use lookup tables instead of actual exp/pow for AQ
[SFUResearch.git] / common / common.h
blobf2a0c54a9bc5f2c6b8659b693fdfd4dd3cb01eea
1 /*****************************************************************************
2 * common.h: h264 encoder
3 *****************************************************************************
4 * Copyright (C) 2003-2008 x264 project
6 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7 * Loren Merritt <lorenm@u.washington.edu>
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., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
22 *****************************************************************************/
24 #ifndef X264_COMMON_H
25 #define X264_COMMON_H
27 /****************************************************************************
28 * Macros
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) do{ type t = a; a = b; b = t; } while(0)
37 #define FIX8(f) ((int)(f*(1<<8)+.5))
39 #define CHECKED_MALLOC( var, size )\
41 var = x264_malloc( size );\
42 if( !var )\
44 x264_log( h, X264_LOG_ERROR, "malloc failed\n" );\
45 goto fail;\
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)
53 #define X264_PCM_COST (386*8)
55 // number of pixels (per thread) in progress at any given time.
56 // 16 for the macroblock in progress + 3 for deblocking + 3 for motion compensation filter + 2 for extra safety
57 #define X264_THREAD_HEIGHT 24
59 /****************************************************************************
60 * Includes
61 ****************************************************************************/
62 #include "osdep.h"
63 #include <stdarg.h>
64 #include <stddef.h>
65 #include <stdlib.h>
66 #include <string.h>
67 #include <assert.h>
68 #include "x264.h"
69 #include "bs.h"
70 #include "set.h"
71 #include "predict.h"
72 #include "pixel.h"
73 #include "mc.h"
74 #include "frame.h"
75 #include "dct.h"
76 #include "cabac.h"
77 #include "quant.h"
79 /****************************************************************************
80 * Generals functions
81 ****************************************************************************/
82 /* x264_malloc : will do or emulate a memalign
83 * you have to use x264_free for buffers allocated with x264_malloc */
84 void *x264_malloc( int );
85 void *x264_realloc( void *p, int i_size );
86 void x264_free( void * );
88 /* x264_slurp_file: malloc space for the whole file and read it */
89 char *x264_slurp_file( const char *filename );
91 /* mdate: return the current date in microsecond */
92 int64_t x264_mdate( void );
94 /* x264_param2string: return a (malloced) string containing most of
95 * the encoding options */
96 char *x264_param2string( x264_param_t *p, int b_res );
98 /* log */
99 void x264_log( x264_t *h, int i_level, const char *psz_fmt, ... );
101 void x264_reduce_fraction( int *n, int *d );
102 void x264_init_vlc_tables();
104 static inline uint8_t x264_clip_uint8( int x )
106 return x&(~255) ? (-x)>>31 : x;
109 static inline int x264_clip3( int v, int i_min, int i_max )
111 return ( (v < i_min) ? i_min : (v > i_max) ? i_max : v );
114 static inline double x264_clip3f( double v, double f_min, double f_max )
116 return ( (v < f_min) ? f_min : (v > f_max) ? f_max : v );
119 static inline int x264_median( int a, int b, int c )
121 int t = (a-b)&((a-b)>>31);
122 a -= t;
123 b += t;
124 b -= (b-c)&((b-c)>>31);
125 b += (a-b)&((a-b)>>31);
126 return b;
129 static inline void x264_median_mv( int16_t *dst, int16_t *a, int16_t *b, int16_t *c )
131 dst[0] = x264_median( a[0], b[0], c[0] );
132 dst[1] = x264_median( a[1], b[1], c[1] );
135 static inline int x264_predictor_difference( int16_t (*mvc)[2], intptr_t i_mvc )
137 int sum = 0, i;
138 for( i = 0; i < i_mvc-1; i++ )
140 sum += abs( mvc[i][0] - mvc[i+1][0] )
141 + abs( mvc[i][1] - mvc[i+1][1] );
143 return sum;
146 /****************************************************************************
148 ****************************************************************************/
149 enum slice_type_e
151 SLICE_TYPE_P = 0,
152 SLICE_TYPE_B = 1,
153 SLICE_TYPE_I = 2,
154 SLICE_TYPE_SP = 3,
155 SLICE_TYPE_SI = 4
158 static const char slice_type_to_char[] = { 'P', 'B', 'I', 'S', 'S' };
160 typedef struct
162 x264_sps_t *sps;
163 x264_pps_t *pps;
165 int i_type;
166 int i_first_mb;
167 int i_last_mb;
169 int i_pps_id;
171 int i_frame_num;
173 int b_mbaff;
174 int b_field_pic;
175 int b_bottom_field;
177 int i_idr_pic_id; /* -1 if nal_type != 5 */
179 int i_poc_lsb;
180 int i_delta_poc_bottom;
182 int i_delta_poc[2];
183 int i_redundant_pic_cnt;
185 int b_direct_spatial_mv_pred;
187 int b_num_ref_idx_override;
188 int i_num_ref_idx_l0_active;
189 int i_num_ref_idx_l1_active;
191 int b_ref_pic_list_reordering_l0;
192 int b_ref_pic_list_reordering_l1;
193 struct {
194 int idc;
195 int arg;
196 } ref_pic_list_order[2][16];
198 int i_cabac_init_idc;
200 int i_qp;
201 int i_qp_delta;
202 int b_sp_for_swidth;
203 int i_qs_delta;
205 /* deblocking filter */
206 int i_disable_deblocking_filter_idc;
207 int i_alpha_c0_offset;
208 int i_beta_offset;
210 } x264_slice_header_t;
212 /* From ffmpeg
214 #define X264_SCAN8_SIZE (6*8)
215 #define X264_SCAN8_0 (4+1*8)
217 static const int x264_scan8[16+2*4+3] =
219 /* Luma */
220 4+1*8, 5+1*8, 4+2*8, 5+2*8,
221 6+1*8, 7+1*8, 6+2*8, 7+2*8,
222 4+3*8, 5+3*8, 4+4*8, 5+4*8,
223 6+3*8, 7+3*8, 6+4*8, 7+4*8,
225 /* Cb */
226 1+1*8, 2+1*8,
227 1+2*8, 2+2*8,
229 /* Cr */
230 1+4*8, 2+4*8,
231 1+5*8, 2+5*8,
233 /* Luma DC */
234 4+5*8,
236 /* Chroma DC */
237 5+5*8, 6+5*8
240 0 1 2 3 4 5 6 7
242 1 B B L L L L
243 2 B B L L L L
244 3 L L L L
245 4 R R L L L L
246 5 R R DyDuDv
249 typedef struct x264_ratecontrol_t x264_ratecontrol_t;
251 struct x264_t
253 /* encoder parameters */
254 x264_param_t param;
256 x264_t *thread[X264_THREAD_MAX];
257 x264_pthread_t thread_handle;
258 int b_thread_active;
259 int i_thread_phase; /* which thread to use for the next frame */
261 /* bitstream output */
262 struct
264 int i_nal;
265 x264_nal_t nal[X264_NAL_MAX];
266 int i_bitstream; /* size of p_bitstream */
267 uint8_t *p_bitstream; /* will hold data for all nal */
268 bs_t bs;
269 int i_frame_size;
270 } out;
272 /**** thread synchronization starts here ****/
274 /* frame number/poc */
275 int i_frame;
277 int i_frame_offset; /* decoding only */
278 int i_frame_num; /* decoding only */
279 int i_poc_msb; /* decoding only */
280 int i_poc_lsb; /* decoding only */
281 int i_poc; /* decoding only */
283 int i_thread_num; /* threads only */
284 int i_nal_type; /* threads only */
285 int i_nal_ref_idc; /* threads only */
287 /* We use only one SPS and one PPS */
288 x264_sps_t sps_array[1];
289 x264_sps_t *sps;
290 x264_pps_t pps_array[1];
291 x264_pps_t *pps;
292 int i_idr_pic_id;
294 /* quantization matrix for decoding, [cqm][qp%6][coef_y][coef_x] */
295 int (*dequant4_mf[4])[4][4]; /* [4][6][4][4] */
296 int (*dequant8_mf[2])[8][8]; /* [2][6][8][8] */
297 /* quantization matrix for trellis, [cqm][qp][coef] */
298 int (*unquant4_mf[4])[16]; /* [4][52][16] */
299 int (*unquant8_mf[2])[64]; /* [2][52][64] */
300 /* quantization matrix for deadzone */
301 uint16_t (*quant4_mf[4])[16]; /* [4][52][16] */
302 uint16_t (*quant8_mf[2])[64]; /* [2][52][64] */
303 uint16_t (*quant4_bias[4])[16]; /* [4][52][16] */
304 uint16_t (*quant8_bias[2])[64]; /* [2][52][64] */
306 const uint8_t *chroma_qp_table; /* includes both the nonlinear luma->chroma mapping and chroma_qp_offset */
308 DECLARE_ALIGNED_16( uint32_t nr_residual_sum[2][64] );
309 DECLARE_ALIGNED_16( uint16_t nr_offset[2][64] );
310 uint32_t nr_count[2];
312 /* Slice header */
313 x264_slice_header_t sh;
315 /* cabac context */
316 x264_cabac_t cabac;
318 struct
320 /* Frames to be encoded (whose types have been decided) */
321 x264_frame_t *current[X264_BFRAME_MAX*4+3];
322 /* Temporary buffer (frames types not yet decided) */
323 x264_frame_t *next[X264_BFRAME_MAX*4+3];
324 /* Unused frames */
325 x264_frame_t *unused[X264_BFRAME_MAX*4 + X264_THREAD_MAX*2 + 16+4];
326 /* For adaptive B decision */
327 x264_frame_t *last_nonb;
329 /* frames used for reference + sentinels */
330 x264_frame_t *reference[16+2];
332 int i_last_idr; /* Frame number of the last IDR */
334 int i_input; /* Number of input frames already accepted */
336 int i_max_dpb; /* Number of frames allocated in the decoded picture buffer */
337 int i_max_ref0;
338 int i_max_ref1;
339 int i_delay; /* Number of frames buffered for B reordering */
340 int b_have_lowres; /* Whether 1/2 resolution luma planes are being used */
341 } frames;
343 /* current frame being encoded */
344 x264_frame_t *fenc;
346 /* frame being reconstructed */
347 x264_frame_t *fdec;
349 /* references lists */
350 int i_ref0;
351 x264_frame_t *fref0[16+3]; /* ref list 0 */
352 int i_ref1;
353 x264_frame_t *fref1[16+3]; /* ref list 1 */
354 int b_ref_reorder[2];
358 /* Current MB DCT coeffs */
359 struct
361 DECLARE_ALIGNED_16( int16_t luma16x16_dc[16] );
362 DECLARE_ALIGNED_16( int16_t chroma_dc[2][4] );
363 // FIXME share memory?
364 DECLARE_ALIGNED_16( int16_t luma8x8[4][64] );
365 DECLARE_ALIGNED_16( int16_t luma4x4[16+8][16] );
366 } dct;
368 /* MB table and cache for current frame/mb */
369 struct
371 int i_mb_count; /* number of mbs in a frame */
373 /* Strides */
374 int i_mb_stride;
375 int i_b8_stride;
376 int i_b4_stride;
378 /* Current index */
379 int i_mb_x;
380 int i_mb_y;
381 int i_mb_xy;
382 int i_b8_xy;
383 int i_b4_xy;
385 /* Search parameters */
386 int i_me_method;
387 int i_subpel_refine;
388 int b_chroma_me;
389 int b_trellis;
390 int b_noise_reduction;
391 int i_psy_rd; /* Psy RD strength--fixed point value*/
392 int i_psy_trellis; /* Psy trellis strength--fixed point value*/
394 int b_interlaced;
396 /* Allowed qpel MV range to stay within the picture + emulated edge pixels */
397 int mv_min[2];
398 int mv_max[2];
399 /* Subpel MV range for motion search.
400 * same mv_min/max but includes levels' i_mv_range. */
401 int mv_min_spel[2];
402 int mv_max_spel[2];
403 /* Fullpel MV range for motion search */
404 int mv_min_fpel[2];
405 int mv_max_fpel[2];
407 /* neighboring MBs */
408 unsigned int i_neighbour;
409 unsigned int i_neighbour8[4]; /* neighbours of each 8x8 or 4x4 block that are available */
410 unsigned int i_neighbour4[16]; /* at the time the block is coded */
411 int i_mb_type_top;
412 int i_mb_type_left;
413 int i_mb_type_topleft;
414 int i_mb_type_topright;
415 int i_mb_prev_xy;
416 int i_mb_top_xy;
418 /**** thread synchronization ends here ****/
419 /* subsequent variables are either thread-local or constant,
420 * and won't be copied from one thread to another */
422 /* mb table */
423 int8_t *type; /* mb type */
424 int8_t *qp; /* mb qp */
425 int16_t *cbp; /* mb cbp: 0x0?: luma, 0x?0: chroma, 0x100: luma dc, 0x0200 and 0x0400: chroma dc (all set for PCM)*/
426 int8_t (*intra4x4_pred_mode)[8]; /* intra4x4 pred mode. for non I4x4 set to I_PRED_4x4_DC(2) */
427 /* actually has only 7 entries; set to 8 for write-combining optimizations */
428 uint8_t (*non_zero_count)[16+4+4]; /* nzc. for I_PCM set to 16 */
429 int8_t *chroma_pred_mode; /* chroma_pred_mode. cabac only. for non intra I_PRED_CHROMA_DC(0) */
430 int16_t (*mv[2])[2]; /* mb mv. set to 0 for intra mb */
431 int16_t (*mvd[2])[2]; /* mb mv difference with predict. set to 0 if intra. cabac only */
432 int8_t *ref[2]; /* mb ref. set to -1 if non used (intra or Lx only) */
433 int16_t (*mvr[2][32])[2]; /* 16x16 mv for each possible ref */
434 int8_t *skipbp; /* block pattern for SKIP or DIRECT (sub)mbs. B-frames + cabac only */
435 int8_t *mb_transform_size; /* transform_size_8x8_flag of each mb */
436 uint8_t *intra_border_backup[2][3]; /* bottom pixels of the previous mb row, used for intra prediction after the framebuffer has been deblocked */
437 uint8_t (*nnz_backup)[16]; /* when using cavlc + 8x8dct, the deblocker uses a modified nnz */
439 /* current value */
440 int i_type;
441 int i_partition;
442 int i_sub_partition[4];
443 int b_transform_8x8;
445 int i_cbp_luma;
446 int i_cbp_chroma;
448 int i_intra16x16_pred_mode;
449 int i_chroma_pred_mode;
451 /* skip flags for i4x4 and i8x8
452 * 0 = encode as normal.
453 * 1 (non-RD only) = the DCT is still in h->dct, restore fdec and skip reconstruction.
454 * 2 (RD only) = the DCT has since been overwritten by RD; restore that too. */
455 int i_skip_intra;
456 /* skip flag for motion compensation */
457 /* if we've already done MC, we don't need to do it again */
458 int b_skip_mc;
460 struct
462 /* space for p_fenc and p_fdec */
463 #define FENC_STRIDE 16
464 #define FDEC_STRIDE 32
465 DECLARE_ALIGNED_16( uint8_t fenc_buf[24*FENC_STRIDE] );
466 DECLARE_ALIGNED_16( uint8_t fdec_buf[27*FDEC_STRIDE] );
468 /* i4x4 and i8x8 backup data, for skipping the encode stage when possible */
469 DECLARE_ALIGNED_16( uint8_t i4x4_fdec_buf[16*16] );
470 DECLARE_ALIGNED_16( uint8_t i8x8_fdec_buf[16*16] );
471 DECLARE_ALIGNED_16( int16_t i8x8_dct_buf[3][64] );
472 DECLARE_ALIGNED_16( int16_t i4x4_dct_buf[15][16] );
474 /* Psy trellis DCT data */
475 DECLARE_ALIGNED_16( int16_t fenc_dct8[4][64] );
476 DECLARE_ALIGNED_16( int16_t fenc_dct4[16][16] );
478 /* Psy RD SATD scores */
479 int fenc_satd[4][4];
480 int fenc_satd_sum;
481 int fenc_sa8d[2][2];
482 int fenc_sa8d_sum;
484 /* pointer over mb of the frame to be compressed */
485 uint8_t *p_fenc[3];
486 /* pointer to the actual source frame, not a block copy */
487 uint8_t *p_fenc_plane[3];
489 /* pointer over mb of the frame to be reconstructed */
490 uint8_t *p_fdec[3];
492 /* pointer over mb of the references */
493 int i_fref[2];
494 uint8_t *p_fref[2][32][4+2]; /* last: lN, lH, lV, lHV, cU, cV */
495 uint16_t *p_integral[2][16];
497 /* fref stride */
498 int i_stride[3];
499 } pic;
501 /* cache */
502 struct
504 /* real intra4x4_pred_mode if I_4X4 or I_8X8, I_PRED_4x4_DC if mb available, -1 if not */
505 int8_t intra4x4_pred_mode[X264_SCAN8_SIZE];
507 /* i_non_zero_count if available else 0x80 */
508 uint8_t non_zero_count[X264_SCAN8_SIZE];
510 /* -1 if unused, -2 if unavailable */
511 DECLARE_ALIGNED_4( int8_t ref[2][X264_SCAN8_SIZE] );
513 /* 0 if not available */
514 DECLARE_ALIGNED_16( int16_t mv[2][X264_SCAN8_SIZE][2] );
515 DECLARE_ALIGNED_8( int16_t mvd[2][X264_SCAN8_SIZE][2] );
517 /* 1 if SKIP or DIRECT. set only for B-frames + CABAC */
518 DECLARE_ALIGNED_4( int8_t skip[X264_SCAN8_SIZE] );
520 DECLARE_ALIGNED_16( int16_t direct_mv[2][X264_SCAN8_SIZE][2] );
521 DECLARE_ALIGNED_4( int8_t direct_ref[2][X264_SCAN8_SIZE] );
522 DECLARE_ALIGNED_4( int16_t pskip_mv[2] );
524 /* number of neighbors (top and left) that used 8x8 dct */
525 int i_neighbour_transform_size;
526 int i_neighbour_interlaced;
527 } cache;
529 /* */
530 int i_qp; /* current qp */
531 int i_chroma_qp;
532 int i_last_qp; /* last qp */
533 int i_last_dqp; /* last delta qp */
534 int b_variable_qp; /* whether qp is allowed to vary per macroblock */
535 int b_lossless;
536 int b_direct_auto_read; /* take stats for --direct auto from the 2pass log */
537 int b_direct_auto_write; /* analyse direct modes, to use and/or save */
539 /* B_direct and weighted prediction */
540 int16_t dist_scale_factor[16][2];
541 int16_t bipred_weight[32][4];
542 /* maps fref1[0]'s ref indices into the current list0 */
543 int8_t map_col_to_list0_buf[2]; // for negative indices
544 int8_t map_col_to_list0[16];
545 } mb;
547 /* rate control encoding only */
548 x264_ratecontrol_t *rc;
550 /* stats */
551 struct
553 /* Current frame stats */
554 struct
556 /* MV bits (MV+Ref+Block Type) */
557 int i_mv_bits;
558 /* Texture bits (DCT coefs) */
559 int i_tex_bits;
560 /* ? */
561 int i_misc_bits;
562 /* MB type counts */
563 int i_mb_count[19];
564 int i_mb_count_i;
565 int i_mb_count_p;
566 int i_mb_count_skip;
567 int i_mb_count_8x8dct[2];
568 int i_mb_count_ref[2][32];
569 int i_mb_partition[17];
570 /* Estimated (SATD) cost as Intra/Predicted frame */
571 /* XXX: both omit the cost of MBs coded as P_SKIP */
572 int i_intra_cost;
573 int i_inter_cost;
574 int i_mbs_analysed;
575 /* Adaptive direct mv pred */
576 int i_direct_score[2];
577 /* Metrics */
578 int64_t i_ssd[3];
579 double f_ssim;
580 } frame;
582 /* Cumulated stats */
584 /* per slice info */
585 int i_slice_count[5];
586 int64_t i_slice_size[5];
587 double f_slice_qp[5];
588 int i_consecutive_bframes[X264_BFRAME_MAX+1];
589 /* */
590 int64_t i_ssd_global[5];
591 double f_psnr_average[5];
592 double f_psnr_mean_y[5];
593 double f_psnr_mean_u[5];
594 double f_psnr_mean_v[5];
595 double f_ssim_mean_y[5];
596 /* */
597 int64_t i_mb_count[5][19];
598 int64_t i_mb_partition[2][17];
599 int64_t i_mb_count_8x8dct[2];
600 int64_t i_mb_count_ref[2][2][32];
601 /* */
602 int i_direct_score[2];
603 int i_direct_frames[2];
605 } stat;
607 /* CPU functions dependents */
608 x264_predict_t predict_16x16[4+3];
609 x264_predict_t predict_8x8c[4+3];
610 x264_predict8x8_t predict_8x8[9+3];
611 x264_predict_t predict_4x4[9+3];
613 x264_pixel_function_t pixf;
614 x264_mc_functions_t mc;
615 x264_dct_function_t dctf;
616 x264_zigzag_function_t zigzagf;
617 x264_quant_function_t quantf;
618 x264_deblock_function_t loopf;
620 #if VISUALIZE
621 struct visualize_t *visualize;
622 #endif
625 // included at the end because it needs x264_t
626 #include "macroblock.h"
628 #ifdef HAVE_MMX
629 #include "x86/util.h"
630 #endif
632 #endif