1 /*****************************************************************************
3 *****************************************************************************
4 * Copyright (C) 2003-2008 x264 project
6 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7 * Loren Merritt <lorenm@u.washington.edu>
8 * Fiona Glaser <fiona@x264.com>
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License as published by
12 * the Free Software Foundation; either version 2 of the License, or
13 * (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
23 *****************************************************************************/
27 #include "common/common.h"
28 #include "common/cpu.h"
32 #include "ratecontrol.h"
33 #include "macroblock.h"
36 #include "common/visualize.h"
39 //#define DEBUG_MB_TYPE
41 #define NALU_OVERHEAD 5 // startcode + NAL type costs 5 bytes per frame
43 #define bs_write_ue bs_write_ue_big
45 static int x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
46 x264_nal_t **pp_nal, int *pi_nal,
47 x264_picture_t *pic_out );
49 /****************************************************************************
51 ******************************* x264 libs **********************************
53 ****************************************************************************/
54 static float x264_psnr( int64_t i_sqe, int64_t i_size )
56 double f_mse = (double)i_sqe / ((double)65025.0 * (double)i_size);
57 if( f_mse <= 0.0000000001 ) /* Max 100dB */
60 return (float)(-10.0 * log( f_mse ) / log( 10.0 ));
63 static void x264_frame_dump( x264_t *h )
65 FILE *f = fopen( h->param.psz_dump_yuv, "r+b" );
69 /* Write the frame in display order */
70 fseek( f, h->fdec->i_frame * h->param.i_height * h->param.i_width * 3/2, SEEK_SET );
71 for( i = 0; i < h->fdec->i_plane; i++ )
72 for( y = 0; y < h->param.i_height >> !!i; y++ )
73 fwrite( &h->fdec->plane[i][y*h->fdec->i_stride[i]], 1, h->param.i_width >> !!i, f );
78 /* Fill "default" values */
79 static void x264_slice_header_init( x264_t *h, x264_slice_header_t *sh,
80 x264_sps_t *sps, x264_pps_t *pps,
81 int i_idr_pic_id, int i_frame, int i_qp )
83 x264_param_t *param = &h->param;
86 /* First we fill all field */
91 sh->i_last_mb = h->mb.i_mb_count - 1;
92 sh->i_pps_id = pps->i_id;
94 sh->i_frame_num = i_frame;
96 sh->b_mbaff = h->param.b_interlaced;
97 sh->b_field_pic = 0; /* no field support for now */
98 sh->b_bottom_field = 0; /* not yet used */
100 sh->i_idr_pic_id = i_idr_pic_id;
102 /* poc stuff, fixed later */
104 sh->i_delta_poc_bottom = 0;
105 sh->i_delta_poc[0] = 0;
106 sh->i_delta_poc[1] = 0;
108 sh->i_redundant_pic_cnt = 0;
110 if( !h->mb.b_direct_auto_read )
112 if( h->mb.b_direct_auto_write )
113 sh->b_direct_spatial_mv_pred = ( h->stat.i_direct_score[1] > h->stat.i_direct_score[0] );
115 sh->b_direct_spatial_mv_pred = ( param->analyse.i_direct_mv_pred == X264_DIRECT_PRED_SPATIAL );
117 /* else b_direct_spatial_mv_pred was read from the 2pass statsfile */
119 sh->b_num_ref_idx_override = 0;
120 sh->i_num_ref_idx_l0_active = 1;
121 sh->i_num_ref_idx_l1_active = 1;
123 sh->b_ref_pic_list_reordering_l0 = h->b_ref_reorder[0];
124 sh->b_ref_pic_list_reordering_l1 = h->b_ref_reorder[1];
126 /* If the ref list isn't in the default order, construct reordering header */
127 /* List1 reordering isn't needed yet */
128 if( sh->b_ref_pic_list_reordering_l0 )
130 int pred_frame_num = i_frame;
131 for( i = 0; i < h->i_ref0; i++ )
133 int diff = h->fref0[i]->i_frame_num - pred_frame_num;
135 x264_log( h, X264_LOG_ERROR, "diff frame num == 0\n" );
136 sh->ref_pic_list_order[0][i].idc = ( diff > 0 );
137 sh->ref_pic_list_order[0][i].arg = abs( diff ) - 1;
138 pred_frame_num = h->fref0[i]->i_frame_num;
142 sh->i_cabac_init_idc = param->i_cabac_init_idc;
145 sh->i_qp_delta = i_qp - pps->i_pic_init_qp;
146 sh->b_sp_for_swidth = 0;
149 /* If effective qp <= 15, deblocking would have no effect anyway */
150 if( param->b_deblocking_filter
151 && ( h->mb.b_variable_qp
152 || 15 < i_qp + 2 * X264_MIN(param->i_deblocking_filter_alphac0, param->i_deblocking_filter_beta) ) )
154 sh->i_disable_deblocking_filter_idc = 0;
158 sh->i_disable_deblocking_filter_idc = 1;
160 sh->i_alpha_c0_offset = param->i_deblocking_filter_alphac0 << 1;
161 sh->i_beta_offset = param->i_deblocking_filter_beta << 1;
164 static void x264_slice_header_write( bs_t *s, x264_slice_header_t *sh, int i_nal_ref_idc )
170 assert( sh->i_first_mb % (2*sh->sps->i_mb_width) == 0 );
171 bs_write_ue( s, sh->i_first_mb >> 1 );
174 bs_write_ue( s, sh->i_first_mb );
176 bs_write_ue( s, sh->i_type + 5 ); /* same type things */
177 bs_write_ue( s, sh->i_pps_id );
178 bs_write( s, sh->sps->i_log2_max_frame_num, sh->i_frame_num & ((1<<sh->sps->i_log2_max_frame_num)-1) );
180 if( !sh->sps->b_frame_mbs_only )
182 bs_write1( s, sh->b_field_pic );
183 if( sh->b_field_pic )
184 bs_write1( s, sh->b_bottom_field );
187 if( sh->i_idr_pic_id >= 0 ) /* NAL IDR */
189 bs_write_ue( s, sh->i_idr_pic_id );
192 if( sh->sps->i_poc_type == 0 )
194 bs_write( s, sh->sps->i_log2_max_poc_lsb, sh->i_poc_lsb & ((1<<sh->sps->i_log2_max_poc_lsb)-1) );
195 if( sh->pps->b_pic_order && !sh->b_field_pic )
197 bs_write_se( s, sh->i_delta_poc_bottom );
200 else if( sh->sps->i_poc_type == 1 && !sh->sps->b_delta_pic_order_always_zero )
202 bs_write_se( s, sh->i_delta_poc[0] );
203 if( sh->pps->b_pic_order && !sh->b_field_pic )
205 bs_write_se( s, sh->i_delta_poc[1] );
209 if( sh->pps->b_redundant_pic_cnt )
211 bs_write_ue( s, sh->i_redundant_pic_cnt );
214 if( sh->i_type == SLICE_TYPE_B )
216 bs_write1( s, sh->b_direct_spatial_mv_pred );
218 if( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_SP || sh->i_type == SLICE_TYPE_B )
220 bs_write1( s, sh->b_num_ref_idx_override );
221 if( sh->b_num_ref_idx_override )
223 bs_write_ue( s, sh->i_num_ref_idx_l0_active - 1 );
224 if( sh->i_type == SLICE_TYPE_B )
226 bs_write_ue( s, sh->i_num_ref_idx_l1_active - 1 );
231 /* ref pic list reordering */
232 if( sh->i_type != SLICE_TYPE_I )
234 bs_write1( s, sh->b_ref_pic_list_reordering_l0 );
235 if( sh->b_ref_pic_list_reordering_l0 )
237 for( i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
239 bs_write_ue( s, sh->ref_pic_list_order[0][i].idc );
240 bs_write_ue( s, sh->ref_pic_list_order[0][i].arg );
246 if( sh->i_type == SLICE_TYPE_B )
248 bs_write1( s, sh->b_ref_pic_list_reordering_l1 );
249 if( sh->b_ref_pic_list_reordering_l1 )
251 for( i = 0; i < sh->i_num_ref_idx_l1_active; i++ )
253 bs_write_ue( s, sh->ref_pic_list_order[1][i].idc );
254 bs_write_ue( s, sh->ref_pic_list_order[1][i].arg );
260 if( ( sh->pps->b_weighted_pred && ( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_SP ) ) ||
261 ( sh->pps->b_weighted_bipred == 1 && sh->i_type == SLICE_TYPE_B ) )
266 if( i_nal_ref_idc != 0 )
268 if( sh->i_idr_pic_id >= 0 )
270 bs_write1( s, 0 ); /* no output of prior pics flag */
271 bs_write1( s, 0 ); /* long term reference flag */
275 bs_write1( s, 0 ); /* adaptive_ref_pic_marking_mode_flag */
279 if( sh->pps->b_cabac && sh->i_type != SLICE_TYPE_I )
281 bs_write_ue( s, sh->i_cabac_init_idc );
283 bs_write_se( s, sh->i_qp_delta ); /* slice qp delta */
285 if( sh->pps->b_deblocking_filter_control )
287 bs_write_ue( s, sh->i_disable_deblocking_filter_idc );
288 if( sh->i_disable_deblocking_filter_idc != 1 )
290 bs_write_se( s, sh->i_alpha_c0_offset >> 1 );
291 bs_write_se( s, sh->i_beta_offset >> 1 );
296 /* If we are within a reasonable distance of the end of the memory allocated for the bitstream, */
297 /* reallocate, adding an arbitrary amount of space (100 kilobytes). */
298 static int x264_bitstream_check_buffer( x264_t *h )
300 uint8_t *bs_bak = h->out.p_bitstream;
301 if( ( h->param.b_cabac && (h->cabac.p_end - h->cabac.p < 2500) )
302 || ( h->out.bs.p_end - h->out.bs.p < 2500 ) )
307 h->out.i_bitstream += 100000;
308 CHECKED_MALLOC( h->out.p_bitstream, h->out.i_bitstream );
309 h->mc.memcpy_aligned( h->out.p_bitstream, bs_bak, (h->out.i_bitstream - 100000) & ~15 );
310 delta = h->out.p_bitstream - bs_bak;
312 h->out.bs.p_start += delta;
313 h->out.bs.p += delta;
314 h->out.bs.p_end = h->out.p_bitstream + h->out.i_bitstream;
316 h->cabac.p_start += delta;
318 h->cabac.p_end = h->out.p_bitstream + h->out.i_bitstream;
320 for( i = 0; i <= h->out.i_nal; i++ )
321 h->out.nal[i].p_payload += delta;
330 /****************************************************************************
332 ****************************************************************************
333 ****************************** External API*********************************
334 ****************************************************************************
336 ****************************************************************************/
338 static int x264_validate_parameters( x264_t *h )
341 if( !(x264_cpu_detect() & X264_CPU_SSE) )
343 x264_log( h, X264_LOG_ERROR, "your cpu does not support SSE1, but x264 was compiled with asm support\n");
344 x264_log( h, X264_LOG_ERROR, "to run x264, recompile without asm support (configure --disable-asm)\n");
348 if( h->param.i_width <= 0 || h->param.i_height <= 0 )
350 x264_log( h, X264_LOG_ERROR, "invalid width x height (%dx%d)\n",
351 h->param.i_width, h->param.i_height );
355 if( h->param.i_width % 2 || h->param.i_height % 2 )
357 x264_log( h, X264_LOG_ERROR, "width or height not divisible by 2 (%dx%d)\n",
358 h->param.i_width, h->param.i_height );
361 if( h->param.i_csp != X264_CSP_I420 )
363 x264_log( h, X264_LOG_ERROR, "invalid CSP (only I420 supported)\n" );
367 if( h->param.i_threads == X264_THREADS_AUTO )
368 h->param.i_threads = x264_cpu_num_processors() * 3/2;
369 h->param.i_threads = x264_clip3( h->param.i_threads, 1, X264_THREAD_MAX );
370 if( h->param.i_threads > 1 )
373 x264_log( h, X264_LOG_WARNING, "not compiled with pthread support!\n");
374 h->param.i_threads = 1;
378 if( h->param.b_interlaced )
380 if( h->param.analyse.i_me_method >= X264_ME_ESA )
382 x264_log( h, X264_LOG_WARNING, "interlace + me=esa is not implemented\n" );
383 h->param.analyse.i_me_method = X264_ME_UMH;
385 if( h->param.analyse.i_direct_mv_pred > X264_DIRECT_PRED_SPATIAL )
387 x264_log( h, X264_LOG_WARNING, "interlace + direct=temporal is not implemented\n" );
388 h->param.analyse.i_direct_mv_pred = X264_DIRECT_PRED_SPATIAL;
392 /* Detect default ffmpeg settings and terminate with an error. */
395 score += h->param.analyse.i_me_range == 0;
396 score += h->param.rc.i_qp_step == 3;
397 score += h->param.i_keyint_max == 12;
398 score += h->param.rc.i_qp_min == 2;
399 score += h->param.rc.i_qp_max == 31;
400 score += h->param.rc.f_qcompress == 0.5;
401 score += fabs(h->param.rc.f_ip_factor - 1.25) < 0.01;
402 score += fabs(h->param.rc.f_pb_factor - 1.25) < 0.01;
403 score += h->param.analyse.inter == 0 && h->param.analyse.i_subpel_refine == 8;
406 x264_log( h, X264_LOG_ERROR, "broken ffmpeg default settings detected\n" );
407 x264_log( h, X264_LOG_ERROR, "use an encoding preset (vpre)\n" );
412 if( h->param.rc.i_rc_method < 0 || h->param.rc.i_rc_method > 2 )
414 x264_log( h, X264_LOG_ERROR, "no ratecontrol method specified\n" );
417 h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, 0, 51 );
418 h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, 0, 51 );
419 if( h->param.rc.i_rc_method == X264_RC_CRF )
421 h->param.rc.i_qp_constant = h->param.rc.f_rf_constant;
422 h->param.rc.i_bitrate = 0;
424 if( (h->param.rc.i_rc_method == X264_RC_CQP || h->param.rc.i_rc_method == X264_RC_CRF)
425 && h->param.rc.i_qp_constant == 0 )
427 h->mb.b_lossless = 1;
428 h->param.i_cqm_preset = X264_CQM_FLAT;
429 h->param.psz_cqm_file = NULL;
430 h->param.rc.i_rc_method = X264_RC_CQP;
431 h->param.rc.f_ip_factor = 1;
432 h->param.rc.f_pb_factor = 1;
433 h->param.analyse.b_psnr = 0;
434 h->param.analyse.b_ssim = 0;
435 h->param.analyse.i_chroma_qp_offset = 0;
436 h->param.analyse.i_trellis = 0;
437 h->param.analyse.b_fast_pskip = 0;
438 h->param.analyse.i_noise_reduction = 0;
439 h->param.analyse.f_psy_rd = 0;
440 h->param.i_bframe = 0;
441 /* 8x8dct is not useful at all in CAVLC lossless */
442 if( !h->param.b_cabac )
443 h->param.analyse.b_transform_8x8 = 0;
445 if( h->param.rc.i_rc_method == X264_RC_CQP )
447 float qp_p = h->param.rc.i_qp_constant;
448 float qp_i = qp_p - 6*log(h->param.rc.f_ip_factor)/log(2);
449 float qp_b = qp_p + 6*log(h->param.rc.f_pb_factor)/log(2);
450 h->param.rc.i_qp_min = x264_clip3( (int)(X264_MIN3( qp_p, qp_i, qp_b )), 0, 51 );
451 h->param.rc.i_qp_max = x264_clip3( (int)(X264_MAX3( qp_p, qp_i, qp_b ) + .999), 0, 51 );
452 h->param.rc.i_aq_mode = 0;
453 h->param.rc.b_mb_tree = 0;
455 h->param.rc.i_qp_max = x264_clip3( h->param.rc.i_qp_max, 0, 51 );
456 h->param.rc.i_qp_min = x264_clip3( h->param.rc.i_qp_min, 0, h->param.rc.i_qp_max );
458 if( ( h->param.i_width % 16 || h->param.i_height % 16 )
459 && h->param.i_height != 1080 && !h->mb.b_lossless )
461 // There's nothing special about 1080 in that the warning still applies to it,
462 // but chances are the user can't help it if his content is already 1080p,
463 // so there's no point in warning in that case.
464 x264_log( h, X264_LOG_WARNING,
465 "width or height not divisible by 16 (%dx%d), compression will suffer.\n",
466 h->param.i_width, h->param.i_height );
469 int max_slices = (h->param.i_height+((16<<h->param.b_interlaced)-1))/(16<<h->param.b_interlaced);
470 h->param.i_slice_count = x264_clip3( h->param.i_slice_count, 0, max_slices );
471 h->param.i_slice_max_size = X264_MAX( h->param.i_slice_max_size, 0 );
472 h->param.i_slice_max_mbs = X264_MAX( h->param.i_slice_max_mbs, 0 );
473 if( h->param.b_interlaced && h->param.i_slice_max_size )
475 x264_log( h, X264_LOG_WARNING, "interlaced + slice-max-size is not implemented\n" );
476 h->param.i_slice_max_size = 0;
478 if( h->param.b_interlaced && h->param.i_slice_max_mbs )
480 x264_log( h, X264_LOG_WARNING, "interlaced + slice-max-mbs is not implemented\n" );
481 h->param.i_slice_max_mbs = 0;
483 if( h->param.i_slice_max_mbs || h->param.i_slice_max_size )
484 h->param.i_slice_count = 0;
486 h->param.i_frame_reference = x264_clip3( h->param.i_frame_reference, 1, 16 );
487 if( h->param.i_keyint_max <= 0 )
488 h->param.i_keyint_max = 1;
489 if( h->param.i_scenecut_threshold < 0 )
490 h->param.i_scenecut_threshold = 0;
491 h->param.i_keyint_min = x264_clip3( h->param.i_keyint_min, 1, h->param.i_keyint_max/2+1 );
492 if( !h->param.analyse.i_subpel_refine && h->param.analyse.i_direct_mv_pred > X264_DIRECT_PRED_SPATIAL )
494 x264_log( h, X264_LOG_WARNING, "subme=0 + direct=temporal is not supported\n" );
495 h->param.analyse.i_direct_mv_pred = X264_DIRECT_PRED_SPATIAL;
497 h->param.i_bframe = x264_clip3( h->param.i_bframe, 0, X264_BFRAME_MAX );
498 if( h->param.i_keyint_max == 1 )
499 h->param.i_bframe = 0;
500 h->param.i_bframe_bias = x264_clip3( h->param.i_bframe_bias, -90, 100 );
501 h->param.b_bframe_pyramid = h->param.b_bframe_pyramid && h->param.i_bframe > 1;
502 if( !h->param.i_bframe )
504 h->param.i_bframe_adaptive = X264_B_ADAPT_NONE;
505 h->param.analyse.i_direct_mv_pred = 0;
506 h->param.analyse.b_weighted_bipred = 0;
508 h->param.rc.i_lookahead = x264_clip3( h->param.rc.i_lookahead, 0, X264_LOOKAHEAD_MAX );
510 int maxrate = X264_MAX( h->param.rc.i_vbv_max_bitrate, h->param.rc.i_bitrate );
511 float bufsize = maxrate ? (float)h->param.rc.i_vbv_buffer_size / maxrate : 0;
512 float fps = h->param.i_fps_num > 0 && h->param.i_fps_den > 0 ? (float) h->param.i_fps_num / h->param.i_fps_den : 25.0;
513 h->param.rc.i_lookahead = X264_MIN( h->param.rc.i_lookahead, X264_MAX( h->param.i_keyint_max, bufsize*fps ) );
516 h->param.rc.f_qcompress = x264_clip3f( h->param.rc.f_qcompress, 0.0, 1.0 );
517 if( !h->param.rc.i_lookahead || h->param.i_keyint_max == 1 || h->param.rc.f_qcompress == 1 )
518 h->param.rc.b_mb_tree = 0;
519 if( h->param.rc.b_stat_read )
520 h->param.rc.i_lookahead = 0;
522 if( h->param.i_sync_lookahead )
523 h->param.i_sync_lookahead = x264_clip3( h->param.i_sync_lookahead, h->param.i_threads + h->param.i_bframe, X264_LOOKAHEAD_MAX );
524 if( h->param.rc.b_stat_read || h->param.i_threads == 1 )
525 h->param.i_sync_lookahead = 0;
527 h->param.i_sync_lookahead = 0;
530 h->mb.b_direct_auto_write = h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO
532 && ( h->param.rc.b_stat_write || !h->param.rc.b_stat_read );
534 h->param.i_deblocking_filter_alphac0 = x264_clip3( h->param.i_deblocking_filter_alphac0, -6, 6 );
535 h->param.i_deblocking_filter_beta = x264_clip3( h->param.i_deblocking_filter_beta, -6, 6 );
536 h->param.analyse.i_luma_deadzone[0] = x264_clip3( h->param.analyse.i_luma_deadzone[0], 0, 32 );
537 h->param.analyse.i_luma_deadzone[1] = x264_clip3( h->param.analyse.i_luma_deadzone[1], 0, 32 );
539 h->param.i_cabac_init_idc = x264_clip3( h->param.i_cabac_init_idc, 0, 2 );
541 if( h->param.i_cqm_preset < X264_CQM_FLAT || h->param.i_cqm_preset > X264_CQM_CUSTOM )
542 h->param.i_cqm_preset = X264_CQM_FLAT;
544 if( h->param.analyse.i_me_method < X264_ME_DIA ||
545 h->param.analyse.i_me_method > X264_ME_TESA )
546 h->param.analyse.i_me_method = X264_ME_HEX;
547 if( h->param.analyse.i_me_range < 4 )
548 h->param.analyse.i_me_range = 4;
549 if( h->param.analyse.i_me_range > 16 && h->param.analyse.i_me_method <= X264_ME_HEX )
550 h->param.analyse.i_me_range = 16;
551 if( h->param.analyse.i_me_method == X264_ME_TESA &&
552 (h->mb.b_lossless || h->param.analyse.i_subpel_refine <= 1) )
553 h->param.analyse.i_me_method = X264_ME_ESA;
554 h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 10 );
555 h->param.analyse.b_mixed_references = h->param.analyse.b_mixed_references && h->param.i_frame_reference > 1;
556 h->param.analyse.inter &= X264_ANALYSE_PSUB16x16|X264_ANALYSE_PSUB8x8|X264_ANALYSE_BSUB16x16|
557 X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
558 h->param.analyse.intra &= X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
559 if( !(h->param.analyse.inter & X264_ANALYSE_PSUB16x16) )
560 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
561 if( !h->param.analyse.b_transform_8x8 )
563 h->param.analyse.inter &= ~X264_ANALYSE_I8x8;
564 h->param.analyse.intra &= ~X264_ANALYSE_I8x8;
566 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
567 if( !h->param.b_cabac )
568 h->param.analyse.i_trellis = 0;
569 h->param.analyse.i_trellis = x264_clip3( h->param.analyse.i_trellis, 0, 2 );
570 if( !h->param.analyse.b_psy )
572 h->param.analyse.f_psy_rd = 0;
573 h->param.analyse.f_psy_trellis = 0;
575 if( !h->param.analyse.i_trellis )
576 h->param.analyse.f_psy_trellis = 0;
577 h->param.analyse.f_psy_rd = x264_clip3f( h->param.analyse.f_psy_rd, 0, 10 );
578 h->param.analyse.f_psy_trellis = x264_clip3f( h->param.analyse.f_psy_trellis, 0, 10 );
579 if( h->param.analyse.i_subpel_refine < 6 )
580 h->param.analyse.f_psy_rd = 0;
581 h->mb.i_psy_rd = FIX8( h->param.analyse.f_psy_rd );
582 /* Psy RDO increases overall quantizers to improve the quality of luma--this indirectly hurts chroma quality */
583 /* so we lower the chroma QP offset to compensate */
584 /* This can be triggered repeatedly on multiple calls to parameter_validate, but since encoding
585 * uses the pps chroma qp offset not the param chroma qp offset, this is not a problem. */
587 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_rd < 0.25 ? 1 : 2;
588 h->mb.i_psy_trellis = FIX8( h->param.analyse.f_psy_trellis / 4 );
589 /* Psy trellis has a similar effect. */
590 if( h->mb.i_psy_trellis )
591 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_trellis < 0.25 ? 1 : 2;
593 h->mb.i_psy_trellis = 0;
594 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
595 h->param.rc.i_aq_mode = x264_clip3( h->param.rc.i_aq_mode, 0, 2 );
596 h->param.rc.f_aq_strength = x264_clip3f( h->param.rc.f_aq_strength, 0, 3 );
597 if( h->param.rc.f_aq_strength == 0 )
598 h->param.rc.i_aq_mode = 0;
599 /* MB-tree requires AQ to be on, even if the strength is zero. */
600 if( !h->param.rc.i_aq_mode && h->param.rc.b_mb_tree )
602 h->param.rc.i_aq_mode = 1;
603 h->param.rc.f_aq_strength = 0;
605 if( h->param.rc.b_mb_tree && h->param.b_bframe_pyramid )
607 x264_log( h, X264_LOG_WARNING, "b-pyramid + mb-tree is not supported\n" );
608 h->param.b_bframe_pyramid = 0;
610 h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
611 if( h->param.analyse.i_subpel_refine == 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
612 h->param.analyse.i_subpel_refine = 9;
615 const x264_level_t *l = x264_levels;
616 if( h->param.i_level_idc < 0 )
618 int maxrate_bak = h->param.rc.i_vbv_max_bitrate;
619 if( h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.i_vbv_buffer_size <= 0 )
620 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate * 2;
621 h->sps = h->sps_array;
622 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
623 do h->param.i_level_idc = l->level_idc;
624 while( l[1].level_idc && x264_validate_levels( h, 0 ) && l++ );
625 h->param.rc.i_vbv_max_bitrate = maxrate_bak;
629 while( l->level_idc && l->level_idc != h->param.i_level_idc )
631 if( l->level_idc == 0 )
633 x264_log( h, X264_LOG_ERROR, "invalid level_idc: %d\n", h->param.i_level_idc );
637 if( h->param.analyse.i_mv_range <= 0 )
638 h->param.analyse.i_mv_range = l->mv_range >> h->param.b_interlaced;
640 h->param.analyse.i_mv_range = x264_clip3(h->param.analyse.i_mv_range, 32, 512 >> h->param.b_interlaced);
643 if( h->param.i_threads > 1 )
645 int r = h->param.analyse.i_mv_range_thread;
649 // half of the available space is reserved and divided evenly among the threads,
650 // the rest is allocated to whichever thread is far enough ahead to use it.
651 // reserving more space increases quality for some videos, but costs more time
652 // in thread synchronization.
653 int max_range = (h->param.i_height + X264_THREAD_HEIGHT) / h->param.i_threads - X264_THREAD_HEIGHT;
656 r = X264_MAX( r, h->param.analyse.i_me_range );
657 r = X264_MIN( r, h->param.analyse.i_mv_range );
658 // round up to use the whole mb row
659 r2 = (r & ~15) + ((-X264_THREAD_HEIGHT) & 15);
662 x264_log( h, X264_LOG_DEBUG, "using mv_range_thread = %d\n", r2 );
663 h->param.analyse.i_mv_range_thread = r2;
666 if( h->param.rc.f_qblur < 0 )
667 h->param.rc.f_qblur = 0;
668 if( h->param.rc.f_complexity_blur < 0 )
669 h->param.rc.f_complexity_blur = 0;
671 h->param.i_sps_id &= 31;
673 if( h->param.i_log_level < X264_LOG_INFO )
675 h->param.analyse.b_psnr = 0;
676 h->param.analyse.b_ssim = 0;
679 /* ensure the booleans are 0 or 1 so they can be used in math */
680 #define BOOLIFY(x) h->param.x = !!h->param.x
682 BOOLIFY( b_deblocking_filter );
683 BOOLIFY( b_interlaced );
684 BOOLIFY( analyse.b_transform_8x8 );
685 BOOLIFY( analyse.b_chroma_me );
686 BOOLIFY( analyse.b_fast_pskip );
687 BOOLIFY( rc.b_stat_write );
688 BOOLIFY( rc.b_stat_read );
689 BOOLIFY( rc.b_mb_tree );
695 static void mbcmp_init( x264_t *h )
697 int satd = !h->mb.b_lossless && h->param.analyse.i_subpel_refine > 1;
698 memcpy( h->pixf.mbcmp, satd ? h->pixf.satd : h->pixf.sad_aligned, sizeof(h->pixf.mbcmp) );
699 memcpy( h->pixf.mbcmp_unaligned, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.mbcmp_unaligned) );
700 h->pixf.intra_mbcmp_x3_16x16 = satd ? h->pixf.intra_satd_x3_16x16 : h->pixf.intra_sad_x3_16x16;
701 h->pixf.intra_mbcmp_x3_8x8c = satd ? h->pixf.intra_satd_x3_8x8c : h->pixf.intra_sad_x3_8x8c;
702 h->pixf.intra_mbcmp_x3_4x4 = satd ? h->pixf.intra_satd_x3_4x4 : h->pixf.intra_sad_x3_4x4;
703 satd &= h->param.analyse.i_me_method == X264_ME_TESA;
704 memcpy( h->pixf.fpelcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.fpelcmp) );
705 memcpy( h->pixf.fpelcmp_x3, satd ? h->pixf.satd_x3 : h->pixf.sad_x3, sizeof(h->pixf.fpelcmp_x3) );
706 memcpy( h->pixf.fpelcmp_x4, satd ? h->pixf.satd_x4 : h->pixf.sad_x4, sizeof(h->pixf.fpelcmp_x4) );
709 static void x264_set_aspect_ratio( x264_t *h, x264_param_t *param, int initial )
712 if( param->vui.i_sar_width > 0 && param->vui.i_sar_height > 0 )
714 int i_w = param->vui.i_sar_width;
715 int i_h = param->vui.i_sar_height;
716 int old_w = h->param.vui.i_sar_width;
717 int old_h = h->param.vui.i_sar_height;
719 x264_reduce_fraction( &i_w, &i_h );
721 while( i_w > 65535 || i_h > 65535 )
727 x264_reduce_fraction( &i_w, &i_h );
729 if( i_w != old_w || i_h != old_h || initial )
731 h->param.vui.i_sar_width = 0;
732 h->param.vui.i_sar_height = 0;
733 if( i_w == 0 || i_h == 0 )
734 x264_log( h, X264_LOG_WARNING, "cannot create valid sample aspect ratio\n" );
737 x264_log( h, initial?X264_LOG_INFO:X264_LOG_DEBUG, "using SAR=%d/%d\n", i_w, i_h );
738 h->param.vui.i_sar_width = i_w;
739 h->param.vui.i_sar_height = i_h;
745 /****************************************************************************
747 ****************************************************************************/
748 x264_t *x264_encoder_open( x264_param_t *param )
752 int i, qp, i_slicetype_length;
754 CHECKED_MALLOCZERO( h, sizeof(x264_t) );
756 /* Create a copy of param */
757 memcpy( &h->param, param, sizeof(x264_param_t) );
759 if( param->param_free )
760 param->param_free( param );
762 if( x264_validate_parameters( h ) < 0 )
765 if( h->param.psz_cqm_file )
766 if( x264_cqm_parse_file( h, h->param.psz_cqm_file ) < 0 )
769 if( h->param.rc.psz_stat_out )
770 h->param.rc.psz_stat_out = strdup( h->param.rc.psz_stat_out );
771 if( h->param.rc.psz_stat_in )
772 h->param.rc.psz_stat_in = strdup( h->param.rc.psz_stat_in );
774 x264_set_aspect_ratio( h, param, 1 );
776 x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den );
783 h->sps = &h->sps_array[0];
784 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
786 h->pps = &h->pps_array[0];
787 x264_pps_init( h->pps, h->param.i_sps_id, &h->param, h->sps );
789 x264_validate_levels( h, 1 );
791 h->chroma_qp_table = i_chroma_qp_table + 12 + h->pps->i_chroma_qp_index_offset;
793 if( x264_cqm_init( h ) < 0 )
796 h->mb.i_mb_count = h->sps->i_mb_width * h->sps->i_mb_height;
799 if( h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS )
800 h->frames.i_delay = X264_MAX(h->param.i_bframe,3)*4;
802 h->frames.i_delay = h->param.i_bframe;
803 if( h->param.rc.b_mb_tree || h->param.rc.i_vbv_buffer_size )
804 h->frames.i_delay = X264_MAX( h->frames.i_delay, h->param.rc.i_lookahead );
805 i_slicetype_length = h->frames.i_delay;
806 h->frames.i_delay += h->param.i_threads - 1;
807 h->frames.i_delay = X264_MIN( h->frames.i_delay, X264_LOOKAHEAD_MAX );
808 h->frames.i_delay += h->param.i_sync_lookahead;
810 h->frames.i_max_ref0 = h->param.i_frame_reference;
811 h->frames.i_max_ref1 = h->sps->vui.i_num_reorder_frames;
812 h->frames.i_max_dpb = h->sps->vui.i_max_dec_frame_buffering;
813 h->frames.b_have_lowres = !h->param.rc.b_stat_read
814 && ( h->param.rc.i_rc_method == X264_RC_ABR
815 || h->param.rc.i_rc_method == X264_RC_CRF
816 || h->param.i_bframe_adaptive
817 || h->param.i_scenecut_threshold
818 || h->param.rc.b_mb_tree );
819 h->frames.b_have_lowres |= h->param.rc.b_stat_read && h->param.rc.i_vbv_buffer_size > 0;
820 h->frames.b_have_sub8x8_esa = !!(h->param.analyse.inter & X264_ANALYSE_PSUB8x8);
822 h->frames.i_last_idr = - h->param.i_keyint_max;
823 h->frames.i_input = 0;
825 CHECKED_MALLOCZERO( h->frames.unused[0], (h->frames.i_delay + 3) * sizeof(x264_frame_t *) );
826 /* Allocate room for max refs plus a few extra just in case. */
827 CHECKED_MALLOCZERO( h->frames.unused[1], (h->param.i_threads + 20) * sizeof(x264_frame_t *) );
828 CHECKED_MALLOCZERO( h->frames.current, (h->param.i_sync_lookahead + h->param.i_bframe
829 + h->param.i_threads + 3) * sizeof(x264_frame_t *) );
836 /* init CPU functions */
837 x264_predict_16x16_init( h->param.cpu, h->predict_16x16 );
838 x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c );
839 x264_predict_8x8_init( h->param.cpu, h->predict_8x8, &h->predict_8x8_filter );
840 x264_predict_4x4_init( h->param.cpu, h->predict_4x4 );
841 if( !h->param.b_cabac )
842 x264_init_vlc_tables();
843 x264_pixel_init( h->param.cpu, &h->pixf );
844 x264_dct_init( h->param.cpu, &h->dctf );
845 x264_zigzag_init( h->param.cpu, &h->zigzagf, h->param.b_interlaced );
846 x264_mc_init( h->param.cpu, &h->mc );
847 x264_quant_init( h, h->param.cpu, &h->quantf );
848 x264_deblock_init( h->param.cpu, &h->loopf );
849 x264_dct_init_weights();
853 p = buf + sprintf( buf, "using cpu capabilities:" );
854 for( i=0; x264_cpu_names[i].flags; i++ )
856 if( !strcmp(x264_cpu_names[i].name, "SSE2")
857 && param->cpu & (X264_CPU_SSE2_IS_FAST|X264_CPU_SSE2_IS_SLOW) )
859 if( !strcmp(x264_cpu_names[i].name, "SSE3")
860 && (param->cpu & X264_CPU_SSSE3 || !(param->cpu & X264_CPU_CACHELINE_64)) )
862 if( !strcmp(x264_cpu_names[i].name, "SSE4.1")
863 && (param->cpu & X264_CPU_SSE42) )
865 if( (param->cpu & x264_cpu_names[i].flags) == x264_cpu_names[i].flags
866 && (!i || x264_cpu_names[i].flags != x264_cpu_names[i-1].flags) )
867 p += sprintf( p, " %s", x264_cpu_names[i].name );
870 p += sprintf( p, " none!" );
871 x264_log( h, X264_LOG_INFO, "%s\n", buf );
873 for( qp = h->param.rc.i_qp_min; qp <= h->param.rc.i_qp_max; qp++ )
874 if( x264_analyse_init_costs( h, qp ) )
876 if( x264_analyse_init_costs( h, X264_LOOKAHEAD_QP ) )
878 if( h->cost_mv[1][2013] != 24 )
880 x264_log( h, X264_LOG_ERROR, "MV cost test failed: x264 has been miscompiled!\n" );
885 h->out.i_bitstream = X264_MAX( 1000000, h->param.i_width * h->param.i_height * 4
886 * ( h->param.rc.i_rc_method == X264_RC_ABR ? pow( 0.95, h->param.rc.i_qp_min )
887 : pow( 0.95, h->param.rc.i_qp_constant ) * X264_MAX( 1, h->param.rc.f_ip_factor )));
889 CHECKED_MALLOC( h->nal_buffer, h->out.i_bitstream * 3/2 + 4 );
890 h->nal_buffer_size = h->out.i_bitstream * 3/2 + 4;
894 for( i = 1; i < h->param.i_threads + !!h->param.i_sync_lookahead; i++ )
895 CHECKED_MALLOC( h->thread[i], sizeof(x264_t) );
897 for( i = 0; i < h->param.i_threads; i++ )
901 h->thread[i]->fdec = x264_frame_pop_unused( h, 1 );
902 if( !h->thread[i]->fdec )
904 CHECKED_MALLOC( h->thread[i]->out.p_bitstream, h->out.i_bitstream );
905 /* Start each thread with room for 8 NAL units; it'll realloc later if needed. */
906 CHECKED_MALLOC( h->thread[i]->out.nal, 8*sizeof(x264_nal_t) );
907 h->thread[i]->out.i_nals_allocated = 8;
908 if( x264_macroblock_cache_init( h->thread[i] ) < 0 )
912 if( x264_lookahead_init( h, i_slicetype_length ) )
915 if( x264_ratecontrol_new( h ) < 0 )
918 if( h->param.psz_dump_yuv )
920 /* create or truncate the reconstructed video file */
921 FILE *f = fopen( h->param.psz_dump_yuv, "w" );
926 x264_log( h, X264_LOG_ERROR, "can't write to fdec.yuv\n" );
931 x264_log( h, X264_LOG_INFO, "profile %s, level %d.%d\n",
932 h->sps->i_profile_idc == PROFILE_BASELINE ? "Baseline" :
933 h->sps->i_profile_idc == PROFILE_MAIN ? "Main" :
934 h->sps->i_profile_idc == PROFILE_HIGH ? "High" :
935 "High 4:4:4 Predictive", h->sps->i_level_idc/10, h->sps->i_level_idc%10 );
943 /****************************************************************************
944 * x264_encoder_reconfig:
945 ****************************************************************************/
946 int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
948 h = h->thread[h->i_thread_phase];
949 x264_set_aspect_ratio( h, param, 0 );
950 #define COPY(var) h->param.var = param->var
951 COPY( i_frame_reference ); // but never uses more refs than initially specified
952 COPY( i_bframe_bias );
953 if( h->param.i_scenecut_threshold )
954 COPY( i_scenecut_threshold ); // can't turn it on or off, only vary the threshold
955 COPY( b_deblocking_filter );
956 COPY( i_deblocking_filter_alphac0 );
957 COPY( i_deblocking_filter_beta );
958 COPY( analyse.intra );
959 COPY( analyse.inter );
960 COPY( analyse.i_direct_mv_pred );
961 /* Scratch buffer prevents me_range from being increased for esa/tesa */
962 if( h->param.analyse.i_me_method < X264_ME_ESA || param->analyse.i_me_range < h->param.analyse.i_me_range )
963 COPY( analyse.i_me_range );
964 COPY( analyse.i_noise_reduction );
965 /* We can't switch out of subme=0 during encoding. */
966 if( h->param.analyse.i_subpel_refine )
967 COPY( analyse.i_subpel_refine );
968 COPY( analyse.i_trellis );
969 COPY( analyse.b_chroma_me );
970 COPY( analyse.b_dct_decimate );
971 COPY( analyse.b_fast_pskip );
972 COPY( analyse.b_mixed_references );
973 COPY( analyse.f_psy_rd );
974 COPY( analyse.f_psy_trellis );
975 // can only twiddle these if they were enabled to begin with:
976 if( h->param.analyse.i_me_method >= X264_ME_ESA || param->analyse.i_me_method < X264_ME_ESA )
977 COPY( analyse.i_me_method );
978 if( h->param.analyse.i_me_method >= X264_ME_ESA && !h->frames.b_have_sub8x8_esa )
979 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
980 if( h->pps->b_transform_8x8_mode )
981 COPY( analyse.b_transform_8x8 );
982 if( h->frames.i_max_ref1 > 1 )
983 COPY( b_bframe_pyramid );
984 COPY( i_slice_max_size );
985 COPY( i_slice_max_mbs );
986 COPY( i_slice_count );
991 return x264_validate_parameters( h );
995 static void x264_nal_start( x264_t *h, int i_type, int i_ref_idc )
997 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
999 nal->i_ref_idc = i_ref_idc;
1000 nal->i_type = i_type;
1003 nal->p_payload= &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1005 static int x264_nal_end( x264_t *h )
1007 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1008 nal->i_payload = &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8] - nal->p_payload;
1011 /* if number of allocated nals is not enough, re-allocate a larger one. */
1012 if( h->out.i_nal >= h->out.i_nals_allocated )
1014 x264_nal_t *new_out = x264_malloc( sizeof(x264_nal_t) * (h->out.i_nals_allocated*2) );
1017 memcpy( new_out, h->out.nal, sizeof(x264_nal_t) * (h->out.i_nals_allocated) );
1018 x264_free( h->out.nal );
1019 h->out.nal = new_out;
1020 h->out.i_nals_allocated *= 2;
1025 static int x264_encoder_encapsulate_nals( x264_t *h )
1027 int nal_size = 0, i;
1028 for( i = 0; i < h->out.i_nal; i++ )
1029 nal_size += h->out.nal[i].i_payload;
1031 /* Worst-case NAL unit escaping: reallocate the buffer if it's too small. */
1032 if( h->nal_buffer_size < nal_size * 3/2 + h->out.i_nal * 4 )
1034 uint8_t *buf = x264_malloc( nal_size * 2 + h->out.i_nal * 4 );
1037 x264_free( h->nal_buffer );
1038 h->nal_buffer = buf;
1041 uint8_t *nal_buffer = h->nal_buffer;
1043 for( i = 0; i < h->out.i_nal; i++ )
1045 int size = x264_nal_encode( nal_buffer, h->param.b_annexb, &h->out.nal[i] );
1046 h->out.nal[i].i_payload = size;
1047 h->out.nal[i].p_payload = nal_buffer;
1051 return nal_buffer - h->nal_buffer;
1054 /****************************************************************************
1055 * x264_encoder_headers:
1056 ****************************************************************************/
1057 int x264_encoder_headers( x264_t *h, x264_nal_t **pp_nal, int *pi_nal )
1060 /* init bitstream context */
1062 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
1064 /* Write SEI, SPS and PPS. */
1065 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
1066 if( x264_sei_version_write( h, &h->out.bs ) )
1068 if( x264_nal_end( h ) )
1071 /* generate sequence parameters */
1072 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
1073 x264_sps_write( &h->out.bs, h->sps );
1074 if( x264_nal_end( h ) )
1077 /* generate picture parameters */
1078 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
1079 x264_pps_write( &h->out.bs, h->pps );
1080 if( x264_nal_end( h ) )
1082 bs_flush( &h->out.bs );
1084 frame_size = x264_encoder_encapsulate_nals( h );
1087 *pi_nal = h->out.i_nal;
1088 *pp_nal = &h->out.nal[0];
1094 static inline void x264_reference_build_list( x264_t *h, int i_poc )
1099 /* build ref list 0/1 */
1102 for( i = 0; h->frames.reference[i]; i++ )
1104 if( h->frames.reference[i]->i_poc < i_poc )
1106 h->fref0[h->i_ref0++] = h->frames.reference[i];
1108 else if( h->frames.reference[i]->i_poc > i_poc )
1110 h->fref1[h->i_ref1++] = h->frames.reference[i];
1114 /* Order ref0 from higher to lower poc */
1118 for( i = 0; i < h->i_ref0 - 1; i++ )
1120 if( h->fref0[i]->i_poc < h->fref0[i+1]->i_poc )
1122 XCHG( x264_frame_t*, h->fref0[i], h->fref0[i+1] );
1128 /* Order ref1 from lower to higher poc (bubble sort) for B-frame */
1132 for( i = 0; i < h->i_ref1 - 1; i++ )
1134 if( h->fref1[i]->i_poc > h->fref1[i+1]->i_poc )
1136 XCHG( x264_frame_t*, h->fref1[i], h->fref1[i+1] );
1143 h->i_ref1 = X264_MIN( h->i_ref1, h->frames.i_max_ref1 );
1144 h->i_ref0 = X264_MIN( h->i_ref0, h->frames.i_max_ref0 );
1145 h->i_ref0 = X264_MIN( h->i_ref0, h->param.i_frame_reference ); // if reconfig() has lowered the limit
1146 assert( h->i_ref0 + h->i_ref1 <= 16 );
1147 h->mb.pic.i_fref[0] = h->i_ref0;
1148 h->mb.pic.i_fref[1] = h->i_ref1;
1151 static void x264_fdec_filter_row( x264_t *h, int mb_y )
1153 /* mb_y is the mb to be encoded next, not the mb to be filtered here */
1154 int b_hpel = h->fdec->b_kept_as_ref;
1155 int b_deblock = !h->sh.i_disable_deblocking_filter_idc;
1156 int b_end = mb_y == h->sps->i_mb_height;
1157 int min_y = mb_y - (1 << h->sh.b_mbaff);
1158 int max_y = b_end ? h->sps->i_mb_height : mb_y;
1159 b_deblock &= b_hpel || h->param.psz_dump_yuv;
1160 if( mb_y & h->sh.b_mbaff )
1168 for( j=0; j<=h->sh.b_mbaff; j++ )
1169 for( i=0; i<3; i++ )
1171 memcpy( h->mb.intra_border_backup[j][i],
1172 h->fdec->plane[i] + ((mb_y*16 >> !!i) + j - 1 - h->sh.b_mbaff) * h->fdec->i_stride[i],
1173 h->sps->i_mb_width*16 >> !!i );
1180 for( y = min_y; y < max_y; y += (1 << h->sh.b_mbaff) )
1181 x264_frame_deblock_row( h, y );
1186 x264_frame_expand_border( h, h->fdec, min_y, b_end );
1187 if( h->param.analyse.i_subpel_refine )
1189 x264_frame_filter( h, h->fdec, min_y, b_end );
1190 x264_frame_expand_border_filtered( h, h->fdec, min_y, b_end );
1194 if( h->param.i_threads > 1 && h->fdec->b_kept_as_ref )
1196 x264_frame_cond_broadcast( h->fdec, mb_y*16 + (b_end ? 10000 : -(X264_THREAD_HEIGHT << h->sh.b_mbaff)) );
1199 min_y = X264_MAX( min_y*16-8, 0 );
1200 max_y = b_end ? h->param.i_height : mb_y*16-8;
1202 if( h->param.analyse.b_psnr )
1205 for( i=0; i<3; i++ )
1206 h->stat.frame.i_ssd[i] +=
1207 x264_pixel_ssd_wxh( &h->pixf,
1208 h->fdec->plane[i] + (min_y>>!!i) * h->fdec->i_stride[i], h->fdec->i_stride[i],
1209 h->fenc->plane[i] + (min_y>>!!i) * h->fenc->i_stride[i], h->fenc->i_stride[i],
1210 h->param.i_width >> !!i, (max_y-min_y) >> !!i );
1213 if( h->param.analyse.b_ssim )
1216 /* offset by 2 pixels to avoid alignment of ssim blocks with dct blocks,
1217 * and overlap by 4 */
1218 min_y += min_y == 0 ? 2 : -6;
1219 h->stat.frame.f_ssim +=
1220 x264_pixel_ssim_wxh( &h->pixf,
1221 h->fdec->plane[0] + 2+min_y*h->fdec->i_stride[0], h->fdec->i_stride[0],
1222 h->fenc->plane[0] + 2+min_y*h->fenc->i_stride[0], h->fenc->i_stride[0],
1223 h->param.i_width-2, max_y-min_y, h->scratch_buffer );
1227 static inline int x264_reference_update( x264_t *h )
1229 if( !h->fdec->b_kept_as_ref )
1231 if( h->param.i_threads > 1 )
1233 x264_frame_push_unused( h, h->fdec );
1234 h->fdec = x264_frame_pop_unused( h, 1 );
1241 /* move frame in the buffer */
1242 x264_frame_push( h->frames.reference, h->fdec );
1243 if( h->frames.reference[h->frames.i_max_dpb] )
1244 x264_frame_push_unused( h, x264_frame_shift( h->frames.reference ) );
1245 h->fdec = x264_frame_pop_unused( h, 1 );
1251 static inline void x264_reference_reset( x264_t *h )
1253 while( h->frames.reference[0] )
1254 x264_frame_push_unused( h, x264_frame_pop( h->frames.reference ) );
1259 static inline void x264_slice_init( x264_t *h, int i_nal_type, int i_global_qp )
1261 /* ------------------------ Create slice header ----------------------- */
1262 if( i_nal_type == NAL_SLICE_IDR )
1264 x264_slice_header_init( h, &h->sh, h->sps, h->pps, h->i_idr_pic_id, h->i_frame_num, i_global_qp );
1267 h->i_idr_pic_id = ( h->i_idr_pic_id + 1 ) % 65536;
1271 x264_slice_header_init( h, &h->sh, h->sps, h->pps, -1, h->i_frame_num, i_global_qp );
1273 /* always set the real higher num of ref frame used */
1274 h->sh.b_num_ref_idx_override = 1;
1275 h->sh.i_num_ref_idx_l0_active = h->i_ref0 <= 0 ? 1 : h->i_ref0;
1276 h->sh.i_num_ref_idx_l1_active = h->i_ref1 <= 0 ? 1 : h->i_ref1;
1279 h->fdec->i_frame_num = h->sh.i_frame_num;
1281 if( h->sps->i_poc_type == 0 )
1283 h->sh.i_poc_lsb = h->fdec->i_poc & ( (1 << h->sps->i_log2_max_poc_lsb) - 1 );
1284 h->sh.i_delta_poc_bottom = 0; /* XXX won't work for field */
1286 else if( h->sps->i_poc_type == 1 )
1288 /* FIXME TODO FIXME */
1292 /* Nothing to do ? */
1295 x264_macroblock_slice_init( h );
1298 static int x264_slice_write( x264_t *h )
1301 int mb_xy, i_mb_x, i_mb_y;
1302 int i, i_list, i_ref, i_skip_bak = 0; /* Shut up GCC. */
1304 x264_cabac_t cabac_bak;
1305 uint8_t cabac_prevbyte_bak = 0; /* Shut up GCC. */
1306 /* Assume no more than 3 bytes of NALU escaping. */
1307 int slice_max_size = h->param.i_slice_max_size > 0 ? (h->param.i_slice_max_size-3-NALU_OVERHEAD)*8 : INT_MAX;
1308 int starting_bits = bs_pos(&h->out.bs);
1311 x264_nal_start( h, h->i_nal_type, h->i_nal_ref_idc );
1314 x264_slice_header_write( &h->out.bs, &h->sh, h->i_nal_ref_idc );
1315 if( h->param.b_cabac )
1317 /* alignment needed */
1318 bs_align_1( &h->out.bs );
1321 x264_cabac_context_init( &h->cabac, h->sh.i_type, h->sh.i_qp, h->sh.i_cabac_init_idc );
1322 x264_cabac_encode_init ( &h->cabac, h->out.bs.p, h->out.bs.p_end );
1324 h->mb.i_last_qp = h->sh.i_qp;
1325 h->mb.i_last_dqp = 0;
1327 i_mb_y = h->sh.i_first_mb / h->sps->i_mb_width;
1328 i_mb_x = h->sh.i_first_mb % h->sps->i_mb_width;
1331 while( (mb_xy = i_mb_x + i_mb_y * h->sps->i_mb_width) <= h->sh.i_last_mb )
1333 int mb_spos = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
1334 if( h->param.i_slice_max_size > 0 )
1336 /* We don't need the contexts because flushing the CABAC encoder has no context
1337 * dependency and macroblocks are only re-encoded in the case where a slice is
1338 * ended (and thus the content of all contexts are thrown away). */
1339 if( h->param.b_cabac )
1341 memcpy( &cabac_bak, &h->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
1342 /* x264's CABAC writer modifies the previous byte during carry, so it has to be
1344 cabac_prevbyte_bak = h->cabac.p[-1];
1349 i_skip_bak = i_skip;
1353 if( i_mb_x == 0 && !h->mb.b_reencode_mb )
1354 x264_fdec_filter_row( h, i_mb_y );
1357 x264_macroblock_cache_load( h, i_mb_x, i_mb_y );
1359 x264_macroblock_analyse( h );
1361 /* encode this macroblock -> be careful it can change the mb type to P_SKIP if needed */
1362 x264_macroblock_encode( h );
1364 if( x264_bitstream_check_buffer( h ) )
1367 if( h->param.b_cabac )
1369 if( mb_xy > h->sh.i_first_mb && !(h->sh.b_mbaff && (i_mb_y&1)) )
1370 x264_cabac_encode_terminal( &h->cabac );
1372 if( IS_SKIP( h->mb.i_type ) )
1373 x264_cabac_mb_skip( h, 1 );
1376 if( h->sh.i_type != SLICE_TYPE_I )
1377 x264_cabac_mb_skip( h, 0 );
1378 x264_macroblock_write_cabac( h, &h->cabac );
1383 if( IS_SKIP( h->mb.i_type ) )
1387 if( h->sh.i_type != SLICE_TYPE_I )
1389 bs_write_ue( &h->out.bs, i_skip ); /* skip run */
1392 x264_macroblock_write_cavlc( h, &h->out.bs );
1396 int total_bits = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
1397 int mb_size = total_bits - mb_spos;
1399 /* We'll just re-encode this last macroblock if we go over the max slice size. */
1400 if( total_bits - starting_bits > slice_max_size && !h->mb.b_reencode_mb )
1402 if( mb_xy != h->sh.i_first_mb )
1404 if( h->param.b_cabac )
1406 memcpy( &h->cabac, &cabac_bak, offsetof(x264_cabac_t, f8_bits_encoded) );
1407 h->cabac.p[-1] = cabac_prevbyte_bak;
1412 i_skip = i_skip_bak;
1414 h->mb.b_reencode_mb = 1;
1415 h->sh.i_last_mb = mb_xy-1;
1420 h->sh.i_last_mb = mb_xy;
1421 h->mb.b_reencode_mb = 0;
1425 h->mb.b_reencode_mb = 0;
1428 if( h->param.b_visualize )
1429 x264_visualize_mb( h );
1433 x264_macroblock_cache_save( h );
1435 /* accumulate mb stats */
1436 h->stat.frame.i_mb_count[h->mb.i_type]++;
1438 if( !IS_INTRA(h->mb.i_type) && !IS_SKIP(h->mb.i_type) && !IS_DIRECT(h->mb.i_type) )
1440 if( h->mb.i_partition != D_8x8 )
1441 h->stat.frame.i_mb_partition[h->mb.i_partition] += 4;
1443 for( i = 0; i < 4; i++ )
1444 h->stat.frame.i_mb_partition[h->mb.i_sub_partition[i]] ++;
1445 if( h->param.i_frame_reference > 1 )
1446 for( i_list = 0; i_list <= (h->sh.i_type == SLICE_TYPE_B); i_list++ )
1447 for( i = 0; i < 4; i++ )
1449 i_ref = h->mb.cache.ref[i_list][ x264_scan8[4*i] ];
1451 h->stat.frame.i_mb_count_ref[i_list][i_ref] ++;
1455 if( h->param.i_log_level >= X264_LOG_INFO )
1457 if( h->mb.i_cbp_luma || h->mb.i_cbp_chroma )
1459 int cbpsum = (h->mb.i_cbp_luma&1) + ((h->mb.i_cbp_luma>>1)&1)
1460 + ((h->mb.i_cbp_luma>>2)&1) + (h->mb.i_cbp_luma>>3);
1461 int b_intra = IS_INTRA(h->mb.i_type);
1462 h->stat.frame.i_mb_cbp[!b_intra + 0] += cbpsum;
1463 h->stat.frame.i_mb_cbp[!b_intra + 2] += h->mb.i_cbp_chroma >= 1;
1464 h->stat.frame.i_mb_cbp[!b_intra + 4] += h->mb.i_cbp_chroma == 2;
1466 if( h->mb.i_cbp_luma && !IS_INTRA(h->mb.i_type) )
1468 h->stat.frame.i_mb_count_8x8dct[0] ++;
1469 h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
1471 if( IS_INTRA(h->mb.i_type) && h->mb.i_type != I_PCM )
1473 if( h->mb.i_type == I_16x16 )
1474 h->stat.frame.i_mb_pred_mode[0][h->mb.i_intra16x16_pred_mode]++;
1475 else if( h->mb.i_type == I_8x8 )
1476 for( i = 0; i < 16; i += 4 )
1477 h->stat.frame.i_mb_pred_mode[1][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
1478 else //if( h->mb.i_type == I_4x4 )
1479 for( i = 0; i < 16; i++ )
1480 h->stat.frame.i_mb_pred_mode[2][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
1484 x264_ratecontrol_mb( h, mb_size );
1488 i_mb_x += i_mb_y & 1;
1489 i_mb_y ^= i_mb_x < h->sps->i_mb_width;
1493 if( i_mb_x == h->sps->i_mb_width )
1500 if( h->param.b_cabac )
1502 x264_cabac_encode_flush( h, &h->cabac );
1503 h->out.bs.p = h->cabac.p;
1508 bs_write_ue( &h->out.bs, i_skip ); /* last skip run */
1509 /* rbsp_slice_trailing_bits */
1510 bs_rbsp_trailing( &h->out.bs );
1511 bs_flush( &h->out.bs );
1513 if( x264_nal_end( h ) )
1516 if( h->sh.i_last_mb == h->mb.i_mb_count-1 )
1518 h->stat.frame.i_misc_bits = bs_pos( &h->out.bs )
1519 + (h->out.i_nal*NALU_OVERHEAD * 8)
1520 - h->stat.frame.i_tex_bits
1521 - h->stat.frame.i_mv_bits;
1522 x264_fdec_filter_row( h, h->sps->i_mb_height );
1528 static void x264_thread_sync_context( x264_t *dst, x264_t *src )
1534 // reference counting
1535 for( f = src->frames.reference; *f; f++ )
1536 (*f)->i_reference_count++;
1537 for( f = dst->frames.reference; *f; f++ )
1538 x264_frame_push_unused( src, *f );
1539 src->fdec->i_reference_count++;
1540 x264_frame_push_unused( src, dst->fdec );
1542 // copy everything except the per-thread pointers and the constants.
1543 memcpy( &dst->i_frame, &src->i_frame, offsetof(x264_t, mb.type) - offsetof(x264_t, i_frame) );
1544 dst->param = src->param;
1545 dst->stat = src->stat;
1548 static void x264_thread_sync_stat( x264_t *dst, x264_t *src )
1552 memcpy( &dst->stat.i_frame_count, &src->stat.i_frame_count, sizeof(dst->stat) - sizeof(dst->stat.frame) );
1555 static void *x264_slices_write( x264_t *h )
1557 int i_slice_num = 0;
1558 if( h->param.i_sync_lookahead )
1559 x264_lower_thread_priority( 10 );
1562 /* Misalign mask has to be set separately for each thread. */
1563 if( h->param.cpu&X264_CPU_SSE_MISALIGN )
1564 x264_cpu_mask_misalign_sse();
1568 if( h->param.b_visualize )
1569 if( x264_visualize_init( h ) )
1574 memset( &h->stat.frame, 0, sizeof(h->stat.frame) );
1575 h->mb.b_reencode_mb = 0;
1576 while( h->sh.i_first_mb < h->mb.i_mb_count )
1578 h->sh.i_last_mb = h->mb.i_mb_count - 1;
1579 if( h->param.i_slice_max_mbs )
1580 h->sh.i_last_mb = h->sh.i_first_mb + h->param.i_slice_max_mbs - 1;
1581 else if( h->param.i_slice_count )
1585 double height = h->sps->i_mb_height >> h->param.b_interlaced;
1586 int width = h->sps->i_mb_width << h->param.b_interlaced;
1587 h->sh.i_last_mb = (int)(height * i_slice_num / h->param.i_slice_count + 0.5) * width - 1;
1589 h->sh.i_last_mb = X264_MIN( h->sh.i_last_mb, h->mb.i_mb_count - 1 );
1590 if( x264_stack_align( x264_slice_write, h ) )
1592 h->sh.i_first_mb = h->sh.i_last_mb + 1;
1596 if( h->param.b_visualize )
1598 x264_visualize_show( h );
1599 x264_visualize_close( h );
1606 /****************************************************************************
1607 * x264_encoder_encode:
1608 * XXX: i_poc : is the poc of the current given picture
1609 * i_frame : is the number of the frame being coded
1610 * ex: type frame poc
1618 ****************************************************************************/
1619 int x264_encoder_encode( x264_t *h,
1620 x264_nal_t **pp_nal, int *pi_nal,
1621 x264_picture_t *pic_in,
1622 x264_picture_t *pic_out )
1624 x264_t *thread_current, *thread_prev, *thread_oldest;
1630 if( h->param.i_threads > 1)
1632 thread_prev = h->thread[ h->i_thread_phase ];
1633 h->i_thread_phase = (h->i_thread_phase + 1) % h->param.i_threads;
1634 thread_current = h->thread[ h->i_thread_phase ];
1635 thread_oldest = h->thread[ (h->i_thread_phase + 1) % h->param.i_threads ];
1636 x264_thread_sync_context( thread_current, thread_prev );
1637 x264_thread_sync_ratecontrol( thread_current, thread_prev, thread_oldest );
1639 // fprintf(stderr, "current: %p prev: %p oldest: %p \n", thread_current, thread_prev, thread_oldest);
1647 // ok to call this before encoding any frames, since the initial values of fdec have b_kept_as_ref=0
1648 if( x264_reference_update( h ) )
1650 h->fdec->i_lines_completed = -1;
1656 /* ------------------- Setup new frame from picture -------------------- */
1657 if( pic_in != NULL )
1659 /* 1: Copy the picture to a frame and move it to a buffer */
1660 x264_frame_t *fenc = x264_frame_pop_unused( h, 0 );
1664 if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 )
1667 if( h->param.i_width != 16 * h->sps->i_mb_width ||
1668 h->param.i_height != 16 * h->sps->i_mb_height )
1669 x264_frame_expand_border_mod16( h, fenc );
1671 fenc->i_frame = h->frames.i_input++;
1673 if( h->frames.b_have_lowres )
1674 x264_frame_init_lowres( h, fenc );
1676 if( h->param.rc.b_mb_tree && h->param.rc.b_stat_read )
1678 if( x264_macroblock_tree_read( h, fenc ) )
1681 else if( h->param.rc.i_aq_mode )
1682 x264_adaptive_quant_frame( h, fenc );
1684 /* 2: Place the frame into the queue for its slice type decision */
1685 x264_lookahead_put_frame( h, fenc );
1687 if( h->frames.i_input <= h->frames.i_delay + 1 - h->param.i_threads )
1689 /* Nothing yet to encode, waiting for filling of buffers */
1690 pic_out->i_type = X264_TYPE_AUTO;
1696 /* signal kills for lookahead thread */
1697 h->lookahead->b_exit_thread = 1;
1698 x264_pthread_cond_broadcast( &h->lookahead->ifbuf.cv_fill );
1702 /* 3: The picture is analyzed in the lookahead */
1703 if( !h->frames.current[0] )
1704 x264_lookahead_get_frames( h );
1706 if( !h->frames.current[0] && x264_lookahead_is_empty( h ) )
1707 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
1709 /* ------------------- Get frame to be encoded ------------------------- */
1710 /* 4: get picture to encode */
1711 h->fenc = x264_frame_shift( h->frames.current );
1712 if( h->fenc->param )
1714 x264_encoder_reconfig( h, h->fenc->param );
1715 if( h->fenc->param->param_free )
1716 h->fenc->param->param_free( h->fenc->param );
1719 if( h->fenc->i_type == X264_TYPE_IDR )
1721 h->frames.i_last_idr = h->fenc->i_frame;
1725 /* ------------------- Setup frame context ----------------------------- */
1726 /* 5: Init data dependent of frame type */
1727 if( h->fenc->i_type == X264_TYPE_IDR )
1729 /* reset ref pictures */
1730 x264_reference_reset( h );
1732 i_nal_type = NAL_SLICE_IDR;
1733 i_nal_ref_idc = NAL_PRIORITY_HIGHEST;
1734 h->sh.i_type = SLICE_TYPE_I;
1736 else if( h->fenc->i_type == X264_TYPE_I )
1738 i_nal_type = NAL_SLICE;
1739 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
1740 h->sh.i_type = SLICE_TYPE_I;
1742 else if( h->fenc->i_type == X264_TYPE_P )
1744 i_nal_type = NAL_SLICE;
1745 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
1746 h->sh.i_type = SLICE_TYPE_P;
1748 else if( h->fenc->i_type == X264_TYPE_BREF )
1750 i_nal_type = NAL_SLICE;
1751 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* maybe add MMCO to forget it? -> low */
1752 h->sh.i_type = SLICE_TYPE_B;
1756 i_nal_type = NAL_SLICE;
1757 i_nal_ref_idc = NAL_PRIORITY_DISPOSABLE;
1758 h->sh.i_type = SLICE_TYPE_B;
1762 h->fenc->i_poc = 2 * (h->fenc->i_frame - h->frames.i_last_idr);
1763 h->fdec->i_type = h->fenc->i_type;
1764 h->fdec->i_frame = h->fenc->i_frame;
1765 h->fenc->b_kept_as_ref =
1766 h->fdec->b_kept_as_ref = i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE && h->param.i_keyint_max > 1;
1770 /* ------------------- Init ----------------------------- */
1771 /* build ref list 0/1 */
1772 x264_reference_build_list( h, h->fdec->i_poc );
1774 /* ---------------------- Write the bitstream -------------------------- */
1775 /* Init bitstream context */
1777 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
1779 if( h->param.b_aud )
1783 if( h->sh.i_type == SLICE_TYPE_I )
1785 else if( h->sh.i_type == SLICE_TYPE_P )
1787 else if( h->sh.i_type == SLICE_TYPE_B )
1792 x264_nal_start( h, NAL_AUD, NAL_PRIORITY_DISPOSABLE );
1793 bs_write( &h->out.bs, 3, pic_type );
1794 bs_rbsp_trailing( &h->out.bs );
1795 if( x264_nal_end( h ) )
1799 h->i_nal_type = i_nal_type;
1800 h->i_nal_ref_idc = i_nal_ref_idc;
1802 int overhead = NALU_OVERHEAD;
1804 /* Write SPS and PPS */
1805 if( i_nal_type == NAL_SLICE_IDR && h->param.b_repeat_headers )
1807 if( h->fenc->i_frame == 0 )
1809 /* identify ourself */
1810 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
1811 if( x264_sei_version_write( h, &h->out.bs ) )
1813 if( x264_nal_end( h ) )
1815 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
1818 /* generate sequence parameters */
1819 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
1820 x264_sps_write( &h->out.bs, h->sps );
1821 if( x264_nal_end( h ) )
1823 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
1825 /* generate picture parameters */
1826 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
1827 x264_pps_write( &h->out.bs, h->pps );
1828 if( x264_nal_end( h ) )
1830 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
1833 /* Init the rate control */
1834 /* FIXME: Include slice header bit cost. */
1835 x264_ratecontrol_start( h, h->fenc->i_qpplus1, overhead*8 );
1836 i_global_qp = x264_ratecontrol_qp( h );
1838 pic_out->i_qpplus1 =
1839 h->fdec->i_qpplus1 = i_global_qp + 1;
1841 if( h->param.rc.b_stat_read && h->sh.i_type != SLICE_TYPE_I )
1842 x264_reference_build_list_optimal( h );
1844 /* Check to see whether we have chosen a reference list ordering different
1845 * from the standard's default. */
1846 h->b_ref_reorder[0] =
1847 h->b_ref_reorder[1] = 0;
1848 for( i = 0; i < h->i_ref0 - 1; i++ )
1849 /* P and B-frames use different default orders. */
1850 if( h->sh.i_type == SLICE_TYPE_P ? h->fref0[i]->i_frame_num < h->fref0[i+1]->i_frame_num
1851 : h->fref0[i]->i_poc < h->fref0[i+1]->i_poc )
1853 h->b_ref_reorder[0] = 1;
1857 if( h->sh.i_type == SLICE_TYPE_B )
1858 x264_macroblock_bipred_init( h );
1860 /* ------------------------ Create slice header ----------------------- */
1861 x264_slice_init( h, i_nal_type, i_global_qp );
1863 if( i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE )
1867 if( h->param.i_threads > 1 )
1869 if( x264_pthread_create( &h->thread_handle, NULL, (void*)x264_slices_write, h ) )
1871 h->b_thread_active = 1;
1874 if( (intptr_t)x264_slices_write( h ) )
1877 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
1880 static int x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
1881 x264_nal_t **pp_nal, int *pi_nal,
1882 x264_picture_t *pic_out )
1884 int i, j, i_list, frame_size;
1885 char psz_message[80];
1887 if( h->b_thread_active )
1890 x264_pthread_join( h->thread_handle, &ret );
1892 return (intptr_t)ret;
1893 h->b_thread_active = 0;
1897 pic_out->i_type = X264_TYPE_AUTO;
1901 x264_frame_push_unused( thread_current, h->fenc );
1903 /* End bitstream, set output */
1904 *pi_nal = h->out.i_nal;
1905 *pp_nal = h->out.nal;
1907 frame_size = x264_encoder_encapsulate_nals( h );
1911 /* Set output picture properties */
1912 if( h->sh.i_type == SLICE_TYPE_I )
1913 pic_out->i_type = h->i_nal_type == NAL_SLICE_IDR ? X264_TYPE_IDR : X264_TYPE_I;
1914 else if( h->sh.i_type == SLICE_TYPE_P )
1915 pic_out->i_type = X264_TYPE_P;
1917 pic_out->i_type = X264_TYPE_B;
1918 pic_out->i_pts = h->fenc->i_pts;
1920 pic_out->img.i_plane = h->fdec->i_plane;
1921 for(i = 0; i < 3; i++)
1923 pic_out->img.i_stride[i] = h->fdec->i_stride[i];
1924 pic_out->img.plane[i] = h->fdec->plane[i];
1927 /* ---------------------- Update encoder state ------------------------- */
1931 if( x264_ratecontrol_end( h, frame_size * 8 ) < 0 )
1934 x264_noise_reduction_update( thread_current );
1936 /* ---------------------- Compute/Print statistics --------------------- */
1937 x264_thread_sync_stat( h, h->thread[0] );
1940 h->stat.i_frame_count[h->sh.i_type]++;
1941 h->stat.i_frame_size[h->sh.i_type] += frame_size;
1942 h->stat.f_frame_qp[h->sh.i_type] += h->fdec->f_qp_avg_aq;
1944 for( i = 0; i < X264_MBTYPE_MAX; i++ )
1945 h->stat.i_mb_count[h->sh.i_type][i] += h->stat.frame.i_mb_count[i];
1946 for( i = 0; i < X264_PARTTYPE_MAX; i++ )
1947 h->stat.i_mb_partition[h->sh.i_type][i] += h->stat.frame.i_mb_partition[i];
1948 for( i = 0; i < 2; i++ )
1949 h->stat.i_mb_count_8x8dct[i] += h->stat.frame.i_mb_count_8x8dct[i];
1950 for( i = 0; i < 6; i++ )
1951 h->stat.i_mb_cbp[i] += h->stat.frame.i_mb_cbp[i];
1952 for( i = 0; i < 3; i++ )
1953 for( j = 0; j < 13; j++ )
1954 h->stat.i_mb_pred_mode[i][j] += h->stat.frame.i_mb_pred_mode[i][j];
1955 if( h->sh.i_type != SLICE_TYPE_I )
1956 for( i_list = 0; i_list < 2; i_list++ )
1957 for( i = 0; i < 32; i++ )
1958 h->stat.i_mb_count_ref[h->sh.i_type][i_list][i] += h->stat.frame.i_mb_count_ref[i_list][i];
1959 if( h->sh.i_type == SLICE_TYPE_P )
1960 h->stat.i_consecutive_bframes[h->fdec->i_frame - h->fref0[0]->i_frame - 1]++;
1961 if( h->sh.i_type == SLICE_TYPE_B )
1963 h->stat.i_direct_frames[ h->sh.b_direct_spatial_mv_pred ] ++;
1964 if( h->mb.b_direct_auto_write )
1966 //FIXME somewhat arbitrary time constants
1967 if( h->stat.i_direct_score[0] + h->stat.i_direct_score[1] > h->mb.i_mb_count )
1969 for( i = 0; i < 2; i++ )
1970 h->stat.i_direct_score[i] = h->stat.i_direct_score[i] * 9/10;
1972 for( i = 0; i < 2; i++ )
1973 h->stat.i_direct_score[i] += h->stat.frame.i_direct_score[i];
1977 psz_message[0] = '\0';
1978 if( h->param.analyse.b_psnr )
1981 h->stat.frame.i_ssd[0],
1982 h->stat.frame.i_ssd[1],
1983 h->stat.frame.i_ssd[2],
1986 h->stat.i_ssd_global[h->sh.i_type] += ssd[0] + ssd[1] + ssd[2];
1987 h->stat.f_psnr_average[h->sh.i_type] += x264_psnr( ssd[0] + ssd[1] + ssd[2], 3 * h->param.i_width * h->param.i_height / 2 );
1988 h->stat.f_psnr_mean_y[h->sh.i_type] += x264_psnr( ssd[0], h->param.i_width * h->param.i_height );
1989 h->stat.f_psnr_mean_u[h->sh.i_type] += x264_psnr( ssd[1], h->param.i_width * h->param.i_height / 4 );
1990 h->stat.f_psnr_mean_v[h->sh.i_type] += x264_psnr( ssd[2], h->param.i_width * h->param.i_height / 4 );
1992 snprintf( psz_message, 80, " PSNR Y:%5.2f U:%5.2f V:%5.2f",
1993 x264_psnr( ssd[0], h->param.i_width * h->param.i_height ),
1994 x264_psnr( ssd[1], h->param.i_width * h->param.i_height / 4),
1995 x264_psnr( ssd[2], h->param.i_width * h->param.i_height / 4) );
1998 if( h->param.analyse.b_ssim )
2000 double ssim_y = h->stat.frame.f_ssim
2001 / (((h->param.i_width-6)>>2) * ((h->param.i_height-6)>>2));
2002 h->stat.f_ssim_mean_y[h->sh.i_type] += ssim_y;
2003 snprintf( psz_message + strlen(psz_message), 80 - strlen(psz_message),
2004 " SSIM Y:%.5f", ssim_y );
2006 psz_message[79] = '\0';
2008 x264_log( h, X264_LOG_DEBUG,
2009 "frame=%4d QP=%.2f NAL=%d Slice:%c Poc:%-3d I:%-4d P:%-4d SKIP:%-4d size=%d bytes%s\n",
2011 h->fdec->f_qp_avg_aq,
2013 h->sh.i_type == SLICE_TYPE_I ? 'I' : (h->sh.i_type == SLICE_TYPE_P ? 'P' : 'B' ),
2015 h->stat.frame.i_mb_count_i,
2016 h->stat.frame.i_mb_count_p,
2017 h->stat.frame.i_mb_count_skip,
2021 // keep stats all in one place
2022 x264_thread_sync_stat( h->thread[0], h );
2023 // for the use of the next frame
2024 x264_thread_sync_stat( thread_current, h );
2026 #ifdef DEBUG_MB_TYPE
2028 static const char mb_chars[] = { 'i', 'i', 'I', 'C', 'P', '8', 'S',
2029 'D', '<', 'X', 'B', 'X', '>', 'B', 'B', 'B', 'B', '8', 'S' };
2031 for( mb_xy = 0; mb_xy < h->sps->i_mb_width * h->sps->i_mb_height; mb_xy++ )
2033 if( h->mb.type[mb_xy] < X264_MBTYPE_MAX && h->mb.type[mb_xy] >= 0 )
2034 fprintf( stderr, "%c ", mb_chars[ h->mb.type[mb_xy] ] );
2036 fprintf( stderr, "? " );
2038 if( (mb_xy+1) % h->sps->i_mb_width == 0 )
2039 fprintf( stderr, "\n" );
2044 if( h->param.psz_dump_yuv )
2045 x264_frame_dump( h );
2050 static void x264_print_intra( int64_t *i_mb_count, double i_count, int b_print_pcm, char *intra )
2052 intra += sprintf( intra, "I16..4%s: %4.1f%% %4.1f%% %4.1f%%",
2053 b_print_pcm ? "..PCM" : "",
2054 i_mb_count[I_16x16]/ i_count,
2055 i_mb_count[I_8x8] / i_count,
2056 i_mb_count[I_4x4] / i_count );
2058 sprintf( intra, " %4.1f%%", i_mb_count[I_PCM] / i_count );
2061 /****************************************************************************
2062 * x264_encoder_close:
2063 ****************************************************************************/
2064 void x264_encoder_close ( x264_t *h )
2066 int64_t i_yuv_size = 3 * h->param.i_width * h->param.i_height / 2;
2067 int64_t i_mb_count_size[2][7] = {{0}};
2069 int i, j, i_list, i_type;
2070 int b_print_pcm = h->stat.i_mb_count[SLICE_TYPE_I][I_PCM]
2071 || h->stat.i_mb_count[SLICE_TYPE_P][I_PCM]
2072 || h->stat.i_mb_count[SLICE_TYPE_B][I_PCM];
2074 x264_lookahead_delete( h );
2076 for( i=0; i<h->param.i_threads; i++ )
2078 // don't strictly have to wait for the other threads, but it's simpler than canceling them
2079 if( h->thread[i]->b_thread_active )
2081 x264_pthread_join( h->thread[i]->thread_handle, NULL );
2082 assert( h->thread[i]->fenc->i_reference_count == 1 );
2083 x264_frame_delete( h->thread[i]->fenc );
2087 if( h->param.i_threads > 1 )
2089 x264_t *thread_prev;
2091 thread_prev = h->thread[h->i_thread_phase];
2092 x264_thread_sync_ratecontrol( h, thread_prev, h );
2093 x264_thread_sync_ratecontrol( thread_prev, thread_prev, h );
2094 h->i_frame = thread_prev->i_frame + 1 - h->param.i_threads;
2098 /* Slices used and PSNR */
2099 for( i=0; i<5; i++ )
2101 static const int slice_order[] = { SLICE_TYPE_I, SLICE_TYPE_SI, SLICE_TYPE_P, SLICE_TYPE_SP, SLICE_TYPE_B };
2102 static const char *slice_name[] = { "P", "B", "I", "SP", "SI" };
2103 int i_slice = slice_order[i];
2105 if( h->stat.i_frame_count[i_slice] > 0 )
2107 const int i_count = h->stat.i_frame_count[i_slice];
2108 if( h->param.analyse.b_psnr )
2110 x264_log( h, X264_LOG_INFO,
2111 "frame %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",
2112 slice_name[i_slice],
2114 h->stat.f_frame_qp[i_slice] / i_count,
2115 (double)h->stat.i_frame_size[i_slice] / i_count,
2116 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,
2117 h->stat.f_psnr_average[i_slice] / i_count,
2118 x264_psnr( h->stat.i_ssd_global[i_slice], i_count * i_yuv_size ) );
2122 x264_log( h, X264_LOG_INFO,
2123 "frame %s:%-5d Avg QP:%5.2f size:%6.0f\n",
2124 slice_name[i_slice],
2126 h->stat.f_frame_qp[i_slice] / i_count,
2127 (double)h->stat.i_frame_size[i_slice] / i_count );
2131 if( h->param.i_bframe && h->stat.i_frame_count[SLICE_TYPE_P] )
2135 // weight by number of frames (including the P-frame) that are in a sequence of N B-frames
2136 for( i=0; i<=h->param.i_bframe; i++ )
2137 den += (i+1) * h->stat.i_consecutive_bframes[i];
2138 for( i=0; i<=h->param.i_bframe; i++ )
2139 p += sprintf( p, " %4.1f%%", 100. * (i+1) * h->stat.i_consecutive_bframes[i] / den );
2140 x264_log( h, X264_LOG_INFO, "consecutive B-frames:%s\n", buf );
2143 for( i_type = 0; i_type < 2; i_type++ )
2144 for( i = 0; i < X264_PARTTYPE_MAX; i++ )
2146 if( i == D_DIRECT_8x8 ) continue; /* direct is counted as its own type */
2147 i_mb_count_size[i_type][x264_mb_partition_pixel_table[i]] += h->stat.i_mb_partition[i_type][i];
2151 if( h->stat.i_frame_count[SLICE_TYPE_I] > 0 )
2153 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_I];
2154 double i_count = h->stat.i_frame_count[SLICE_TYPE_I] * h->mb.i_mb_count / 100.0;
2155 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
2156 x264_log( h, X264_LOG_INFO, "mb I %s\n", buf );
2158 if( h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
2160 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_P];
2161 double i_count = h->stat.i_frame_count[SLICE_TYPE_P] * h->mb.i_mb_count / 100.0;
2162 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_P];
2163 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
2164 x264_log( h, X264_LOG_INFO,
2165 "mb P %s P16..4: %4.1f%% %4.1f%% %4.1f%% %4.1f%% %4.1f%% skip:%4.1f%%\n",
2167 i_mb_size[PIXEL_16x16] / (i_count*4),
2168 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
2169 i_mb_size[PIXEL_8x8] / (i_count*4),
2170 (i_mb_size[PIXEL_8x4] + i_mb_size[PIXEL_4x8]) / (i_count*4),
2171 i_mb_size[PIXEL_4x4] / (i_count*4),
2172 i_mb_count[P_SKIP] / i_count );
2174 if( h->stat.i_frame_count[SLICE_TYPE_B] > 0 )
2176 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_B];
2177 double i_count = h->stat.i_frame_count[SLICE_TYPE_B] * h->mb.i_mb_count / 100.0;
2178 double i_mb_list_count;
2179 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_B];
2180 int64_t list_count[3] = {0}; /* 0 == L0, 1 == L1, 2 == BI */
2181 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
2182 for( i = 0; i < X264_PARTTYPE_MAX; i++ )
2183 for( j = 0; j < 2; j++ )
2185 int l0 = x264_mb_type_list_table[i][0][j];
2186 int l1 = x264_mb_type_list_table[i][1][j];
2188 list_count[l1+l0*l1] += h->stat.i_mb_count[SLICE_TYPE_B][i] * 2;
2190 list_count[0] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L0_8x8];
2191 list_count[1] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L1_8x8];
2192 list_count[2] += h->stat.i_mb_partition[SLICE_TYPE_B][D_BI_8x8];
2193 i_mb_count[B_DIRECT] += (h->stat.i_mb_partition[SLICE_TYPE_B][D_DIRECT_8x8]+2)/4;
2194 i_mb_list_count = (list_count[0] + list_count[1] + list_count[2]) / 100.0;
2195 x264_log( h, X264_LOG_INFO,
2196 "mb B %s B16..8: %4.1f%% %4.1f%% %4.1f%% direct:%4.1f%% skip:%4.1f%% L0:%4.1f%% L1:%4.1f%% BI:%4.1f%%\n",
2198 i_mb_size[PIXEL_16x16] / (i_count*4),
2199 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
2200 i_mb_size[PIXEL_8x8] / (i_count*4),
2201 i_mb_count[B_DIRECT] / i_count,
2202 i_mb_count[B_SKIP] / i_count,
2203 list_count[0] / i_mb_list_count,
2204 list_count[1] / i_mb_list_count,
2205 list_count[2] / i_mb_list_count );
2208 x264_ratecontrol_summary( h );
2210 if( h->stat.i_frame_count[SLICE_TYPE_I] + h->stat.i_frame_count[SLICE_TYPE_P] + h->stat.i_frame_count[SLICE_TYPE_B] > 0 )
2212 #define SUM3(p) (p[SLICE_TYPE_I] + p[SLICE_TYPE_P] + p[SLICE_TYPE_B])
2213 #define SUM3b(p,o) (p[SLICE_TYPE_I][o] + p[SLICE_TYPE_P][o] + p[SLICE_TYPE_B][o])
2214 int64_t i_i8x8 = SUM3b( h->stat.i_mb_count, I_8x8 );
2215 int64_t i_intra = i_i8x8 + SUM3b( h->stat.i_mb_count, I_4x4 )
2216 + SUM3b( h->stat.i_mb_count, I_16x16 );
2217 int64_t i_all_intra = i_intra + SUM3b( h->stat.i_mb_count, I_PCM);
2218 const int i_count = h->stat.i_frame_count[SLICE_TYPE_I] +
2219 h->stat.i_frame_count[SLICE_TYPE_P] +
2220 h->stat.i_frame_count[SLICE_TYPE_B];
2221 int64_t i_mb_count = i_count * h->mb.i_mb_count;
2222 float fps = (float) h->param.i_fps_num / h->param.i_fps_den;
2223 float f_bitrate = fps * SUM3(h->stat.i_frame_size) / i_count / 125;
2225 if( h->pps->b_transform_8x8_mode )
2228 if( h->stat.i_mb_count_8x8dct[0] )
2229 sprintf( buf, " inter:%.1f%%", 100. * h->stat.i_mb_count_8x8dct[1] / h->stat.i_mb_count_8x8dct[0] );
2230 x264_log( h, X264_LOG_INFO, "8x8 transform intra:%.1f%%%s\n", 100. * i_i8x8 / i_intra, buf );
2233 if( h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO
2234 && h->stat.i_frame_count[SLICE_TYPE_B] )
2236 x264_log( h, X264_LOG_INFO, "direct mvs spatial:%.1f%% temporal:%.1f%%\n",
2237 h->stat.i_direct_frames[1] * 100. / h->stat.i_frame_count[SLICE_TYPE_B],
2238 h->stat.i_direct_frames[0] * 100. / h->stat.i_frame_count[SLICE_TYPE_B] );
2242 if( i_mb_count != i_all_intra )
2243 sprintf( buf, " inter: %.1f%% %.1f%% %.1f%%",
2244 h->stat.i_mb_cbp[1] * 100.0 / ((i_mb_count - i_all_intra)*4),
2245 h->stat.i_mb_cbp[3] * 100.0 / ((i_mb_count - i_all_intra) ),
2246 h->stat.i_mb_cbp[5] * 100.0 / ((i_mb_count - i_all_intra)) );
2247 x264_log( h, X264_LOG_INFO, "coded y,uvDC,uvAC intra: %.1f%% %.1f%% %.1f%%%s\n",
2248 h->stat.i_mb_cbp[0] * 100.0 / (i_all_intra*4),
2249 h->stat.i_mb_cbp[2] * 100.0 / (i_all_intra ),
2250 h->stat.i_mb_cbp[4] * 100.0 / (i_all_intra ), buf );
2252 int64_t fixed_pred_modes[3][9] = {{0}};
2253 int64_t sum_pred_modes[3] = {0};
2254 for( i = 0; i <= I_PRED_16x16_DC_128; i++ )
2256 fixed_pred_modes[0][x264_mb_pred_mode16x16_fix[i]] += h->stat.i_mb_pred_mode[0][i];
2257 sum_pred_modes[0] += h->stat.i_mb_pred_mode[0][i];
2259 if( sum_pred_modes[0] )
2260 x264_log( h, X264_LOG_INFO, "i16 v,h,dc,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
2261 fixed_pred_modes[0][0] * 100.0 / sum_pred_modes[0],
2262 fixed_pred_modes[0][1] * 100.0 / sum_pred_modes[0],
2263 fixed_pred_modes[0][2] * 100.0 / sum_pred_modes[0],
2264 fixed_pred_modes[0][3] * 100.0 / sum_pred_modes[0] );
2265 for( i = 1; i <= 2; i++ )
2267 for( j = 0; j <= I_PRED_8x8_DC_128; j++ )
2269 fixed_pred_modes[i][x264_mb_pred_mode4x4_fix(j)] += h->stat.i_mb_pred_mode[i][j];
2270 sum_pred_modes[i] += h->stat.i_mb_pred_mode[i][j];
2272 if( sum_pred_modes[i] )
2273 x264_log( h, X264_LOG_INFO, "i%d v,h,dc,ddl,ddr,vr,hd,vl,hu: %2.0f%% %2.0f%% %2.0f%% %2.0f%% %2.0f%% %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n", (3-i)*4,
2274 fixed_pred_modes[i][0] * 100.0 / sum_pred_modes[i],
2275 fixed_pred_modes[i][1] * 100.0 / sum_pred_modes[i],
2276 fixed_pred_modes[i][2] * 100.0 / sum_pred_modes[i],
2277 fixed_pred_modes[i][3] * 100.0 / sum_pred_modes[i],
2278 fixed_pred_modes[i][4] * 100.0 / sum_pred_modes[i],
2279 fixed_pred_modes[i][5] * 100.0 / sum_pred_modes[i],
2280 fixed_pred_modes[i][6] * 100.0 / sum_pred_modes[i],
2281 fixed_pred_modes[i][7] * 100.0 / sum_pred_modes[i],
2282 fixed_pred_modes[i][8] * 100.0 / sum_pred_modes[i] );
2285 for( i_list = 0; i_list < 2; i_list++ )
2288 for( i_slice = 0; i_slice < 2; i_slice++ )
2293 for( i = 0; i < 32; i++ )
2294 if( h->stat.i_mb_count_ref[i_slice][i_list][i] )
2296 i_den += h->stat.i_mb_count_ref[i_slice][i_list][i];
2301 for( i = 0; i <= i_max; i++ )
2302 p += sprintf( p, " %4.1f%%", 100. * h->stat.i_mb_count_ref[i_slice][i_list][i] / i_den );
2303 x264_log( h, X264_LOG_INFO, "ref %c L%d:%s\n", "PB"[i_slice], i_list, buf );
2307 if( h->param.analyse.b_ssim )
2309 x264_log( h, X264_LOG_INFO,
2310 "SSIM Mean Y:%.7f\n",
2311 SUM3( h->stat.f_ssim_mean_y ) / i_count );
2313 if( h->param.analyse.b_psnr )
2315 x264_log( h, X264_LOG_INFO,
2316 "PSNR Mean Y:%6.3f U:%6.3f V:%6.3f Avg:%6.3f Global:%6.3f kb/s:%.2f\n",
2317 SUM3( h->stat.f_psnr_mean_y ) / i_count,
2318 SUM3( h->stat.f_psnr_mean_u ) / i_count,
2319 SUM3( h->stat.f_psnr_mean_v ) / i_count,
2320 SUM3( h->stat.f_psnr_average ) / i_count,
2321 x264_psnr( SUM3( h->stat.i_ssd_global ), i_count * i_yuv_size ),
2325 x264_log( h, X264_LOG_INFO, "kb/s:%.2f\n", f_bitrate );
2329 x264_ratecontrol_delete( h );
2332 if( h->param.rc.psz_stat_out )
2333 free( h->param.rc.psz_stat_out );
2334 if( h->param.rc.psz_stat_in )
2335 free( h->param.rc.psz_stat_in );
2337 x264_cqm_delete( h );
2339 x264_analyse_free_costs( h );
2341 if( h->param.i_threads > 1)
2342 h = h->thread[h->i_thread_phase];
2345 x264_frame_delete_list( h->frames.unused[0] );
2346 x264_frame_delete_list( h->frames.unused[1] );
2347 x264_frame_delete_list( h->frames.current );
2351 for( i = h->param.i_threads - 1; i >= 0; i-- )
2353 x264_frame_t **frame;
2355 for( frame = h->thread[i]->frames.reference; *frame; frame++ )
2357 assert( (*frame)->i_reference_count > 0 );
2358 (*frame)->i_reference_count--;
2359 if( (*frame)->i_reference_count == 0 )
2360 x264_frame_delete( *frame );
2362 frame = &h->thread[i]->fdec;
2363 assert( (*frame)->i_reference_count > 0 );
2364 (*frame)->i_reference_count--;
2365 if( (*frame)->i_reference_count == 0 )
2366 x264_frame_delete( *frame );
2368 x264_macroblock_cache_end( h->thread[i] );
2369 x264_free( h->thread[i]->out.p_bitstream );
2370 x264_free( h->thread[i]->out.nal);
2371 x264_free( h->thread[i] );
2375 /****************************************************************************
2376 * x264_encoder_delayed_frames:
2377 ****************************************************************************/
2378 int x264_encoder_delayed_frames( x264_t *h )
2380 int delayed_frames = 0;
2382 for( i=0; i<h->param.i_threads; i++ )
2383 delayed_frames += h->thread[i]->b_thread_active;
2384 h = h->thread[h->i_thread_phase];
2385 for( i=0; h->frames.current[i]; i++ )
2387 x264_pthread_mutex_lock( &h->lookahead->ofbuf.mutex );
2388 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
2389 x264_pthread_mutex_lock( &h->lookahead->next.mutex );
2390 delayed_frames += h->lookahead->ifbuf.i_size + h->lookahead->next.i_size + h->lookahead->ofbuf.i_size;
2391 x264_pthread_mutex_unlock( &h->lookahead->next.mutex );
2392 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
2393 x264_pthread_mutex_unlock( &h->lookahead->ofbuf.mutex );
2394 return delayed_frames;