1 /*****************************************************************************
3 *****************************************************************************
4 * Copyright (C) 2003 Laurent Aimar
5 * $Id: encoder.c,v 1.1 2004/06/03 19:27:08 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 *****************************************************************************/
26 #include "common/common.h"
27 #include "common/cpu.h"
31 #include "ratecontrol.h"
32 #include "macroblock.h"
35 #include "common/visualize.h"
38 //#define DEBUG_MB_TYPE
39 //#define DEBUG_DUMP_FRAME
40 //#define DEBUG_BENCHMARK
42 #ifdef DEBUG_BENCHMARK
43 static int64_t i_mtime_encode_frame = 0;
44 static int64_t i_mtime_analyse = 0;
45 static int64_t i_mtime_encode = 0;
46 static int64_t i_mtime_write = 0;
47 static int64_t i_mtime_filter = 0;
48 #define TIMER_START( d ) \
50 int64_t d##start = x264_mdate();
52 #define TIMER_STOP( d ) \
53 d += x264_mdate() - d##start;\
56 #define TIMER_START( d )
57 #define TIMER_STOP( d )
60 #define NALU_OVERHEAD 5 // startcode + NAL type costs 5 bytes per frame
62 static void x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
63 x264_nal_t **pp_nal, int *pi_nal,
64 x264_picture_t *pic_out );
66 /****************************************************************************
68 ******************************* x264 libs **********************************
70 ****************************************************************************/
71 static float x264_psnr( int64_t i_sqe, int64_t i_size )
73 double f_mse = (double)i_sqe / ((double)65025.0 * (double)i_size);
74 if( f_mse <= 0.0000000001 ) /* Max 100dB */
77 return (float)(-10.0 * log( f_mse ) / log( 10.0 ));
80 #ifdef DEBUG_DUMP_FRAME
81 static void x264_frame_dump( x264_t *h, x264_frame_t *fr, char *name )
83 FILE *f = fopen( name, "r+b" );
88 /* Write the frame in display order */
89 fseek( f, fr->i_frame * h->param.i_height * h->param.i_width * 3 / 2, SEEK_SET );
91 for( i = 0; i < fr->i_plane; i++ )
93 for( y = 0; y < h->param.i_height / ( i == 0 ? 1 : 2 ); y++ )
95 fwrite( &fr->plane[i][y*fr->i_stride[i]], 1, h->param.i_width / ( i == 0 ? 1 : 2 ), f );
103 /* Fill "default" values */
104 static void x264_slice_header_init( x264_t *h, x264_slice_header_t *sh,
105 x264_sps_t *sps, x264_pps_t *pps,
106 int i_idr_pic_id, int i_frame, int i_qp )
108 x264_param_t *param = &h->param;
111 /* First we fill all field */
116 sh->i_last_mb = h->sps->i_mb_width * h->sps->i_mb_height;
117 sh->i_pps_id = pps->i_id;
119 sh->i_frame_num = i_frame;
121 sh->b_mbaff = h->param.b_interlaced;
122 sh->b_field_pic = 0; /* no field support for now */
123 sh->b_bottom_field = 0; /* not yet used */
125 sh->i_idr_pic_id = i_idr_pic_id;
127 /* poc stuff, fixed later */
129 sh->i_delta_poc_bottom = 0;
130 sh->i_delta_poc[0] = 0;
131 sh->i_delta_poc[1] = 0;
133 sh->i_redundant_pic_cnt = 0;
135 if( !h->mb.b_direct_auto_read )
137 if( h->mb.b_direct_auto_write )
138 sh->b_direct_spatial_mv_pred = ( h->stat.i_direct_score[1] > h->stat.i_direct_score[0] );
140 sh->b_direct_spatial_mv_pred = ( param->analyse.i_direct_mv_pred == X264_DIRECT_PRED_SPATIAL );
142 /* else b_direct_spatial_mv_pred was read from the 2pass statsfile */
144 sh->b_num_ref_idx_override = 0;
145 sh->i_num_ref_idx_l0_active = 1;
146 sh->i_num_ref_idx_l1_active = 1;
148 sh->b_ref_pic_list_reordering_l0 = h->b_ref_reorder[0];
149 sh->b_ref_pic_list_reordering_l1 = h->b_ref_reorder[1];
151 /* If the ref list isn't in the default order, construct reordering header */
152 /* List1 reordering isn't needed yet */
153 if( sh->b_ref_pic_list_reordering_l0 )
155 int pred_frame_num = i_frame;
156 for( i = 0; i < h->i_ref0; i++ )
158 int diff = h->fref0[i]->i_frame_num - pred_frame_num;
160 x264_log( h, X264_LOG_ERROR, "diff frame num == 0\n" );
161 sh->ref_pic_list_order[0][i].idc = ( diff > 0 );
162 sh->ref_pic_list_order[0][i].arg = abs( diff ) - 1;
163 pred_frame_num = h->fref0[i]->i_frame_num;
167 sh->i_cabac_init_idc = param->i_cabac_init_idc;
170 sh->i_qp_delta = i_qp - pps->i_pic_init_qp;
171 sh->b_sp_for_swidth = 0;
174 /* If effective qp <= 15, deblocking would have no effect anyway */
175 if( param->b_deblocking_filter
176 && ( h->mb.b_variable_qp
177 || 15 < i_qp + 2 * X264_MAX(param->i_deblocking_filter_alphac0, param->i_deblocking_filter_beta) ) )
179 sh->i_disable_deblocking_filter_idc = 0;
183 sh->i_disable_deblocking_filter_idc = 1;
185 sh->i_alpha_c0_offset = param->i_deblocking_filter_alphac0 << 1;
186 sh->i_beta_offset = param->i_deblocking_filter_beta << 1;
189 static void x264_slice_header_write( bs_t *s, x264_slice_header_t *sh, int i_nal_ref_idc )
195 assert( sh->i_first_mb % (2*sh->sps->i_mb_width) == 0 );
196 bs_write_ue( s, sh->i_first_mb >> 1 );
199 bs_write_ue( s, sh->i_first_mb );
201 bs_write_ue( s, sh->i_type + 5 ); /* same type things */
202 bs_write_ue( s, sh->i_pps_id );
203 bs_write( s, sh->sps->i_log2_max_frame_num, sh->i_frame_num );
205 if( !sh->sps->b_frame_mbs_only )
207 bs_write1( s, sh->b_field_pic );
208 if ( sh->b_field_pic )
209 bs_write1( s, sh->b_bottom_field );
212 if( sh->i_idr_pic_id >= 0 ) /* NAL IDR */
214 bs_write_ue( s, sh->i_idr_pic_id );
217 if( sh->sps->i_poc_type == 0 )
219 bs_write( s, sh->sps->i_log2_max_poc_lsb, sh->i_poc_lsb );
220 if( sh->pps->b_pic_order && !sh->b_field_pic )
222 bs_write_se( s, sh->i_delta_poc_bottom );
225 else if( sh->sps->i_poc_type == 1 && !sh->sps->b_delta_pic_order_always_zero )
227 bs_write_se( s, sh->i_delta_poc[0] );
228 if( sh->pps->b_pic_order && !sh->b_field_pic )
230 bs_write_se( s, sh->i_delta_poc[1] );
234 if( sh->pps->b_redundant_pic_cnt )
236 bs_write_ue( s, sh->i_redundant_pic_cnt );
239 if( sh->i_type == SLICE_TYPE_B )
241 bs_write1( s, sh->b_direct_spatial_mv_pred );
243 if( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_SP || sh->i_type == SLICE_TYPE_B )
245 bs_write1( s, sh->b_num_ref_idx_override );
246 if( sh->b_num_ref_idx_override )
248 bs_write_ue( s, sh->i_num_ref_idx_l0_active - 1 );
249 if( sh->i_type == SLICE_TYPE_B )
251 bs_write_ue( s, sh->i_num_ref_idx_l1_active - 1 );
256 /* ref pic list reordering */
257 if( sh->i_type != SLICE_TYPE_I )
259 bs_write1( s, sh->b_ref_pic_list_reordering_l0 );
260 if( sh->b_ref_pic_list_reordering_l0 )
262 for( i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
264 bs_write_ue( s, sh->ref_pic_list_order[0][i].idc );
265 bs_write_ue( s, sh->ref_pic_list_order[0][i].arg );
271 if( sh->i_type == SLICE_TYPE_B )
273 bs_write1( s, sh->b_ref_pic_list_reordering_l1 );
274 if( sh->b_ref_pic_list_reordering_l1 )
276 for( i = 0; i < sh->i_num_ref_idx_l1_active; i++ )
278 bs_write_ue( s, sh->ref_pic_list_order[1][i].idc );
279 bs_write_ue( s, sh->ref_pic_list_order[1][i].arg );
285 if( ( sh->pps->b_weighted_pred && ( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_SP ) ) ||
286 ( sh->pps->b_weighted_bipred == 1 && sh->i_type == SLICE_TYPE_B ) )
291 if( i_nal_ref_idc != 0 )
293 if( sh->i_idr_pic_id >= 0 )
295 bs_write1( s, 0 ); /* no output of prior pics flag */
296 bs_write1( s, 0 ); /* long term reference flag */
300 bs_write1( s, 0 ); /* adaptive_ref_pic_marking_mode_flag */
304 if( sh->pps->b_cabac && sh->i_type != SLICE_TYPE_I )
306 bs_write_ue( s, sh->i_cabac_init_idc );
308 bs_write_se( s, sh->i_qp_delta ); /* slice qp delta */
310 if( sh->pps->b_deblocking_filter_control )
312 bs_write_ue( s, sh->i_disable_deblocking_filter_idc );
313 if( sh->i_disable_deblocking_filter_idc != 1 )
315 bs_write_se( s, sh->i_alpha_c0_offset >> 1 );
316 bs_write_se( s, sh->i_beta_offset >> 1 );
321 /****************************************************************************
323 ****************************************************************************
324 ****************************** External API*********************************
325 ****************************************************************************
327 ****************************************************************************/
329 static int x264_validate_parameters( x264_t *h )
331 if( h->param.i_width <= 0 || h->param.i_height <= 0 )
333 x264_log( h, X264_LOG_ERROR, "invalid width x height (%dx%d)\n",
334 h->param.i_width, h->param.i_height );
338 if( h->param.i_width % 2 || h->param.i_height % 2 )
340 x264_log( h, X264_LOG_ERROR, "width or height not divisible by 2 (%dx%d)\n",
341 h->param.i_width, h->param.i_height );
344 if( h->param.i_csp != X264_CSP_I420 )
346 x264_log( h, X264_LOG_ERROR, "invalid CSP (only I420 supported)\n" );
350 if( h->param.i_threads == 0 )
351 h->param.i_threads = x264_cpu_num_processors() * 3/2;
352 h->param.i_threads = x264_clip3( h->param.i_threads, 1, X264_THREAD_MAX );
353 if( h->param.i_threads > 1 )
356 x264_log( h, X264_LOG_WARNING, "not compiled with pthread support!\n");
357 h->param.i_threads = 1;
359 if( h->param.i_scenecut_threshold >= 0 )
360 h->param.b_pre_scenecut = 1;
364 if( h->param.b_interlaced )
366 if( h->param.analyse.i_me_method >= X264_ME_ESA )
368 x264_log( h, X264_LOG_WARNING, "interlace + me=esa is not implemented\n" );
369 h->param.analyse.i_me_method = X264_ME_UMH;
371 if( h->param.analyse.i_direct_mv_pred > X264_DIRECT_PRED_SPATIAL )
373 x264_log( h, X264_LOG_WARNING, "interlace + direct=temporal is not implemented\n" );
374 h->param.analyse.i_direct_mv_pred = X264_DIRECT_PRED_SPATIAL;
378 if( h->param.rc.i_rc_method < 0 || h->param.rc.i_rc_method > 2 )
380 x264_log( h, X264_LOG_ERROR, "no ratecontrol method specified\n" );
383 h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, 0, 51 );
384 h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, 0, 51 );
385 if( h->param.rc.i_rc_method == X264_RC_CRF )
386 h->param.rc.i_qp_constant = h->param.rc.f_rf_constant;
387 if( (h->param.rc.i_rc_method == X264_RC_CQP || h->param.rc.i_rc_method == X264_RC_CRF)
388 && h->param.rc.i_qp_constant == 0 )
390 h->mb.b_lossless = 1;
391 h->param.i_cqm_preset = X264_CQM_FLAT;
392 h->param.psz_cqm_file = NULL;
393 h->param.rc.i_rc_method = X264_RC_CQP;
394 h->param.rc.f_ip_factor = 1;
395 h->param.rc.f_pb_factor = 1;
396 h->param.analyse.b_transform_8x8 = 0;
397 h->param.analyse.b_psnr = 0;
398 h->param.analyse.b_ssim = 0;
399 h->param.analyse.i_chroma_qp_offset = 0;
400 h->param.analyse.i_trellis = 0;
401 h->param.analyse.b_fast_pskip = 0;
402 h->param.analyse.i_noise_reduction = 0;
403 h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 1, 6 );
404 h->param.rc.i_aq_mode = 0;
406 if( h->param.rc.i_rc_method == X264_RC_CQP )
408 float qp_p = h->param.rc.i_qp_constant;
409 float qp_i = qp_p - 6*log(h->param.rc.f_ip_factor)/log(2);
410 float qp_b = qp_p + 6*log(h->param.rc.f_pb_factor)/log(2);
411 h->param.rc.i_qp_min = x264_clip3( (int)(X264_MIN3( qp_p, qp_i, qp_b )), 0, 51 );
412 h->param.rc.i_qp_max = x264_clip3( (int)(X264_MAX3( qp_p, qp_i, qp_b ) + .999), 0, 51 );
415 if( ( h->param.i_width % 16 || h->param.i_height % 16 )
416 && h->param.i_height != 1080 && !h->mb.b_lossless )
418 // There's nothing special about 1080 in that the warning still applies to it,
419 // but chances are the user can't help it if his content is already 1080p,
420 // so there's no point in warning in that case.
421 x264_log( h, X264_LOG_WARNING,
422 "width or height not divisible by 16 (%dx%d), compression will suffer.\n",
423 h->param.i_width, h->param.i_height );
426 h->param.i_frame_reference = x264_clip3( h->param.i_frame_reference, 1, 16 );
427 if( h->param.i_keyint_max <= 0 )
428 h->param.i_keyint_max = 1;
429 h->param.i_keyint_min = x264_clip3( h->param.i_keyint_min, 1, h->param.i_keyint_max/2+1 );
431 h->param.i_bframe = x264_clip3( h->param.i_bframe, 0, X264_BFRAME_MAX );
432 h->param.i_bframe_bias = x264_clip3( h->param.i_bframe_bias, -90, 100 );
433 h->param.b_bframe_pyramid = h->param.b_bframe_pyramid && h->param.i_bframe > 1;
434 h->param.b_bframe_adaptive = h->param.b_bframe_adaptive && h->param.i_bframe > 0;
435 h->param.analyse.b_weighted_bipred = h->param.analyse.b_weighted_bipred && h->param.i_bframe > 0;
436 h->mb.b_direct_auto_write = h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO
438 && ( h->param.rc.b_stat_write || !h->param.rc.b_stat_read );
439 if( h->param.i_scenecut_threshold < 0 )
440 h->param.b_pre_scenecut = 0;
442 h->param.i_deblocking_filter_alphac0 = x264_clip3( h->param.i_deblocking_filter_alphac0, -6, 6 );
443 h->param.i_deblocking_filter_beta = x264_clip3( h->param.i_deblocking_filter_beta, -6, 6 );
444 h->param.analyse.i_luma_deadzone[0] = x264_clip3( h->param.analyse.i_luma_deadzone[0], 0, 32 );
445 h->param.analyse.i_luma_deadzone[1] = x264_clip3( h->param.analyse.i_luma_deadzone[1], 0, 32 );
447 h->param.i_cabac_init_idc = x264_clip3( h->param.i_cabac_init_idc, 0, 2 );
449 if( h->param.i_cqm_preset < X264_CQM_FLAT || h->param.i_cqm_preset > X264_CQM_CUSTOM )
450 h->param.i_cqm_preset = X264_CQM_FLAT;
452 if( h->param.analyse.i_me_method < X264_ME_DIA ||
453 h->param.analyse.i_me_method > X264_ME_TESA )
454 h->param.analyse.i_me_method = X264_ME_HEX;
455 if( h->param.analyse.i_me_range < 4 )
456 h->param.analyse.i_me_range = 4;
457 if( h->param.analyse.i_me_range > 16 && h->param.analyse.i_me_method <= X264_ME_HEX )
458 h->param.analyse.i_me_range = 16;
459 if( h->param.analyse.i_me_method == X264_ME_TESA &&
460 (h->mb.b_lossless || h->param.analyse.i_subpel_refine <= 1) )
461 h->param.analyse.i_me_method = X264_ME_ESA;
462 h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 1, 7 );
463 h->param.analyse.b_bframe_rdo = h->param.analyse.b_bframe_rdo && h->param.analyse.i_subpel_refine >= 6;
464 h->param.analyse.b_mixed_references = h->param.analyse.b_mixed_references && h->param.i_frame_reference > 1;
465 h->param.analyse.inter &= X264_ANALYSE_PSUB16x16|X264_ANALYSE_PSUB8x8|X264_ANALYSE_BSUB16x16|
466 X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
467 h->param.analyse.intra &= X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
468 if( !(h->param.analyse.inter & X264_ANALYSE_PSUB16x16) )
469 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
470 if( !h->param.analyse.b_transform_8x8 )
472 h->param.analyse.inter &= ~X264_ANALYSE_I8x8;
473 h->param.analyse.intra &= ~X264_ANALYSE_I8x8;
475 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
476 if( !h->param.b_cabac )
477 h->param.analyse.i_trellis = 0;
478 h->param.analyse.i_trellis = x264_clip3( h->param.analyse.i_trellis, 0, 2 );
479 h->param.rc.i_aq_mode = x264_clip3( h->param.rc.i_aq_mode, 0, 2 );
480 if( h->param.rc.f_aq_strength <= 0 )
481 h->param.rc.i_aq_mode = 0;
482 /* VAQ effectively replaces qcomp, so qcomp is raised towards 1 to compensate. */
483 if( h->param.rc.i_aq_mode == X264_AQ_GLOBAL )
484 h->param.rc.f_qcompress = x264_clip3f(h->param.rc.f_qcompress + h->param.rc.f_aq_strength / 0.7, 0, 1);
485 h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
488 const x264_level_t *l = x264_levels;
489 while( l->level_idc != 0 && l->level_idc != h->param.i_level_idc )
491 if( l->level_idc == 0 )
493 x264_log( h, X264_LOG_ERROR, "invalid level_idc: %d\n", h->param.i_level_idc );
496 if( h->param.analyse.i_mv_range <= 0 )
497 h->param.analyse.i_mv_range = l->mv_range;
499 h->param.analyse.i_mv_range = x264_clip3(h->param.analyse.i_mv_range, 32, 512);
500 if( h->param.analyse.i_direct_8x8_inference < 0 )
501 h->param.analyse.i_direct_8x8_inference = l->direct8x8;
504 if( h->param.i_threads > 1 )
506 int r = h->param.analyse.i_mv_range_thread;
510 // half of the available space is reserved and divided evenly among the threads,
511 // the rest is allocated to whichever thread is far enough ahead to use it.
512 // reserving more space increases quality for some videos, but costs more time
513 // in thread synchronization.
514 int max_range = (h->param.i_height + X264_THREAD_HEIGHT) / h->param.i_threads - X264_THREAD_HEIGHT;
517 r = X264_MAX( r, h->param.analyse.i_me_range );
518 r = X264_MIN( r, h->param.analyse.i_mv_range );
519 // round up to use the whole mb row
520 r2 = (r & ~15) + ((-X264_THREAD_HEIGHT) & 15);
523 x264_log( h, X264_LOG_DEBUG, "using mv_range_thread = %d\n", r2 );
524 h->param.analyse.i_mv_range_thread = r2;
527 if( h->param.rc.f_qblur < 0 )
528 h->param.rc.f_qblur = 0;
529 if( h->param.rc.f_complexity_blur < 0 )
530 h->param.rc.f_complexity_blur = 0;
532 h->param.i_sps_id &= 31;
534 if( h->param.i_log_level < X264_LOG_INFO )
536 h->param.analyse.b_psnr = 0;
537 h->param.analyse.b_ssim = 0;
540 /* ensure the booleans are 0 or 1 so they can be used in math */
541 #define BOOLIFY(x) h->param.x = !!h->param.x
543 BOOLIFY( b_deblocking_filter );
544 BOOLIFY( b_interlaced );
545 BOOLIFY( analyse.b_transform_8x8 );
546 BOOLIFY( analyse.i_direct_8x8_inference );
547 BOOLIFY( analyse.b_bidir_me );
548 BOOLIFY( analyse.b_chroma_me );
549 BOOLIFY( analyse.b_fast_pskip );
550 BOOLIFY( rc.b_stat_write );
551 BOOLIFY( rc.b_stat_read );
557 static void mbcmp_init( x264_t *h )
559 int satd = !h->mb.b_lossless && h->param.analyse.i_subpel_refine > 1;
560 memcpy( h->pixf.mbcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.mbcmp) );
561 satd &= h->param.analyse.i_me_method == X264_ME_TESA;
562 memcpy( h->pixf.fpelcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.fpelcmp) );
563 memcpy( h->pixf.fpelcmp_x3, satd ? h->pixf.satd_x3 : h->pixf.sad_x3, sizeof(h->pixf.fpelcmp_x3) );
564 memcpy( h->pixf.fpelcmp_x4, satd ? h->pixf.satd_x4 : h->pixf.sad_x4, sizeof(h->pixf.fpelcmp_x4) );
567 /****************************************************************************
569 ****************************************************************************/
570 x264_t *x264_encoder_open ( x264_param_t *param )
572 x264_t *h = x264_malloc( sizeof( x264_t ) );
575 memset( h, 0, sizeof( x264_t ) );
577 /* Create a copy of param */
578 memcpy( &h->param, param, sizeof( x264_param_t ) );
580 if( x264_validate_parameters( h ) < 0 )
586 if( h->param.psz_cqm_file )
587 if( x264_cqm_parse_file( h, h->param.psz_cqm_file ) < 0 )
593 if( h->param.rc.psz_stat_out )
594 h->param.rc.psz_stat_out = strdup( h->param.rc.psz_stat_out );
595 if( h->param.rc.psz_stat_in )
596 h->param.rc.psz_stat_in = strdup( h->param.rc.psz_stat_in );
597 if( h->param.rc.psz_rc_eq )
598 h->param.rc.psz_rc_eq = strdup( h->param.rc.psz_rc_eq );
601 if( h->param.vui.i_sar_width > 0 && h->param.vui.i_sar_height > 0 )
603 int i_w = param->vui.i_sar_width;
604 int i_h = param->vui.i_sar_height;
606 x264_reduce_fraction( &i_w, &i_h );
608 while( i_w > 65535 || i_h > 65535 )
614 h->param.vui.i_sar_width = 0;
615 h->param.vui.i_sar_height = 0;
616 if( i_w == 0 || i_h == 0 )
618 x264_log( h, X264_LOG_WARNING, "cannot create valid sample aspect ratio\n" );
622 x264_log( h, X264_LOG_INFO, "using SAR=%d/%d\n", i_w, i_h );
623 h->param.vui.i_sar_width = i_w;
624 h->param.vui.i_sar_height = i_h;
628 x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den );
635 h->sps = &h->sps_array[0];
636 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
638 h->pps = &h->pps_array[0];
639 x264_pps_init( h->pps, h->param.i_sps_id, &h->param, h->sps);
641 x264_validate_levels( h );
643 if( x264_cqm_init( h ) < 0 )
649 h->mb.i_mb_count = h->sps->i_mb_width * h->sps->i_mb_height;
652 h->frames.i_delay = h->param.i_bframe + h->param.i_threads - 1;
653 h->frames.i_max_ref0 = h->param.i_frame_reference;
654 h->frames.i_max_ref1 = h->sps->vui.i_num_reorder_frames;
655 h->frames.i_max_dpb = h->sps->vui.i_max_dec_frame_buffering;
656 h->frames.b_have_lowres = !h->param.rc.b_stat_read
657 && ( h->param.rc.i_rc_method == X264_RC_ABR
658 || h->param.rc.i_rc_method == X264_RC_CRF
659 || h->param.b_bframe_adaptive
660 || h->param.b_pre_scenecut );
662 h->frames.i_last_idr = - h->param.i_keyint_max;
663 h->frames.i_input = 0;
664 h->frames.last_nonb = NULL;
671 /* init CPU functions */
672 x264_predict_16x16_init( h->param.cpu, h->predict_16x16 );
673 x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c );
674 x264_predict_8x8_init( h->param.cpu, h->predict_8x8 );
675 x264_predict_4x4_init( h->param.cpu, h->predict_4x4 );
677 x264_pixel_init( h->param.cpu, &h->pixf );
678 x264_dct_init( h->param.cpu, &h->dctf );
679 x264_zigzag_init( h->param.cpu, &h->zigzagf, h->param.b_interlaced );
680 x264_mc_init( h->param.cpu, &h->mc );
681 x264_quant_init( h, h->param.cpu, &h->quantf );
682 x264_deblock_init( h->param.cpu, &h->loopf );
683 x264_dct_init_weights();
687 x264_log( h, X264_LOG_INFO, "using cpu capabilities: %s%s%s%s%s%s%s%s%s%s\n",
688 param->cpu&X264_CPU_MMX ? "MMX " : "",
689 param->cpu&X264_CPU_MMXEXT ? "MMXEXT " : "",
690 param->cpu&X264_CPU_SSE ? "SSE " : "",
691 param->cpu&X264_CPU_SSE2 ? "SSE2 " : "",
692 param->cpu&X264_CPU_SSE3 ? "SSE3 " : "",
693 param->cpu&X264_CPU_SSSE3 ? "SSSE3 " : "",
694 param->cpu&X264_CPU_3DNOW ? "3DNow! " : "",
695 param->cpu&X264_CPU_ALTIVEC ? "Altivec " : "",
696 param->cpu&X264_CPU_CACHELINE_SPLIT ?
697 param->cpu&X264_CPU_CACHELINE_32 ? "Cache32 " :
698 param->cpu&X264_CPU_CACHELINE_64 ? "Cache64 " : "Cache? " : "",
699 param->cpu ? "" : "none!" );
702 h->out.i_bitstream = X264_MAX( 1000000, h->param.i_width * h->param.i_height * 4
703 * ( h->param.rc.i_rc_method == X264_RC_ABR ? pow( 0.95, h->param.rc.i_qp_min )
704 : pow( 0.95, h->param.rc.i_qp_constant ) * X264_MAX( 1, h->param.rc.f_ip_factor )));
708 for( i = 1; i < h->param.i_threads; i++ )
709 h->thread[i] = x264_malloc( sizeof(x264_t) );
711 for( i = 0; i < h->param.i_threads; i++ )
715 h->thread[i]->fdec = x264_frame_pop_unused( h );
716 h->thread[i]->out.p_bitstream = x264_malloc( h->out.i_bitstream );
717 if( x264_macroblock_cache_init( h->thread[i] ) < 0 )
721 if( x264_ratecontrol_new( h ) < 0 )
724 #ifdef DEBUG_DUMP_FRAME
726 /* create or truncate the reconstructed video file */
727 FILE *f = fopen( "fdec.yuv", "w" );
732 x264_log( h, X264_LOG_ERROR, "can't write to fdec.yuv\n" );
742 /****************************************************************************
743 * x264_encoder_reconfig:
744 ****************************************************************************/
745 int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
747 #define COPY(var) h->param.var = param->var
748 COPY( i_frame_reference ); // but never uses more refs than initially specified
749 COPY( i_bframe_bias );
750 if( h->param.i_scenecut_threshold >= 0 && param->i_scenecut_threshold >= 0 )
751 COPY( i_scenecut_threshold ); // can't turn it on or off, only vary the threshold
752 COPY( b_deblocking_filter );
753 COPY( i_deblocking_filter_alphac0 );
754 COPY( i_deblocking_filter_beta );
755 COPY( analyse.intra );
756 COPY( analyse.inter );
757 COPY( analyse.i_direct_mv_pred );
758 COPY( analyse.i_me_method );
759 COPY( analyse.i_me_range );
760 COPY( analyse.i_noise_reduction );
761 COPY( analyse.i_subpel_refine );
762 COPY( analyse.i_trellis );
763 COPY( analyse.b_bidir_me );
764 COPY( analyse.b_bframe_rdo );
765 COPY( analyse.b_chroma_me );
766 COPY( analyse.b_dct_decimate );
767 COPY( analyse.b_fast_pskip );
768 COPY( analyse.b_mixed_references );
769 // can only twiddle these if they were enabled to begin with:
770 if( h->pps->b_transform_8x8_mode )
771 COPY( analyse.b_transform_8x8 );
772 if( h->frames.i_max_ref1 > 1 )
773 COPY( b_bframe_pyramid );
778 return x264_validate_parameters( h );
782 static void x264_nal_start( x264_t *h, int i_type, int i_ref_idc )
784 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
786 nal->i_ref_idc = i_ref_idc;
787 nal->i_type = i_type;
790 nal->p_payload= &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
792 static void x264_nal_end( x264_t *h )
794 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
795 nal->i_payload = &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8] - nal->p_payload;
799 /****************************************************************************
800 * x264_encoder_headers:
801 ****************************************************************************/
802 int x264_encoder_headers( x264_t *h, x264_nal_t **pp_nal, int *pi_nal )
804 /* init bitstream context */
806 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
808 /* Put SPS and PPS */
809 if( h->i_frame == 0 )
811 /* identify ourself */
812 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
813 x264_sei_version_write( h, &h->out.bs );
816 /* generate sequence parameters */
817 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
818 x264_sps_write( &h->out.bs, h->sps );
821 /* generate picture parameters */
822 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
823 x264_pps_write( &h->out.bs, h->pps );
827 *pi_nal = h->out.i_nal;
828 *pp_nal = &h->out.nal[0];
834 static inline void x264_reference_build_list( x264_t *h, int i_poc )
839 /* build ref list 0/1 */
842 for( i = 0; h->frames.reference[i]; i++ )
844 if( h->frames.reference[i]->i_poc < i_poc )
846 h->fref0[h->i_ref0++] = h->frames.reference[i];
848 else if( h->frames.reference[i]->i_poc > i_poc )
850 h->fref1[h->i_ref1++] = h->frames.reference[i];
854 /* Order ref0 from higher to lower poc */
858 for( i = 0; i < h->i_ref0 - 1; i++ )
860 if( h->fref0[i]->i_poc < h->fref0[i+1]->i_poc )
862 XCHG( x264_frame_t*, h->fref0[i], h->fref0[i+1] );
868 /* Order ref1 from lower to higher poc (bubble sort) for B-frame */
872 for( i = 0; i < h->i_ref1 - 1; i++ )
874 if( h->fref1[i]->i_poc > h->fref1[i+1]->i_poc )
876 XCHG( x264_frame_t*, h->fref1[i], h->fref1[i+1] );
883 /* In the standard, a P-frame's ref list is sorted by frame_num.
884 * We use POC, but check whether explicit reordering is needed */
885 h->b_ref_reorder[0] =
886 h->b_ref_reorder[1] = 0;
887 if( h->sh.i_type == SLICE_TYPE_P )
889 for( i = 0; i < h->i_ref0 - 1; i++ )
890 if( h->fref0[i]->i_frame_num < h->fref0[i+1]->i_frame_num )
892 h->b_ref_reorder[0] = 1;
897 h->i_ref1 = X264_MIN( h->i_ref1, h->frames.i_max_ref1 );
898 h->i_ref0 = X264_MIN( h->i_ref0, h->frames.i_max_ref0 );
899 h->i_ref0 = X264_MIN( h->i_ref0, h->param.i_frame_reference ); // if reconfig() has lowered the limit
900 assert( h->i_ref0 + h->i_ref1 <= 16 );
901 h->mb.pic.i_fref[0] = h->i_ref0;
902 h->mb.pic.i_fref[1] = h->i_ref1;
905 static void x264_fdec_filter_row( x264_t *h, int mb_y )
907 /* mb_y is the mb to be encoded next, not the mb to be filtered here */
908 int b_hpel = h->fdec->b_kept_as_ref;
909 int b_deblock = !h->sh.i_disable_deblocking_filter_idc;
910 int b_end = mb_y == h->sps->i_mb_height;
911 int min_y = mb_y - (1 << h->sh.b_mbaff);
912 #ifndef DEBUG_DUMP_FRAME
915 if( mb_y & h->sh.b_mbaff )
923 for( j=0; j<=h->sh.b_mbaff; j++ )
926 memcpy( h->mb.intra_border_backup[j][i],
927 h->fdec->plane[i] + ((mb_y*16 >> !!i) + j - 1 - h->sh.b_mbaff) * h->fdec->i_stride[i],
928 h->sps->i_mb_width*16 >> !!i );
934 int max_y = b_end ? h->sps->i_mb_height : mb_y;
936 for( y = min_y; y < max_y; y += (1 << h->sh.b_mbaff) )
937 x264_frame_deblock_row( h, y );
942 x264_frame_expand_border( h, h->fdec, min_y, b_end );
943 x264_frame_filter( h, h->fdec, min_y, b_end );
944 x264_frame_expand_border_filtered( h, h->fdec, min_y, b_end );
947 if( h->param.i_threads > 1 && h->fdec->b_kept_as_ref )
949 x264_frame_cond_broadcast( h->fdec, mb_y*16 + (b_end ? 10000 : -(X264_THREAD_HEIGHT << h->sh.b_mbaff)) );
953 static inline void x264_reference_update( x264_t *h )
957 if( h->fdec->i_frame >= 0 )
960 if( !h->fdec->b_kept_as_ref )
962 if( h->param.i_threads > 1 )
964 x264_frame_push_unused( h, h->fdec );
965 h->fdec = x264_frame_pop_unused( h );
970 /* move lowres copy of the image to the ref frame */
971 for( i = 0; i < 4; i++)
973 XCHG( uint8_t*, h->fdec->lowres[i], h->fenc->lowres[i] );
974 XCHG( uint8_t*, h->fdec->buffer_lowres[i], h->fenc->buffer_lowres[i] );
977 /* adaptive B decision needs a pointer, since it can't use the ref lists */
978 if( h->sh.i_type != SLICE_TYPE_B )
979 h->frames.last_nonb = h->fdec;
981 /* move frame in the buffer */
982 x264_frame_push( h->frames.reference, h->fdec );
983 if( h->frames.reference[h->frames.i_max_dpb] )
984 x264_frame_push_unused( h, x264_frame_shift( h->frames.reference ) );
985 h->fdec = x264_frame_pop_unused( h );
988 static inline void x264_reference_reset( x264_t *h )
990 while( h->frames.reference[0] )
991 x264_frame_push_unused( h, x264_frame_pop( h->frames.reference ) );
996 static inline void x264_slice_init( x264_t *h, int i_nal_type, int i_global_qp )
998 /* ------------------------ Create slice header ----------------------- */
999 if( i_nal_type == NAL_SLICE_IDR )
1001 x264_slice_header_init( h, &h->sh, h->sps, h->pps, h->i_idr_pic_id, h->i_frame_num, i_global_qp );
1004 h->i_idr_pic_id = ( h->i_idr_pic_id + 1 ) % 65536;
1008 x264_slice_header_init( h, &h->sh, h->sps, h->pps, -1, h->i_frame_num, i_global_qp );
1010 /* always set the real higher num of ref frame used */
1011 h->sh.b_num_ref_idx_override = 1;
1012 h->sh.i_num_ref_idx_l0_active = h->i_ref0 <= 0 ? 1 : h->i_ref0;
1013 h->sh.i_num_ref_idx_l1_active = h->i_ref1 <= 0 ? 1 : h->i_ref1;
1016 h->fdec->i_frame_num = h->sh.i_frame_num;
1018 if( h->sps->i_poc_type == 0 )
1020 h->sh.i_poc_lsb = h->fdec->i_poc & ( (1 << h->sps->i_log2_max_poc_lsb) - 1 );
1021 h->sh.i_delta_poc_bottom = 0; /* XXX won't work for field */
1023 else if( h->sps->i_poc_type == 1 )
1025 /* FIXME TODO FIXME */
1029 /* Nothing to do ? */
1032 x264_macroblock_slice_init( h );
1035 static void x264_slice_write( x264_t *h )
1042 memset( &h->stat.frame, 0, sizeof(h->stat.frame) );
1045 x264_nal_start( h, h->i_nal_type, h->i_nal_ref_idc );
1048 x264_slice_header_write( &h->out.bs, &h->sh, h->i_nal_ref_idc );
1049 if( h->param.b_cabac )
1051 /* alignment needed */
1052 bs_align_1( &h->out.bs );
1055 x264_cabac_context_init( &h->cabac, h->sh.i_type, h->sh.i_qp, h->sh.i_cabac_init_idc );
1056 x264_cabac_encode_init ( &h->cabac, h->out.bs.p, h->out.bs.p_end );
1058 h->mb.i_last_qp = h->sh.i_qp;
1059 h->mb.i_last_dqp = 0;
1061 for( mb_xy = h->sh.i_first_mb, i_skip = 0; mb_xy < h->sh.i_last_mb; )
1063 const int i_mb_y = mb_xy / h->sps->i_mb_width;
1064 const int i_mb_x = mb_xy % h->sps->i_mb_width;
1065 int mb_spos = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
1068 x264_fdec_filter_row( h, i_mb_y );
1071 x264_macroblock_cache_load( h, i_mb_x, i_mb_y );
1073 /* analyse parameters
1074 * Slice I: choose I_4x4 or I_16x16 mode
1075 * Slice P: choose between using P mode or intra (4x4 or 16x16)
1077 TIMER_START( i_mtime_analyse );
1078 x264_macroblock_analyse( h );
1079 TIMER_STOP( i_mtime_analyse );
1081 /* encode this macroblock -> be careful it can change the mb type to P_SKIP if needed */
1082 TIMER_START( i_mtime_encode );
1083 x264_macroblock_encode( h );
1084 TIMER_STOP( i_mtime_encode );
1086 TIMER_START( i_mtime_write );
1087 if( h->param.b_cabac )
1089 if( mb_xy > h->sh.i_first_mb && !(h->sh.b_mbaff && (i_mb_y&1)) )
1090 x264_cabac_encode_terminal( &h->cabac );
1092 if( IS_SKIP( h->mb.i_type ) )
1093 x264_cabac_mb_skip( h, 1 );
1096 if( h->sh.i_type != SLICE_TYPE_I )
1097 x264_cabac_mb_skip( h, 0 );
1098 x264_macroblock_write_cabac( h, &h->cabac );
1103 if( IS_SKIP( h->mb.i_type ) )
1107 if( h->sh.i_type != SLICE_TYPE_I )
1109 bs_write_ue( &h->out.bs, i_skip ); /* skip run */
1112 x264_macroblock_write_cavlc( h, &h->out.bs );
1115 TIMER_STOP( i_mtime_write );
1118 if( h->param.b_visualize )
1119 x264_visualize_mb( h );
1123 x264_macroblock_cache_save( h );
1125 /* accumulate mb stats */
1126 h->stat.frame.i_mb_count[h->mb.i_type]++;
1127 if( !IS_SKIP(h->mb.i_type) && !IS_INTRA(h->mb.i_type) && !IS_DIRECT(h->mb.i_type) )
1129 if( h->mb.i_partition != D_8x8 )
1130 h->stat.frame.i_mb_count_size[ x264_mb_partition_pixel_table[ h->mb.i_partition ] ] += 4;
1132 for( i = 0; i < 4; i++ )
1133 h->stat.frame.i_mb_count_size[ x264_mb_partition_pixel_table[ h->mb.i_sub_partition[i] ] ] ++;
1134 if( h->param.i_frame_reference > 1 )
1136 for( i = 0; i < 4; i++ )
1138 int i_ref = h->mb.cache.ref[0][ x264_scan8[4*i] ];
1140 h->stat.frame.i_mb_count_ref[i_ref] ++;
1144 if( h->mb.i_cbp_luma && !IS_INTRA(h->mb.i_type) )
1146 h->stat.frame.i_mb_count_8x8dct[0] ++;
1147 h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
1150 x264_ratecontrol_mb( h, bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac) - mb_spos );
1154 if( (i_mb_y&1) && i_mb_x == h->sps->i_mb_width - 1 )
1157 mb_xy += 1 - h->sps->i_mb_width;
1159 mb_xy += h->sps->i_mb_width;
1165 if( h->param.b_cabac )
1167 x264_cabac_encode_flush( h, &h->cabac );
1168 h->out.bs.p = h->cabac.p;
1173 bs_write_ue( &h->out.bs, i_skip ); /* last skip run */
1174 /* rbsp_slice_trailing_bits */
1175 bs_rbsp_trailing( &h->out.bs );
1180 /* Compute misc bits */
1181 h->stat.frame.i_misc_bits = bs_pos( &h->out.bs )
1183 - h->stat.frame.i_itex_bits
1184 - h->stat.frame.i_ptex_bits
1185 - h->stat.frame.i_hdr_bits;
1188 static void x264_thread_sync_context( x264_t *dst, x264_t *src )
1194 // reference counting
1195 for( f = src->frames.reference; *f; f++ )
1196 (*f)->i_reference_count++;
1197 for( f = dst->frames.reference; *f; f++ )
1198 x264_frame_push_unused( src, *f );
1199 src->fdec->i_reference_count++;
1200 x264_frame_push_unused( src, dst->fdec );
1202 // copy everything except the per-thread pointers and the constants.
1203 memcpy( &dst->i_frame, &src->i_frame, offsetof(x264_t, mb.type) - offsetof(x264_t, i_frame) );
1204 memcpy( &dst->mb.i_type, &src->mb.i_type, offsetof(x264_t, rc) - offsetof(x264_t, mb.i_type) );
1205 dst->stat = src->stat;
1208 static void x264_thread_sync_stat( x264_t *dst, x264_t *src )
1212 memcpy( &dst->stat.i_slice_count, &src->stat.i_slice_count, sizeof(dst->stat) - sizeof(dst->stat.frame) );
1215 static int x264_slices_write( x264_t *h )
1220 if( h->param.b_visualize )
1221 x264_visualize_init( h );
1224 x264_stack_align( x264_slice_write, h );
1225 i_frame_size = h->out.nal[h->out.i_nal-1].i_payload;
1226 x264_fdec_filter_row( h, h->sps->i_mb_height );
1229 if( h->param.b_visualize )
1231 x264_visualize_show( h );
1232 x264_visualize_close( h );
1236 h->out.i_frame_size = i_frame_size;
1240 /****************************************************************************
1241 * x264_encoder_encode:
1242 * XXX: i_poc : is the poc of the current given picture
1243 * i_frame : is the number of the frame being coded
1244 * ex: type frame poc
1252 ****************************************************************************/
1253 int x264_encoder_encode( x264_t *h,
1254 x264_nal_t **pp_nal, int *pi_nal,
1255 x264_picture_t *pic_in,
1256 x264_picture_t *pic_out )
1258 x264_t *thread_current, *thread_prev, *thread_oldest;
1264 if( h->param.i_threads > 1)
1266 int i = ++h->i_thread_phase;
1267 int t = h->param.i_threads;
1268 thread_current = h->thread[ i%t ];
1269 thread_prev = h->thread[ (i-1)%t ];
1270 thread_oldest = h->thread[ (i+1)%t ];
1271 x264_thread_sync_context( thread_current, thread_prev );
1272 x264_thread_sync_ratecontrol( thread_current, thread_prev, thread_oldest );
1274 // fprintf(stderr, "current: %p prev: %p oldest: %p \n", thread_current, thread_prev, thread_oldest);
1283 // ok to call this before encoding any frames, since the initial values of fdec have b_kept_as_ref=0
1284 x264_reference_update( h );
1285 h->fdec->i_lines_completed = -1;
1291 /* ------------------- Setup new frame from picture -------------------- */
1292 TIMER_START( i_mtime_encode_frame );
1293 if( pic_in != NULL )
1295 /* 1: Copy the picture to a frame and move it to a buffer */
1296 x264_frame_t *fenc = x264_frame_pop_unused( h );
1298 if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 )
1301 if( h->param.i_width != 16 * h->sps->i_mb_width ||
1302 h->param.i_height != 16 * h->sps->i_mb_height )
1303 x264_frame_expand_border_mod16( h, fenc );
1305 fenc->i_frame = h->frames.i_input++;
1307 x264_frame_push( h->frames.next, fenc );
1309 if( h->frames.b_have_lowres )
1310 x264_frame_init_lowres( h, fenc );
1312 if( h->frames.i_input <= h->frames.i_delay + 1 - h->param.i_threads )
1314 /* Nothing yet to encode */
1315 /* waiting for filling bframe buffer */
1316 pic_out->i_type = X264_TYPE_AUTO;
1321 if( h->frames.current[0] == NULL )
1324 /* 2: Select frame types */
1325 if( h->frames.next[0] == NULL )
1327 x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
1331 x264_slicetype_decide( h );
1333 /* 3: move some B-frames and 1 non-B to encode queue */
1334 while( IS_X264_TYPE_B( h->frames.next[bframes]->i_type ) )
1336 x264_frame_push( h->frames.current, x264_frame_shift( &h->frames.next[bframes] ) );
1337 /* FIXME: when max B-frames > 3, BREF may no longer be centered after GOP closing */
1338 if( h->param.b_bframe_pyramid && bframes > 1 )
1340 x264_frame_t *mid = x264_frame_shift( &h->frames.next[bframes/2] );
1341 mid->i_type = X264_TYPE_BREF;
1342 x264_frame_push( h->frames.current, mid );
1346 x264_frame_push( h->frames.current, x264_frame_shift( h->frames.next ) );
1348 TIMER_STOP( i_mtime_encode_frame );
1350 /* ------------------- Get frame to be encoded ------------------------- */
1351 /* 4: get picture to encode */
1352 h->fenc = x264_frame_shift( h->frames.current );
1353 if( h->fenc == NULL )
1355 /* Nothing yet to encode (ex: waiting for I/P with B frames) */
1356 /* waiting for filling bframe buffer */
1357 pic_out->i_type = X264_TYPE_AUTO;
1363 if( h->fenc->i_type == X264_TYPE_IDR )
1365 h->frames.i_last_idr = h->fenc->i_frame;
1368 /* ------------------- Setup frame context ----------------------------- */
1369 /* 5: Init data dependent of frame type */
1370 TIMER_START( i_mtime_encode_frame );
1371 if( h->fenc->i_type == X264_TYPE_IDR )
1373 /* reset ref pictures */
1374 x264_reference_reset( h );
1376 i_nal_type = NAL_SLICE_IDR;
1377 i_nal_ref_idc = NAL_PRIORITY_HIGHEST;
1378 h->sh.i_type = SLICE_TYPE_I;
1380 else if( h->fenc->i_type == X264_TYPE_I )
1382 i_nal_type = NAL_SLICE;
1383 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
1384 h->sh.i_type = SLICE_TYPE_I;
1386 else if( h->fenc->i_type == X264_TYPE_P )
1388 i_nal_type = NAL_SLICE;
1389 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
1390 h->sh.i_type = SLICE_TYPE_P;
1392 else if( h->fenc->i_type == X264_TYPE_BREF )
1394 i_nal_type = NAL_SLICE;
1395 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* maybe add MMCO to forget it? -> low */
1396 h->sh.i_type = SLICE_TYPE_B;
1400 i_nal_type = NAL_SLICE;
1401 i_nal_ref_idc = NAL_PRIORITY_DISPOSABLE;
1402 h->sh.i_type = SLICE_TYPE_B;
1406 h->fenc->i_poc = 2 * (h->fenc->i_frame - h->frames.i_last_idr);
1407 h->fdec->i_type = h->fenc->i_type;
1408 h->fdec->i_frame = h->fenc->i_frame;
1409 h->fenc->b_kept_as_ref =
1410 h->fdec->b_kept_as_ref = i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE && h->param.i_keyint_max > 1;
1414 /* ------------------- Init ----------------------------- */
1415 /* build ref list 0/1 */
1416 x264_reference_build_list( h, h->fdec->i_poc );
1418 /* Init the rate control */
1419 x264_ratecontrol_start( h, h->fenc->i_qpplus1 );
1420 i_global_qp = x264_ratecontrol_qp( h );
1422 pic_out->i_qpplus1 =
1423 h->fdec->i_qpplus1 = i_global_qp + 1;
1425 if( h->sh.i_type == SLICE_TYPE_B )
1426 x264_macroblock_bipred_init( h );
1428 /* ------------------------ Create slice header ----------------------- */
1429 x264_slice_init( h, i_nal_type, i_global_qp );
1431 if( i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE )
1434 /* ---------------------- Write the bitstream -------------------------- */
1435 /* Init bitstream context */
1437 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
1442 if(h->sh.i_type == SLICE_TYPE_I)
1444 else if(h->sh.i_type == SLICE_TYPE_P)
1446 else if(h->sh.i_type == SLICE_TYPE_B)
1451 x264_nal_start(h, NAL_AUD, NAL_PRIORITY_DISPOSABLE);
1452 bs_write(&h->out.bs, 3, pic_type);
1453 bs_rbsp_trailing(&h->out.bs);
1457 h->i_nal_type = i_nal_type;
1458 h->i_nal_ref_idc = i_nal_ref_idc;
1460 /* Write SPS and PPS */
1461 if( i_nal_type == NAL_SLICE_IDR && h->param.b_repeat_headers )
1463 if( h->fenc->i_frame == 0 )
1465 /* identify ourself */
1466 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
1467 x264_sei_version_write( h, &h->out.bs );
1471 /* generate sequence parameters */
1472 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
1473 x264_sps_write( &h->out.bs, h->sps );
1476 /* generate picture parameters */
1477 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
1478 x264_pps_write( &h->out.bs, h->pps );
1483 if( h->param.i_threads > 1 )
1485 x264_pthread_create( &h->thread_handle, NULL, (void*)x264_slices_write, h );
1486 h->b_thread_active = 1;
1489 x264_slices_write( h );
1491 /* restore CPU state (before using float again) */
1492 x264_cpu_restore( h->param.cpu );
1494 if( h->sh.i_type == SLICE_TYPE_P && !h->param.rc.b_stat_read
1495 && h->param.i_scenecut_threshold >= 0
1496 && !h->param.b_pre_scenecut )
1498 const int *mbs = h->stat.frame.i_mb_count;
1499 int i_mb_i = mbs[I_16x16] + mbs[I_8x8] + mbs[I_4x4];
1500 int i_mb_p = mbs[P_L0] + mbs[P_8x8];
1501 int i_mb_s = mbs[P_SKIP];
1502 int i_mb = h->sps->i_mb_width * h->sps->i_mb_height;
1503 int64_t i_inter_cost = h->stat.frame.i_inter_cost;
1504 int64_t i_intra_cost = h->stat.frame.i_intra_cost;
1507 int i_gop_size = h->fenc->i_frame - h->frames.i_last_idr;
1508 float f_thresh_max = h->param.i_scenecut_threshold / 100.0;
1509 /* magic numbers pulled out of thin air */
1510 float f_thresh_min = f_thresh_max * h->param.i_keyint_min
1511 / ( h->param.i_keyint_max * 4 );
1512 if( h->param.i_keyint_min == h->param.i_keyint_max )
1513 f_thresh_min= f_thresh_max;
1515 /* macroblock_analyse() doesn't further analyse skipped mbs,
1516 * so we have to guess their cost */
1517 if( h->stat.frame.i_mbs_analysed > 0 )
1518 i_intra_cost = i_intra_cost * i_mb / h->stat.frame.i_mbs_analysed;
1520 if( i_gop_size < h->param.i_keyint_min / 4 )
1521 f_bias = f_thresh_min / 4;
1522 else if( i_gop_size <= h->param.i_keyint_min )
1523 f_bias = f_thresh_min * i_gop_size / h->param.i_keyint_min;
1526 f_bias = f_thresh_min
1527 + ( f_thresh_max - f_thresh_min )
1528 * ( i_gop_size - h->param.i_keyint_min )
1529 / ( h->param.i_keyint_max - h->param.i_keyint_min );
1531 f_bias = X264_MIN( f_bias, 1.0 );
1533 /* Bad P will be reencoded as I */
1534 if( h->stat.frame.i_mbs_analysed > 0 &&
1535 i_inter_cost >= (1.0 - f_bias) * i_intra_cost )
1539 x264_log( h, X264_LOG_DEBUG, "scene cut at %d Icost:%.0f Pcost:%.0f ratio:%.4f bias:%.4f gop:%d (imb:%d pmb:%d smb:%d)\n",
1541 (double)i_intra_cost, (double)i_inter_cost,
1542 1. - (double)i_inter_cost / i_intra_cost,
1544 i_mb_i, i_mb_p, i_mb_s );
1546 /* Restore frame num */
1549 for( b = 0; h->frames.current[b] && IS_X264_TYPE_B( h->frames.current[b]->i_type ); b++ );
1552 /* If using B-frames, force GOP to be closed.
1553 * Even if this frame is going to be I and not IDR, forcing a
1554 * P-frame before the scenecut will probably help compression.
1556 * We don't yet know exactly which frame is the scene cut, so
1557 * we can't assign an I-frame. Instead, change the previous
1558 * B-frame to P, and rearrange coding order. */
1560 if( h->param.b_bframe_adaptive || b > 1 )
1561 h->fenc->i_type = X264_TYPE_AUTO;
1562 x264_frame_sort_pts( h->frames.current );
1563 x264_frame_unshift( h->frames.next, h->fenc );
1564 h->fenc = h->frames.current[b-1];
1565 h->frames.current[b-1] = NULL;
1566 h->fenc->i_type = X264_TYPE_P;
1567 x264_frame_sort_dts( h->frames.current );
1569 /* Do IDR if needed */
1570 else if( i_gop_size >= h->param.i_keyint_min )
1575 /* Reinit field of fenc */
1576 h->fenc->i_type = X264_TYPE_IDR;
1579 /* Put enqueued frames back in the pool */
1580 while( h->frames.current[0] )
1581 x264_frame_push( h->frames.next, x264_frame_shift( h->frames.current ) );
1582 x264_frame_sort_pts( h->frames.next );
1586 h->fenc->i_type = X264_TYPE_I;
1592 x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
1596 static void x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
1597 x264_nal_t **pp_nal, int *pi_nal,
1598 x264_picture_t *pic_out )
1601 char psz_message[80];
1603 if( h->b_thread_active )
1605 x264_pthread_join( h->thread_handle, NULL );
1606 h->b_thread_active = 0;
1610 pic_out->i_type = X264_TYPE_AUTO;
1614 x264_frame_push_unused( thread_current, h->fenc );
1616 /* End bitstream, set output */
1617 *pi_nal = h->out.i_nal;
1618 *pp_nal = h->out.nal;
1621 /* Set output picture properties */
1622 if( h->sh.i_type == SLICE_TYPE_I )
1623 pic_out->i_type = h->i_nal_type == NAL_SLICE_IDR ? X264_TYPE_IDR : X264_TYPE_I;
1624 else if( h->sh.i_type == SLICE_TYPE_P )
1625 pic_out->i_type = X264_TYPE_P;
1627 pic_out->i_type = X264_TYPE_B;
1628 pic_out->i_pts = h->fenc->i_pts;
1630 pic_out->img.i_plane = h->fdec->i_plane;
1631 for(i = 0; i < 4; i++){
1632 pic_out->img.i_stride[i] = h->fdec->i_stride[i];
1633 pic_out->img.plane[i] = h->fdec->plane[i];
1636 /* ---------------------- Update encoder state ------------------------- */
1639 x264_cpu_restore( h->param.cpu );
1640 x264_ratecontrol_end( h, h->out.i_frame_size * 8 );
1642 /* restore CPU state (before using float again) */
1643 x264_cpu_restore( h->param.cpu );
1645 x264_noise_reduction_update( h );
1647 TIMER_STOP( i_mtime_encode_frame );
1649 /* ---------------------- Compute/Print statistics --------------------- */
1650 x264_thread_sync_stat( h, h->thread[0] );
1653 h->stat.i_slice_count[h->sh.i_type]++;
1654 h->stat.i_slice_size[h->sh.i_type] += h->out.i_frame_size + NALU_OVERHEAD;
1655 h->stat.f_slice_qp[h->sh.i_type] += h->fdec->f_qp_avg_aq;
1657 for( i = 0; i < X264_MBTYPE_MAX; i++ )
1658 h->stat.i_mb_count[h->sh.i_type][i] += h->stat.frame.i_mb_count[i];
1659 for( i = 0; i < 2; i++ )
1660 h->stat.i_mb_count_8x8dct[i] += h->stat.frame.i_mb_count_8x8dct[i];
1661 if( h->sh.i_type != SLICE_TYPE_I )
1663 for( i = 0; i < 7; i++ )
1664 h->stat.i_mb_count_size[h->sh.i_type][i] += h->stat.frame.i_mb_count_size[i];
1665 for( i = 0; i < 32; i++ )
1666 h->stat.i_mb_count_ref[h->sh.i_type][i] += h->stat.frame.i_mb_count_ref[i];
1668 if( h->sh.i_type == SLICE_TYPE_B )
1670 h->stat.i_direct_frames[ h->sh.b_direct_spatial_mv_pred ] ++;
1671 if( h->mb.b_direct_auto_write )
1673 //FIXME somewhat arbitrary time constants
1674 if( h->stat.i_direct_score[0] + h->stat.i_direct_score[1] > h->mb.i_mb_count )
1676 for( i = 0; i < 2; i++ )
1677 h->stat.i_direct_score[i] = h->stat.i_direct_score[i] * 9/10;
1679 for( i = 0; i < 2; i++ )
1680 h->stat.i_direct_score[i] += h->stat.frame.i_direct_score[i];
1684 psz_message[0] = '\0';
1685 if( h->param.analyse.b_psnr )
1689 for( i=0; i<3; i++ )
1691 sqe[i] = x264_pixel_ssd_wxh( &h->pixf,
1692 h->fdec->plane[i], h->fdec->i_stride[i],
1693 h->fenc->plane[i], h->fenc->i_stride[i],
1694 h->param.i_width >> !!i, h->param.i_height >> !!i );
1696 x264_cpu_restore( h->param.cpu );
1698 h->stat.i_sqe_global[h->sh.i_type] += sqe[0] + sqe[1] + sqe[2];
1699 h->stat.f_psnr_average[h->sh.i_type] += x264_psnr( sqe[0] + sqe[1] + sqe[2], 3 * h->param.i_width * h->param.i_height / 2 );
1700 h->stat.f_psnr_mean_y[h->sh.i_type] += x264_psnr( sqe[0], h->param.i_width * h->param.i_height );
1701 h->stat.f_psnr_mean_u[h->sh.i_type] += x264_psnr( sqe[1], h->param.i_width * h->param.i_height / 4 );
1702 h->stat.f_psnr_mean_v[h->sh.i_type] += x264_psnr( sqe[2], h->param.i_width * h->param.i_height / 4 );
1704 snprintf( psz_message, 80, " PSNR Y:%5.2f U:%5.2f V:%5.2f",
1705 x264_psnr( sqe[0], h->param.i_width * h->param.i_height ),
1706 x264_psnr( sqe[1], h->param.i_width * h->param.i_height / 4),
1707 x264_psnr( sqe[2], h->param.i_width * h->param.i_height / 4) );
1710 if( h->param.analyse.b_ssim )
1712 // offset by 2 pixels to avoid alignment of ssim blocks with dct blocks
1713 float ssim_y = x264_pixel_ssim_wxh( &h->pixf,
1714 h->fdec->plane[0] + 2+2*h->fdec->i_stride[0], h->fdec->i_stride[0],
1715 h->fenc->plane[0] + 2+2*h->fenc->i_stride[0], h->fenc->i_stride[0],
1716 h->param.i_width-2, h->param.i_height-2 );
1717 h->stat.f_ssim_mean_y[h->sh.i_type] += ssim_y;
1718 snprintf( psz_message + strlen(psz_message), 80 - strlen(psz_message),
1719 " SSIM Y:%.5f", ssim_y );
1721 psz_message[79] = '\0';
1723 x264_log( h, X264_LOG_DEBUG,
1724 "frame=%4d QP=%.2f NAL=%d Slice:%c Poc:%-3d I:%-4d P:%-4d SKIP:%-4d size=%d bytes%s\n",
1726 h->fdec->f_qp_avg_aq,
1728 h->sh.i_type == SLICE_TYPE_I ? 'I' : (h->sh.i_type == SLICE_TYPE_P ? 'P' : 'B' ),
1730 h->stat.frame.i_mb_count_i,
1731 h->stat.frame.i_mb_count_p,
1732 h->stat.frame.i_mb_count_skip,
1733 h->out.i_frame_size,
1736 // keep stats all in one place
1737 x264_thread_sync_stat( h->thread[0], h );
1738 // for the use of the next frame
1739 x264_thread_sync_stat( thread_current, h );
1741 #ifdef DEBUG_MB_TYPE
1743 static const char mb_chars[] = { 'i', 'i', 'I', 'C', 'P', '8', 'S',
1744 'D', '<', 'X', 'B', 'X', '>', 'B', 'B', 'B', 'B', '8', 'S' };
1746 for( mb_xy = 0; mb_xy < h->sps->i_mb_width * h->sps->i_mb_height; mb_xy++ )
1748 if( h->mb.type[mb_xy] < X264_MBTYPE_MAX && h->mb.type[mb_xy] >= 0 )
1749 fprintf( stderr, "%c ", mb_chars[ h->mb.type[mb_xy] ] );
1751 fprintf( stderr, "? " );
1753 if( (mb_xy+1) % h->sps->i_mb_width == 0 )
1754 fprintf( stderr, "\n" );
1759 #ifdef DEBUG_DUMP_FRAME
1760 /* Dump reconstructed frame */
1761 x264_frame_dump( h, h->fdec, "fdec.yuv" );
1765 /****************************************************************************
1766 * x264_encoder_close:
1767 ****************************************************************************/
1768 void x264_encoder_close ( x264_t *h )
1770 #ifdef DEBUG_BENCHMARK
1771 int64_t i_mtime_total = i_mtime_analyse + i_mtime_encode + i_mtime_write + i_mtime_filter + 1;
1773 int64_t i_yuv_size = 3 * h->param.i_width * h->param.i_height / 2;
1776 for( i=0; i<h->param.i_threads; i++ )
1778 // don't strictly have to wait for the other threads, but it's simpler than cancelling them
1779 if( h->thread[i]->b_thread_active )
1780 x264_pthread_join( h->thread[i]->thread_handle, NULL );
1783 #ifdef DEBUG_BENCHMARK
1784 x264_log( h, X264_LOG_INFO,
1785 "analyse=%d(%lldms) encode=%d(%lldms) write=%d(%lldms) filter=%d(%lldms)\n",
1786 (int)(100*i_mtime_analyse/i_mtime_total), i_mtime_analyse/1000,
1787 (int)(100*i_mtime_encode/i_mtime_total), i_mtime_encode/1000,
1788 (int)(100*i_mtime_write/i_mtime_total), i_mtime_write/1000,
1789 (int)(100*i_mtime_filter/i_mtime_total), i_mtime_filter/1000 );
1792 /* Slices used and PSNR */
1793 for( i=0; i<5; i++ )
1795 static const int slice_order[] = { SLICE_TYPE_I, SLICE_TYPE_SI, SLICE_TYPE_P, SLICE_TYPE_SP, SLICE_TYPE_B };
1796 static const char *slice_name[] = { "P", "B", "I", "SP", "SI" };
1797 int i_slice = slice_order[i];
1799 if( h->stat.i_slice_count[i_slice] > 0 )
1801 const int i_count = h->stat.i_slice_count[i_slice];
1802 if( h->param.analyse.b_psnr )
1804 x264_log( h, X264_LOG_INFO,
1805 "slice %s:%-5d Avg QP:%5.2f size:%6.0f PSNR Mean Y:%5.2f U:%5.2f V:%5.2f Avg:%5.2f Global:%5.2f\n",
1806 slice_name[i_slice],
1808 h->stat.f_slice_qp[i_slice] / i_count,
1809 (double)h->stat.i_slice_size[i_slice] / i_count,
1810 h->stat.f_psnr_mean_y[i_slice] / i_count, h->stat.f_psnr_mean_u[i_slice] / i_count, h->stat.f_psnr_mean_v[i_slice] / i_count,
1811 h->stat.f_psnr_average[i_slice] / i_count,
1812 x264_psnr( h->stat.i_sqe_global[i_slice], i_count * i_yuv_size ) );
1816 x264_log( h, X264_LOG_INFO,
1817 "slice %s:%-5d Avg QP:%5.2f size:%6.0f\n",
1818 slice_name[i_slice],
1820 h->stat.f_slice_qp[i_slice] / i_count,
1821 (double)h->stat.i_slice_size[i_slice] / i_count );
1827 if( h->stat.i_slice_count[SLICE_TYPE_I] > 0 )
1829 const int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_I];
1830 const double i_count = h->stat.i_slice_count[SLICE_TYPE_I] * h->mb.i_mb_count / 100.0;
1831 x264_log( h, X264_LOG_INFO,
1832 "mb I I16..4: %4.1f%% %4.1f%% %4.1f%%\n",
1833 i_mb_count[I_16x16]/ i_count,
1834 i_mb_count[I_8x8] / i_count,
1835 i_mb_count[I_4x4] / i_count );
1837 if( h->stat.i_slice_count[SLICE_TYPE_P] > 0 )
1839 const int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_P];
1840 const int64_t *i_mb_size = h->stat.i_mb_count_size[SLICE_TYPE_P];
1841 const double i_count = h->stat.i_slice_count[SLICE_TYPE_P] * h->mb.i_mb_count / 100.0;
1842 x264_log( h, X264_LOG_INFO,
1843 "mb P I16..4: %4.1f%% %4.1f%% %4.1f%% P16..4: %4.1f%% %4.1f%% %4.1f%% %4.1f%% %4.1f%% skip:%4.1f%%\n",
1844 i_mb_count[I_16x16]/ i_count,
1845 i_mb_count[I_8x8] / i_count,
1846 i_mb_count[I_4x4] / i_count,
1847 i_mb_size[PIXEL_16x16] / (i_count*4),
1848 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
1849 i_mb_size[PIXEL_8x8] / (i_count*4),
1850 (i_mb_size[PIXEL_8x4] + i_mb_size[PIXEL_4x8]) / (i_count*4),
1851 i_mb_size[PIXEL_4x4] / (i_count*4),
1852 i_mb_count[P_SKIP] / i_count );
1854 if( h->stat.i_slice_count[SLICE_TYPE_B] > 0 )
1856 const int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_B];
1857 const int64_t *i_mb_size = h->stat.i_mb_count_size[SLICE_TYPE_B];
1858 const double i_count = h->stat.i_slice_count[SLICE_TYPE_B] * h->mb.i_mb_count / 100.0;
1859 x264_log( h, X264_LOG_INFO,
1860 "mb B I16..4: %4.1f%% %4.1f%% %4.1f%% B16..8: %4.1f%% %4.1f%% %4.1f%% direct:%4.1f%% skip:%4.1f%%\n",
1861 i_mb_count[I_16x16] / i_count,
1862 i_mb_count[I_8x8] / i_count,
1863 i_mb_count[I_4x4] / i_count,
1864 i_mb_size[PIXEL_16x16] / (i_count*4),
1865 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
1866 i_mb_size[PIXEL_8x8] / (i_count*4),
1867 i_mb_count[B_DIRECT] / i_count,
1868 i_mb_count[B_SKIP] / i_count );
1871 x264_ratecontrol_summary( h );
1873 if( h->stat.i_slice_count[SLICE_TYPE_I] + h->stat.i_slice_count[SLICE_TYPE_P] + h->stat.i_slice_count[SLICE_TYPE_B] > 0 )
1875 const int i_count = h->stat.i_slice_count[SLICE_TYPE_I] +
1876 h->stat.i_slice_count[SLICE_TYPE_P] +
1877 h->stat.i_slice_count[SLICE_TYPE_B];
1878 float fps = (float) h->param.i_fps_num / h->param.i_fps_den;
1879 #define SUM3(p) (p[SLICE_TYPE_I] + p[SLICE_TYPE_P] + p[SLICE_TYPE_B])
1880 #define SUM3b(p,o) (p[SLICE_TYPE_I][o] + p[SLICE_TYPE_P][o] + p[SLICE_TYPE_B][o])
1881 float f_bitrate = fps * SUM3(h->stat.i_slice_size) / i_count / 125;
1883 if( h->pps->b_transform_8x8_mode )
1885 int64_t i_i8x8 = SUM3b( h->stat.i_mb_count, I_8x8 );
1886 int64_t i_intra = i_i8x8 + SUM3b( h->stat.i_mb_count, I_4x4 )
1887 + SUM3b( h->stat.i_mb_count, I_16x16 );
1888 x264_log( h, X264_LOG_INFO, "8x8 transform intra:%.1f%% inter:%.1f%%\n",
1889 100. * i_i8x8 / i_intra,
1890 100. * h->stat.i_mb_count_8x8dct[1] / h->stat.i_mb_count_8x8dct[0] );
1893 if( h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO
1894 && h->stat.i_slice_count[SLICE_TYPE_B] )
1896 x264_log( h, X264_LOG_INFO, "direct mvs spatial:%.1f%% temporal:%.1f%%\n",
1897 h->stat.i_direct_frames[1] * 100. / h->stat.i_slice_count[SLICE_TYPE_B],
1898 h->stat.i_direct_frames[0] * 100. / h->stat.i_slice_count[SLICE_TYPE_B] );
1901 if( h->frames.i_max_ref0 > 1 )
1904 for( i_slice = 0; i_slice < 2; i_slice++ )
1910 for( i = 0; i < h->frames.i_max_ref0 << h->param.b_interlaced; i++ )
1911 if( h->stat.i_mb_count_ref[i_slice][i] )
1913 i_den += h->stat.i_mb_count_ref[i_slice][i];
1918 for( i = 0; i <= i_max; i++ )
1919 p += sprintf( p, " %4.1f%%", 100. * h->stat.i_mb_count_ref[i_slice][i] / i_den );
1920 x264_log( h, X264_LOG_INFO, "ref %c %s\n", i_slice==SLICE_TYPE_P ? 'P' : 'B', buf );
1924 if( h->param.analyse.b_ssim )
1926 x264_log( h, X264_LOG_INFO,
1927 "SSIM Mean Y:%.7f\n",
1928 SUM3( h->stat.f_ssim_mean_y ) / i_count );
1930 if( h->param.analyse.b_psnr )
1932 x264_log( h, X264_LOG_INFO,
1933 "PSNR Mean Y:%6.3f U:%6.3f V:%6.3f Avg:%6.3f Global:%6.3f kb/s:%.2f\n",
1934 SUM3( h->stat.f_psnr_mean_y ) / i_count,
1935 SUM3( h->stat.f_psnr_mean_u ) / i_count,
1936 SUM3( h->stat.f_psnr_mean_v ) / i_count,
1937 SUM3( h->stat.f_psnr_average ) / i_count,
1938 x264_psnr( SUM3( h->stat.i_sqe_global ), i_count * i_yuv_size ),
1942 x264_log( h, X264_LOG_INFO, "kb/s:%.1f\n", f_bitrate );
1946 for( i = 0; h->frames.current[i]; i++ )
1947 x264_frame_delete( h->frames.current[i] );
1948 for( i = 0; h->frames.next[i]; i++ )
1949 x264_frame_delete( h->frames.next[i] );
1950 for( i = 0; h->frames.unused[i]; i++ )
1951 x264_frame_delete( h->frames.unused[i] );
1952 for( i = 0; h->frames.reference[i]; i++ )
1953 x264_frame_delete( h->frames.reference[i] );
1956 x264_ratecontrol_delete( h );
1959 if( h->param.rc.psz_stat_out )
1960 free( h->param.rc.psz_stat_out );
1961 if( h->param.rc.psz_stat_in )
1962 free( h->param.rc.psz_stat_in );
1963 if( h->param.rc.psz_rc_eq )
1964 free( h->param.rc.psz_rc_eq );
1966 x264_cqm_delete( h );
1967 for( i = h->param.i_threads - 1; i >= 0; i-- )
1969 x264_macroblock_cache_end( h->thread[i] );
1970 x264_free( h->thread[i]->out.p_bitstream );
1971 x264_free( h->thread[i] );