAN8_SIZE][2] );
477 int8_t direct_ref[2][X264_SCAN8_SIZE];
480 /* number of neighbors (top and left) that used 8x8 dct */
481 int i_neighbour_transform_size;
482 int i_neighbour_interlaced;
486 int i_qp; /* current qp */
488 int i_last_qp; /* last qp */
489 int i_last_dqp; /* last delta qp */
490 int b_variable_qp; /* whether qp is allowed to vary per macroblock */
492 int b_direct_auto_read; /* take stats for --direct auto from the 2pass log */
493 int b_direct_auto_write; /* analyse direct modes, to use and/or save */
495 /* B_direct and weighted prediction */
496 int16_t dist_scale_factor[16][2];
497 int16_t bipred_weight[32][4];
498 /* maps fref1[0]'s ref indices into the current list0 */
499 int8_t map_col_to_list0_buf[2]; // for negative indices
500 int8_t map_col_to_list0[16];
503 /* rate control encoding only */
504 x264_ratecontrol_t *rc;
509 /* Current frame stats */
512 /* Headers bits (MV+Ref+MB Block Type */
514 /* Texture bits (Intra/Predicted) */
524 int i_mb_count_8x8dct[2];
525 int i_mb_count_size[7];
526 int i_mb_count_ref[32];
527 /* Estimated (SATD) cost as Intra/Predicted frame */
528 /* XXX: both omit the cost of MBs coded as P_SKIP */
532 /* Adaptive direct mv pred */
533 int i_direct_score[2];
536 /* Cumulated stats */
539 int i_slice_count[5];
540 int64_t i_slice_size[5];
541 double f_slice_qp[5];
543 int64_t i_sqe_global[5];
544 double f_psnr_average[5];
545 double f_psnr_mean_y[5];
546 double f_psnr_mean_u[5];
547 double f_psnr_mean_v[5];
548 double f_ssim_mean_y[5];
550 int64_t i_mb_count[5][19];
551 int64_t i_mb_count_8x8dct[2];
552 int64_t i_mb_count_size[2][7];
553 int64_t i_mb_count_ref[2][32];
555 int i_direct_score[2];
556 int i_direct_frames[2];
560 /* CPU functions dependents */
561 x264_predict_t predict_16x16[4+3];
562 x264_predict_t predict_8x8c[4+3];
563 x264_predict8x8_t predict_8x8[9+3];
564 x264_predict_t predict_4x4[9+3];
566 x264_pixel_function_t pixf;
567 x264_mc_functions_t mc;
568 x264_dct_function_t dctf;
569 x264_zigzag_function_t zigzagf;
570 x264_quant_function_t quantf;
571 x264_deblock_function_t loopf;
574 struct visualize_t *visualize;
578 // included at the end because it needs x264_t
579 #include "macroblock.h"