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"
37 #include "common/visualize.h"
40 //#define DEBUG_MB_TYPE
42 #define bs_write_ue bs_write_ue_big
44 static int x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
45 x264_nal_t **pp_nal, int *pi_nal,
46 x264_picture_t *pic_out );
48 /****************************************************************************
50 ******************************* x264 libs **********************************
52 ****************************************************************************/
53 static float x264_psnr( int64_t i_sqe, int64_t i_size )
55 double f_mse = (double)i_sqe / ((double)65025.0 * (double)i_size);
56 if( f_mse <= 0.0000000001 ) /* Max 100dB */
59 return -10.0 * log10( f_mse );
62 static void x264_frame_dump( x264_t *h )
64 FILE *f = fopen( h->param.psz_dump_yuv, "r+b" );
67 /* Write the frame in display order */
68 fseek( f, (uint64_t)h->fdec->i_frame * h->param.i_height * h->param.i_width * 3/2, SEEK_SET );
69 for( int i = 0; i < h->fdec->i_plane; i++ )
70 for( int y = 0; y < h->param.i_height >> !!i; y++ )
71 fwrite( &h->fdec->plane[i][y*h->fdec->i_stride[i]], 1, h->param.i_width >> !!i, f );
76 /* Fill "default" values */
77 static void x264_slice_header_init( x264_t *h, x264_slice_header_t *sh,
78 x264_sps_t *sps, x264_pps_t *pps,
79 int i_idr_pic_id, int i_frame, int i_qp )
81 x264_param_t *param = &h->param;
83 /* First we fill all fields */
88 sh->i_last_mb = h->mb.i_mb_count - 1;
89 sh->i_pps_id = pps->i_id;
91 sh->i_frame_num = i_frame;
93 sh->b_mbaff = h->param.b_interlaced;
94 sh->b_field_pic = 0; /* no field support for now */
95 sh->b_bottom_field = 0; /* not yet used */
97 sh->i_idr_pic_id = i_idr_pic_id;
99 /* poc stuff, fixed later */
101 sh->i_delta_poc_bottom = 0;
102 sh->i_delta_poc[0] = 0;
103 sh->i_delta_poc[1] = 0;
105 sh->i_redundant_pic_cnt = 0;
107 h->mb.b_direct_auto_write = h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO
109 && ( h->param.rc.b_stat_write || !h->param.rc.b_stat_read );
111 if( !h->mb.b_direct_auto_read && sh->i_type == SLICE_TYPE_B )
113 if( h->fref1[0]->i_poc_l0ref0 == h->fref0[0]->i_poc )
115 if( h->mb.b_direct_auto_write )
116 sh->b_direct_spatial_mv_pred = ( h->stat.i_direct_score[1] > h->stat.i_direct_score[0] );
118 sh->b_direct_spatial_mv_pred = ( param->analyse.i_direct_mv_pred == X264_DIRECT_PRED_SPATIAL );
122 h->mb.b_direct_auto_write = 0;
123 sh->b_direct_spatial_mv_pred = 1;
126 /* else b_direct_spatial_mv_pred was read from the 2pass statsfile */
128 sh->b_num_ref_idx_override = 0;
129 sh->i_num_ref_idx_l0_active = 1;
130 sh->i_num_ref_idx_l1_active = 1;
132 sh->b_ref_pic_list_reordering_l0 = h->b_ref_reorder[0];
133 sh->b_ref_pic_list_reordering_l1 = h->b_ref_reorder[1];
135 /* If the ref list isn't in the default order, construct reordering header */
136 /* List1 reordering isn't needed yet */
137 if( sh->b_ref_pic_list_reordering_l0 )
139 int pred_frame_num = i_frame;
140 for( int i = 0; i < h->i_ref0; i++ )
142 int diff = h->fref0[i]->i_frame_num - pred_frame_num;
143 sh->ref_pic_list_order[0][i].idc = ( diff > 0 );
144 sh->ref_pic_list_order[0][i].arg = (abs(diff) - 1) & ((1 << sps->i_log2_max_frame_num) - 1);
145 pred_frame_num = h->fref0[i]->i_frame_num;
149 sh->i_cabac_init_idc = param->i_cabac_init_idc;
152 sh->i_qp_delta = i_qp - pps->i_pic_init_qp;
153 sh->b_sp_for_swidth = 0;
156 int deblock_thresh = i_qp + 2 * X264_MIN(param->i_deblocking_filter_alphac0, param->i_deblocking_filter_beta);
157 /* If effective qp <= 15, deblocking would have no effect anyway */
158 if( param->b_deblocking_filter && (h->mb.b_variable_qp || 15 < deblock_thresh ) )
159 sh->i_disable_deblocking_filter_idc = param->b_sliced_threads ? 2 : 0;
161 sh->i_disable_deblocking_filter_idc = 1;
162 sh->i_alpha_c0_offset = param->i_deblocking_filter_alphac0 << 1;
163 sh->i_beta_offset = param->i_deblocking_filter_beta << 1;
166 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 */
188 bs_write_ue( s, sh->i_idr_pic_id );
190 if( sh->sps->i_poc_type == 0 )
192 bs_write( s, sh->sps->i_log2_max_poc_lsb, sh->i_poc & ((1<<sh->sps->i_log2_max_poc_lsb)-1) );
193 if( sh->pps->b_pic_order && !sh->b_field_pic )
194 bs_write_se( s, sh->i_delta_poc_bottom );
196 else if( sh->sps->i_poc_type == 1 && !sh->sps->b_delta_pic_order_always_zero )
198 bs_write_se( s, sh->i_delta_poc[0] );
199 if( sh->pps->b_pic_order && !sh->b_field_pic )
200 bs_write_se( s, sh->i_delta_poc[1] );
203 if( sh->pps->b_redundant_pic_cnt )
204 bs_write_ue( s, sh->i_redundant_pic_cnt );
206 if( sh->i_type == SLICE_TYPE_B )
207 bs_write1( s, sh->b_direct_spatial_mv_pred );
209 if( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_SP || sh->i_type == SLICE_TYPE_B )
211 bs_write1( s, sh->b_num_ref_idx_override );
212 if( sh->b_num_ref_idx_override )
214 bs_write_ue( s, sh->i_num_ref_idx_l0_active - 1 );
215 if( sh->i_type == SLICE_TYPE_B )
216 bs_write_ue( s, sh->i_num_ref_idx_l1_active - 1 );
220 /* ref pic list reordering */
221 if( sh->i_type != SLICE_TYPE_I )
223 bs_write1( s, sh->b_ref_pic_list_reordering_l0 );
224 if( sh->b_ref_pic_list_reordering_l0 )
226 for( int i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
228 bs_write_ue( s, sh->ref_pic_list_order[0][i].idc );
229 bs_write_ue( s, sh->ref_pic_list_order[0][i].arg );
234 if( sh->i_type == SLICE_TYPE_B )
236 bs_write1( s, sh->b_ref_pic_list_reordering_l1 );
237 if( sh->b_ref_pic_list_reordering_l1 )
239 for( int i = 0; i < sh->i_num_ref_idx_l1_active; i++ )
241 bs_write_ue( s, sh->ref_pic_list_order[1][i].idc );
242 bs_write_ue( s, sh->ref_pic_list_order[1][i].arg );
248 if( sh->pps->b_weighted_pred && ( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_SP ) )
250 /* pred_weight_table() */
251 bs_write_ue( s, sh->weight[0][0].i_denom );
252 bs_write_ue( s, sh->weight[0][1].i_denom );
253 for( int i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
255 int luma_weight_l0_flag = !!sh->weight[i][0].weightfn;
256 int chroma_weight_l0_flag = !!sh->weight[i][1].weightfn || !!sh->weight[i][2].weightfn;
257 bs_write1( s, luma_weight_l0_flag );
258 if( luma_weight_l0_flag )
260 bs_write_se( s, sh->weight[i][0].i_scale );
261 bs_write_se( s, sh->weight[i][0].i_offset );
263 bs_write1( s, chroma_weight_l0_flag );
264 if( chroma_weight_l0_flag )
266 for( int j = 1; j < 3; j++ )
268 bs_write_se( s, sh->weight[i][j].i_scale );
269 bs_write_se( s, sh->weight[i][j].i_offset );
274 else if( sh->pps->b_weighted_bipred == 1 && sh->i_type == SLICE_TYPE_B )
279 if( i_nal_ref_idc != 0 )
281 if( sh->i_idr_pic_id >= 0 )
283 bs_write1( s, 0 ); /* no output of prior pics flag */
284 bs_write1( s, 0 ); /* long term reference flag */
288 bs_write1( s, sh->i_mmco_command_count > 0 ); /* adaptive_ref_pic_marking_mode_flag */
289 if( sh->i_mmco_command_count > 0 )
291 for( int i = 0; i < sh->i_mmco_command_count; i++ )
293 bs_write_ue( s, 1 ); /* mark short term ref as unused */
294 bs_write_ue( s, sh->mmco[i].i_difference_of_pic_nums - 1 );
296 bs_write_ue( s, 0 ); /* end command list */
301 if( sh->pps->b_cabac && sh->i_type != SLICE_TYPE_I )
302 bs_write_ue( s, sh->i_cabac_init_idc );
304 bs_write_se( s, sh->i_qp_delta ); /* slice qp delta */
306 if( sh->pps->b_deblocking_filter_control )
308 bs_write_ue( s, sh->i_disable_deblocking_filter_idc );
309 if( sh->i_disable_deblocking_filter_idc != 1 )
311 bs_write_se( s, sh->i_alpha_c0_offset >> 1 );
312 bs_write_se( s, sh->i_beta_offset >> 1 );
317 /* If we are within a reasonable distance of the end of the memory allocated for the bitstream, */
318 /* reallocate, adding an arbitrary amount of space (100 kilobytes). */
319 static int x264_bitstream_check_buffer( x264_t *h )
321 uint8_t *bs_bak = h->out.p_bitstream;
322 if( (h->param.b_cabac && (h->cabac.p_end - h->cabac.p < 2500)) ||
323 (h->out.bs.p_end - h->out.bs.p < 2500) )
325 h->out.i_bitstream += 100000;
326 CHECKED_MALLOC( h->out.p_bitstream, h->out.i_bitstream );
327 h->mc.memcpy_aligned( h->out.p_bitstream, bs_bak, (h->out.i_bitstream - 100000) & ~15 );
328 intptr_t delta = h->out.p_bitstream - bs_bak;
330 h->out.bs.p_start += delta;
331 h->out.bs.p += delta;
332 h->out.bs.p_end = h->out.p_bitstream + h->out.i_bitstream;
334 h->cabac.p_start += delta;
336 h->cabac.p_end = h->out.p_bitstream + h->out.i_bitstream;
338 for( int i = 0; i <= h->out.i_nal; i++ )
339 h->out.nal[i].p_payload += delta;
348 /****************************************************************************
350 ****************************************************************************
351 ****************************** External API*********************************
352 ****************************************************************************
354 ****************************************************************************/
356 static int x264_validate_parameters( x264_t *h )
359 if( !(x264_cpu_detect() & X264_CPU_SSE) )
361 x264_log( h, X264_LOG_ERROR, "your cpu does not support SSE1, but x264 was compiled with asm support\n");
362 x264_log( h, X264_LOG_ERROR, "to run x264, recompile without asm support (configure --disable-asm)\n");
366 if( h->param.i_width <= 0 || h->param.i_height <= 0 )
368 x264_log( h, X264_LOG_ERROR, "invalid width x height (%dx%d)\n",
369 h->param.i_width, h->param.i_height );
373 if( h->param.i_width % 2 || h->param.i_height % 2 )
375 x264_log( h, X264_LOG_ERROR, "width or height not divisible by 2 (%dx%d)\n",
376 h->param.i_width, h->param.i_height );
379 int i_csp = h->param.i_csp & X264_CSP_MASK;
380 if( i_csp != X264_CSP_I420 && i_csp != X264_CSP_YV12 )
382 x264_log( h, X264_LOG_ERROR, "invalid CSP (only I420/YV12 supported)\n" );
386 if( h->param.i_threads == X264_THREADS_AUTO )
387 h->param.i_threads = x264_cpu_num_processors() * (h->param.b_sliced_threads?2:3)/2;
388 h->param.i_threads = x264_clip3( h->param.i_threads, 1, X264_THREAD_MAX );
389 if( h->param.i_threads > 1 )
392 x264_log( h, X264_LOG_WARNING, "not compiled with pthread support!\n");
393 h->param.i_threads = 1;
395 /* Avoid absurdly small thread slices as they can reduce performance
396 * and VBV compliance. Capped at an arbitrary 4 rows per thread. */
397 if( h->param.b_sliced_threads )
399 int max_threads = (h->param.i_height+15)/16 / 4;
400 h->param.i_threads = X264_MIN( h->param.i_threads, max_threads );
404 h->param.b_sliced_threads = 0;
405 h->i_thread_frames = h->param.b_sliced_threads ? 1 : h->param.i_threads;
407 if( h->param.b_interlaced )
409 if( h->param.analyse.i_me_method >= X264_ME_ESA )
411 x264_log( h, X264_LOG_WARNING, "interlace + me=esa is not implemented\n" );
412 h->param.analyse.i_me_method = X264_ME_UMH;
414 if( h->param.analyse.i_weighted_pred > 0 )
416 x264_log( h, X264_LOG_WARNING, "interlace + weightp is not implemented\n" );
417 h->param.analyse.i_weighted_pred = X264_WEIGHTP_NONE;
421 /* Detect default ffmpeg settings and terminate with an error. */
424 score += h->param.analyse.i_me_range == 0;
425 score += h->param.rc.i_qp_step == 3;
426 score += h->param.i_keyint_max == 12;
427 score += h->param.rc.i_qp_min == 2;
428 score += h->param.rc.i_qp_max == 31;
429 score += h->param.rc.f_qcompress == 0.5;
430 score += fabs(h->param.rc.f_ip_factor - 1.25) < 0.01;
431 score += fabs(h->param.rc.f_pb_factor - 1.25) < 0.01;
432 score += h->param.analyse.inter == 0 && h->param.analyse.i_subpel_refine == 8;
435 x264_log( h, X264_LOG_ERROR, "broken ffmpeg default settings detected\n" );
436 x264_log( h, X264_LOG_ERROR, "use an encoding preset (vpre)\n" );
441 if( h->param.rc.i_rc_method < 0 || h->param.rc.i_rc_method > 2 )
443 x264_log( h, X264_LOG_ERROR, "no ratecontrol method specified\n" );
446 h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, 0, 51 );
447 h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, 0, 51 );
448 if( h->param.rc.i_rc_method == X264_RC_CRF )
450 h->param.rc.i_qp_constant = h->param.rc.f_rf_constant;
451 h->param.rc.i_bitrate = 0;
453 if( (h->param.rc.i_rc_method == X264_RC_CQP || h->param.rc.i_rc_method == X264_RC_CRF)
454 && h->param.rc.i_qp_constant == 0 )
456 h->mb.b_lossless = 1;
457 h->param.i_cqm_preset = X264_CQM_FLAT;
458 h->param.psz_cqm_file = NULL;
459 h->param.rc.i_rc_method = X264_RC_CQP;
460 h->param.rc.f_ip_factor = 1;
461 h->param.rc.f_pb_factor = 1;
462 h->param.analyse.b_psnr = 0;
463 h->param.analyse.b_ssim = 0;
464 h->param.analyse.i_chroma_qp_offset = 0;
465 h->param.analyse.i_trellis = 0;
466 h->param.analyse.b_fast_pskip = 0;
467 h->param.analyse.i_noise_reduction = 0;
468 h->param.analyse.b_psy = 0;
469 h->param.i_bframe = 0;
470 /* 8x8dct is not useful at all in CAVLC lossless */
471 if( !h->param.b_cabac )
472 h->param.analyse.b_transform_8x8 = 0;
474 if( h->param.rc.i_rc_method == X264_RC_CQP )
476 float qp_p = h->param.rc.i_qp_constant;
477 float qp_i = qp_p - 6*log2f( h->param.rc.f_ip_factor );
478 float qp_b = qp_p + 6*log2f( h->param.rc.f_pb_factor );
479 h->param.rc.i_qp_min = x264_clip3( (int)(X264_MIN3( qp_p, qp_i, qp_b )), 0, 51 );
480 h->param.rc.i_qp_max = x264_clip3( (int)(X264_MAX3( qp_p, qp_i, qp_b ) + .999), 0, 51 );
481 h->param.rc.i_aq_mode = 0;
482 h->param.rc.b_mb_tree = 0;
484 h->param.rc.i_qp_max = x264_clip3( h->param.rc.i_qp_max, 0, 51 );
485 h->param.rc.i_qp_min = x264_clip3( h->param.rc.i_qp_min, 0, h->param.rc.i_qp_max );
486 if( h->param.rc.i_vbv_buffer_size )
488 if( h->param.rc.i_rc_method == X264_RC_CQP )
490 x264_log( h, X264_LOG_WARNING, "VBV is incompatible with constant QP, ignored.\n" );
491 h->param.rc.i_vbv_max_bitrate = 0;
492 h->param.rc.i_vbv_buffer_size = 0;
494 else if( h->param.rc.i_vbv_max_bitrate == 0 )
496 if( h->param.rc.i_rc_method == X264_RC_ABR )
498 x264_log( h, X264_LOG_WARNING, "VBV maxrate unspecified, assuming CBR\n" );
499 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
503 x264_log( h, X264_LOG_WARNING, "VBV bufsize set but maxrate unspecified, ignored\n" );
504 h->param.rc.i_vbv_buffer_size = 0;
507 else if( h->param.rc.i_vbv_max_bitrate < h->param.rc.i_bitrate &&
508 h->param.rc.i_rc_method == X264_RC_ABR )
510 x264_log( h, X264_LOG_WARNING, "max bitrate less than average bitrate, assuming CBR\n" );
511 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
514 else if( h->param.rc.i_vbv_max_bitrate )
516 x264_log( h, X264_LOG_WARNING, "VBV maxrate specified, but no bufsize, ignored\n" );
517 h->param.rc.i_vbv_max_bitrate = 0;
520 if( h->param.b_interlaced && h->param.i_slice_max_size )
522 x264_log( h, X264_LOG_WARNING, "interlaced + slice-max-size is not implemented\n" );
523 h->param.i_slice_max_size = 0;
525 if( h->param.b_interlaced && h->param.i_slice_max_mbs )
527 x264_log( h, X264_LOG_WARNING, "interlaced + slice-max-mbs is not implemented\n" );
528 h->param.i_slice_max_mbs = 0;
530 int max_slices = (h->param.i_height+((16<<h->param.b_interlaced)-1))/(16<<h->param.b_interlaced);
531 if( h->param.b_sliced_threads )
532 h->param.i_slice_count = x264_clip3( h->param.i_threads, 0, max_slices );
535 h->param.i_slice_count = x264_clip3( h->param.i_slice_count, 0, max_slices );
536 h->param.i_slice_max_size = X264_MAX( h->param.i_slice_max_size, 0 );
537 h->param.i_slice_max_mbs = X264_MAX( h->param.i_slice_max_mbs, 0 );
538 if( h->param.i_slice_max_mbs || h->param.i_slice_max_size )
539 h->param.i_slice_count = 0;
542 h->param.i_frame_reference = x264_clip3( h->param.i_frame_reference, 1, 16 );
543 if( h->param.i_keyint_max <= 0 )
544 h->param.i_keyint_max = 1;
545 if( h->param.i_scenecut_threshold < 0 )
546 h->param.i_scenecut_threshold = 0;
547 if( !h->param.analyse.i_subpel_refine && h->param.analyse.i_direct_mv_pred > X264_DIRECT_PRED_SPATIAL )
549 x264_log( h, X264_LOG_WARNING, "subme=0 + direct=temporal is not supported\n" );
550 h->param.analyse.i_direct_mv_pred = X264_DIRECT_PRED_SPATIAL;
552 h->param.i_bframe = x264_clip3( h->param.i_bframe, 0, X264_BFRAME_MAX );
553 if( h->param.i_keyint_max == 1 )
555 h->param.i_bframe = 0;
556 h->param.b_intra_refresh = 0;
558 h->param.i_bframe_bias = x264_clip3( h->param.i_bframe_bias, -90, 100 );
559 if( h->param.i_bframe <= 1 )
560 h->param.i_bframe_pyramid = X264_B_PYRAMID_NONE;
561 h->param.i_bframe_pyramid = x264_clip3( h->param.i_bframe_pyramid, X264_B_PYRAMID_NONE, X264_B_PYRAMID_NORMAL );
562 if( !h->param.i_bframe )
564 h->param.i_bframe_adaptive = X264_B_ADAPT_NONE;
565 h->param.analyse.i_direct_mv_pred = 0;
566 h->param.analyse.b_weighted_bipred = 0;
568 if( h->param.b_intra_refresh && h->param.i_bframe_pyramid == X264_B_PYRAMID_NORMAL )
570 x264_log( h, X264_LOG_WARNING, "b-pyramid normal + intra-refresh is not supported\n" );
571 h->param.i_bframe_pyramid = X264_B_PYRAMID_STRICT;
573 if( h->param.b_intra_refresh && h->param.i_frame_reference > 1 )
575 x264_log( h, X264_LOG_WARNING, "ref > 1 + intra-refresh is not supported\n" );
576 h->param.i_frame_reference = 1;
578 if( h->param.i_keyint_min == X264_KEYINT_MIN_AUTO )
579 h->param.i_keyint_min = h->param.i_keyint_max / 10;
580 h->param.i_keyint_min = x264_clip3( h->param.i_keyint_min, 1, h->param.i_keyint_max/2+1 );
581 h->param.rc.i_lookahead = x264_clip3( h->param.rc.i_lookahead, 0, X264_LOOKAHEAD_MAX );
583 int maxrate = X264_MAX( h->param.rc.i_vbv_max_bitrate, h->param.rc.i_bitrate );
584 float bufsize = maxrate ? (float)h->param.rc.i_vbv_buffer_size / maxrate : 0;
585 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;
586 h->param.rc.i_lookahead = X264_MIN( h->param.rc.i_lookahead, X264_MAX( h->param.i_keyint_max, bufsize*fps ) );
589 if( !h->param.i_timebase_num || !h->param.i_timebase_den )
591 h->param.i_timebase_num = h->param.i_fps_den;
592 h->param.i_timebase_den = h->param.i_fps_num;
595 h->param.rc.f_qcompress = x264_clip3f( h->param.rc.f_qcompress, 0.0, 1.0 );
596 if( !h->param.rc.i_lookahead || h->param.i_keyint_max == 1 || h->param.rc.f_qcompress == 1 )
597 h->param.rc.b_mb_tree = 0;
598 if( h->param.rc.b_stat_read )
599 h->param.rc.i_lookahead = 0;
601 if( h->param.i_sync_lookahead )
602 h->param.i_sync_lookahead = x264_clip3( h->param.i_sync_lookahead, h->i_thread_frames + h->param.i_bframe, X264_LOOKAHEAD_MAX );
603 if( h->param.rc.b_stat_read || h->i_thread_frames == 1 )
604 h->param.i_sync_lookahead = 0;
606 h->param.i_sync_lookahead = 0;
609 h->param.i_deblocking_filter_alphac0 = x264_clip3( h->param.i_deblocking_filter_alphac0, -6, 6 );
610 h->param.i_deblocking_filter_beta = x264_clip3( h->param.i_deblocking_filter_beta, -6, 6 );
611 h->param.analyse.i_luma_deadzone[0] = x264_clip3( h->param.analyse.i_luma_deadzone[0], 0, 32 );
612 h->param.analyse.i_luma_deadzone[1] = x264_clip3( h->param.analyse.i_luma_deadzone[1], 0, 32 );
614 h->param.i_cabac_init_idc = x264_clip3( h->param.i_cabac_init_idc, 0, 2 );
616 if( h->param.i_cqm_preset < X264_CQM_FLAT || h->param.i_cqm_preset > X264_CQM_CUSTOM )
617 h->param.i_cqm_preset = X264_CQM_FLAT;
619 if( h->param.analyse.i_me_method < X264_ME_DIA ||
620 h->param.analyse.i_me_method > X264_ME_TESA )
621 h->param.analyse.i_me_method = X264_ME_HEX;
622 if( h->param.analyse.i_me_range < 4 )
623 h->param.analyse.i_me_range = 4;
624 if( h->param.analyse.i_me_range > 16 && h->param.analyse.i_me_method <= X264_ME_HEX )
625 h->param.analyse.i_me_range = 16;
626 if( h->param.analyse.i_me_method == X264_ME_TESA &&
627 (h->mb.b_lossless || h->param.analyse.i_subpel_refine <= 1) )
628 h->param.analyse.i_me_method = X264_ME_ESA;
629 h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 10 );
630 h->param.analyse.b_mixed_references = h->param.analyse.b_mixed_references && h->param.i_frame_reference > 1;
631 h->param.analyse.inter &= X264_ANALYSE_PSUB16x16|X264_ANALYSE_PSUB8x8|X264_ANALYSE_BSUB16x16|
632 X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
633 h->param.analyse.intra &= X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
634 if( !(h->param.analyse.inter & X264_ANALYSE_PSUB16x16) )
635 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
636 if( !h->param.analyse.b_transform_8x8 )
638 h->param.analyse.inter &= ~X264_ANALYSE_I8x8;
639 h->param.analyse.intra &= ~X264_ANALYSE_I8x8;
641 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
642 if( !h->param.b_cabac )
643 h->param.analyse.i_trellis = 0;
644 h->param.analyse.i_trellis = x264_clip3( h->param.analyse.i_trellis, 0, 2 );
645 if( !h->param.analyse.b_psy )
647 h->param.analyse.f_psy_rd = 0;
648 h->param.analyse.f_psy_trellis = 0;
650 if( !h->param.analyse.i_trellis )
651 h->param.analyse.f_psy_trellis = 0;
652 h->param.analyse.f_psy_rd = x264_clip3f( h->param.analyse.f_psy_rd, 0, 10 );
653 h->param.analyse.f_psy_trellis = x264_clip3f( h->param.analyse.f_psy_trellis, 0, 10 );
654 if( h->param.analyse.i_subpel_refine < 6 )
655 h->param.analyse.f_psy_rd = 0;
656 h->mb.i_psy_rd = FIX8( h->param.analyse.f_psy_rd );
657 /* Psy RDO increases overall quantizers to improve the quality of luma--this indirectly hurts chroma quality */
658 /* so we lower the chroma QP offset to compensate */
659 /* This can be triggered repeatedly on multiple calls to parameter_validate, but since encoding
660 * uses the pps chroma qp offset not the param chroma qp offset, this is not a problem. */
662 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_rd < 0.25 ? 1 : 2;
663 h->mb.i_psy_trellis = FIX8( h->param.analyse.f_psy_trellis / 4 );
664 /* Psy trellis has a similar effect. */
665 if( h->mb.i_psy_trellis )
666 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_trellis < 0.25 ? 1 : 2;
667 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
668 h->param.rc.i_aq_mode = x264_clip3( h->param.rc.i_aq_mode, 0, 2 );
669 h->param.rc.f_aq_strength = x264_clip3f( h->param.rc.f_aq_strength, 0, 3 );
670 if( h->param.rc.f_aq_strength == 0 )
671 h->param.rc.i_aq_mode = 0;
672 /* MB-tree requires AQ to be on, even if the strength is zero. */
673 if( !h->param.rc.i_aq_mode && h->param.rc.b_mb_tree )
675 h->param.rc.i_aq_mode = 1;
676 h->param.rc.f_aq_strength = 0;
678 h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
679 if( h->param.analyse.i_subpel_refine == 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
680 h->param.analyse.i_subpel_refine = 9;
683 const x264_level_t *l = x264_levels;
684 if( h->param.i_level_idc < 0 )
686 int maxrate_bak = h->param.rc.i_vbv_max_bitrate;
687 if( h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.i_vbv_buffer_size <= 0 )
688 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate * 2;
689 h->sps = h->sps_array;
690 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
691 do h->param.i_level_idc = l->level_idc;
692 while( l[1].level_idc && x264_validate_levels( h, 0 ) && l++ );
693 h->param.rc.i_vbv_max_bitrate = maxrate_bak;
697 while( l->level_idc && l->level_idc != h->param.i_level_idc )
699 if( l->level_idc == 0 )
701 x264_log( h, X264_LOG_ERROR, "invalid level_idc: %d\n", h->param.i_level_idc );
705 if( h->param.analyse.i_mv_range <= 0 )
706 h->param.analyse.i_mv_range = l->mv_range >> h->param.b_interlaced;
708 h->param.analyse.i_mv_range = x264_clip3(h->param.analyse.i_mv_range, 32, 512 >> h->param.b_interlaced);
711 h->param.analyse.i_weighted_pred = x264_clip3( h->param.analyse.i_weighted_pred, 0, X264_WEIGHTP_SMART );
712 if( !h->param.analyse.i_weighted_pred && h->param.rc.b_mb_tree && h->param.analyse.b_psy && !h->param.b_interlaced )
713 h->param.analyse.i_weighted_pred = X264_WEIGHTP_FAKE;
715 if( h->i_thread_frames > 1 )
717 int r = h->param.analyse.i_mv_range_thread;
721 // half of the available space is reserved and divided evenly among the threads,
722 // the rest is allocated to whichever thread is far enough ahead to use it.
723 // reserving more space increases quality for some videos, but costs more time
724 // in thread synchronization.
725 int max_range = (h->param.i_height + X264_THREAD_HEIGHT) / h->i_thread_frames - X264_THREAD_HEIGHT;
728 r = X264_MAX( r, h->param.analyse.i_me_range );
729 r = X264_MIN( r, h->param.analyse.i_mv_range );
730 // round up to use the whole mb row
731 r2 = (r & ~15) + ((-X264_THREAD_HEIGHT) & 15);
734 x264_log( h, X264_LOG_DEBUG, "using mv_range_thread = %d\n", r2 );
735 h->param.analyse.i_mv_range_thread = r2;
738 if( h->param.rc.f_qblur < 0 )
739 h->param.rc.f_qblur = 0;
740 if( h->param.rc.f_complexity_blur < 0 )
741 h->param.rc.f_complexity_blur = 0;
743 h->param.i_sps_id &= 31;
745 if( h->param.i_log_level < X264_LOG_INFO )
747 h->param.analyse.b_psnr = 0;
748 h->param.analyse.b_ssim = 0;
751 if( h->param.b_interlaced )
752 h->param.b_pic_struct = 1;
754 if( h->param.i_nal_hrd && !h->param.rc.i_vbv_buffer_size )
756 x264_log( h, X264_LOG_WARNING, "NAL HRD parameters require VBV parameters\n" );
757 h->param.i_nal_hrd = X264_NAL_HRD_NONE;
760 if( h->param.i_nal_hrd == X264_NAL_HRD_CBR &&
761 (h->param.rc.i_bitrate != h->param.rc.i_vbv_max_bitrate || !h->param.rc.i_vbv_max_bitrate) )
763 x264_log( h, X264_LOG_WARNING, "CBR HRD requires constant bitrate\n" );
764 h->param.i_nal_hrd = X264_NAL_HRD_VBR;
767 /* ensure the booleans are 0 or 1 so they can be used in math */
768 #define BOOLIFY(x) h->param.x = !!h->param.x
770 BOOLIFY( b_constrained_intra );
771 BOOLIFY( b_deblocking_filter );
772 BOOLIFY( b_deterministic );
773 BOOLIFY( b_sliced_threads );
774 BOOLIFY( b_interlaced );
775 BOOLIFY( b_intra_refresh );
776 BOOLIFY( b_visualize );
778 BOOLIFY( b_repeat_headers );
780 BOOLIFY( b_vfr_input );
781 BOOLIFY( b_pic_struct );
782 BOOLIFY( analyse.b_transform_8x8 );
783 BOOLIFY( analyse.b_weighted_bipred );
784 BOOLIFY( analyse.b_chroma_me );
785 BOOLIFY( analyse.b_mixed_references );
786 BOOLIFY( analyse.b_fast_pskip );
787 BOOLIFY( analyse.b_dct_decimate );
788 BOOLIFY( analyse.b_psy );
789 BOOLIFY( analyse.b_psnr );
790 BOOLIFY( analyse.b_ssim );
791 BOOLIFY( rc.b_stat_write );
792 BOOLIFY( rc.b_stat_read );
793 BOOLIFY( rc.b_mb_tree );
799 static void mbcmp_init( x264_t *h )
801 int satd = !h->mb.b_lossless && h->param.analyse.i_subpel_refine > 1;
802 memcpy( h->pixf.mbcmp, satd ? h->pixf.satd : h->pixf.sad_aligned, sizeof(h->pixf.mbcmp) );
803 memcpy( h->pixf.mbcmp_unaligned, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.mbcmp_unaligned) );
804 h->pixf.intra_mbcmp_x3_16x16 = satd ? h->pixf.intra_satd_x3_16x16 : h->pixf.intra_sad_x3_16x16;
805 h->pixf.intra_mbcmp_x3_8x8c = satd ? h->pixf.intra_satd_x3_8x8c : h->pixf.intra_sad_x3_8x8c;
806 h->pixf.intra_mbcmp_x3_4x4 = satd ? h->pixf.intra_satd_x3_4x4 : h->pixf.intra_sad_x3_4x4;
807 satd &= h->param.analyse.i_me_method == X264_ME_TESA;
808 memcpy( h->pixf.fpelcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.fpelcmp) );
809 memcpy( h->pixf.fpelcmp_x3, satd ? h->pixf.satd_x3 : h->pixf.sad_x3, sizeof(h->pixf.fpelcmp_x3) );
810 memcpy( h->pixf.fpelcmp_x4, satd ? h->pixf.satd_x4 : h->pixf.sad_x4, sizeof(h->pixf.fpelcmp_x4) );
813 static void x264_set_aspect_ratio( x264_t *h, x264_param_t *param, int initial )
816 if( param->vui.i_sar_width > 0 && param->vui.i_sar_height > 0 )
818 uint32_t i_w = param->vui.i_sar_width;
819 uint32_t i_h = param->vui.i_sar_height;
820 uint32_t old_w = h->param.vui.i_sar_width;
821 uint32_t old_h = h->param.vui.i_sar_height;
823 x264_reduce_fraction( &i_w, &i_h );
825 while( i_w > 65535 || i_h > 65535 )
831 x264_reduce_fraction( &i_w, &i_h );
833 if( i_w != old_w || i_h != old_h || initial )
835 h->param.vui.i_sar_width = 0;
836 h->param.vui.i_sar_height = 0;
837 if( i_w == 0 || i_h == 0 )
838 x264_log( h, X264_LOG_WARNING, "cannot create valid sample aspect ratio\n" );
841 x264_log( h, initial?X264_LOG_INFO:X264_LOG_DEBUG, "using SAR=%d/%d\n", i_w, i_h );
842 h->param.vui.i_sar_width = i_w;
843 h->param.vui.i_sar_height = i_h;
849 /****************************************************************************
851 ****************************************************************************/
852 x264_t *x264_encoder_open( x264_param_t *param )
856 int qp, i_slicetype_length;
858 CHECKED_MALLOCZERO( h, sizeof(x264_t) );
860 /* Create a copy of param */
861 memcpy( &h->param, param, sizeof(x264_param_t) );
863 if( param->param_free )
864 param->param_free( param );
866 if( x264_validate_parameters( h ) < 0 )
869 if( h->param.psz_cqm_file )
870 if( x264_cqm_parse_file( h, h->param.psz_cqm_file ) < 0 )
873 if( h->param.rc.psz_stat_out )
874 h->param.rc.psz_stat_out = strdup( h->param.rc.psz_stat_out );
875 if( h->param.rc.psz_stat_in )
876 h->param.rc.psz_stat_in = strdup( h->param.rc.psz_stat_in );
878 x264_set_aspect_ratio( h, &h->param, 1 );
880 x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den );
881 x264_reduce_fraction( &h->param.i_timebase_num, &h->param.i_timebase_den );
887 uint64_t new_timebase_den = h->param.i_timebase_den;
888 if( h->param.b_dts_compress )
890 /* h->i_dts_compress_multiplier == h->frames.i_bframe_delay + 1 */
891 h->i_dts_compress_multiplier = h->param.i_bframe ? (h->param.i_bframe_pyramid ? 3 : 2) : 1;
892 if( h->i_dts_compress_multiplier != 1 )
894 new_timebase_den = h->param.i_timebase_den * h->i_dts_compress_multiplier;
895 x264_log( h, X264_LOG_DEBUG, "DTS compresion changed timebase: %u/%u -> %u/%"PRIu64"\n",
896 h->param.i_timebase_num, h->param.i_timebase_den,
897 h->param.i_timebase_num, new_timebase_den );
901 h->i_dts_compress_multiplier = 1;
903 if( new_timebase_den * 2 > UINT32_MAX )
905 x264_log( h, X264_LOG_ERROR, "Effective timebase denominator %"PRIu64" exceeds H.264 maximum\n", new_timebase_den );
908 h->param.i_timebase_den = new_timebase_den;
910 h->sps = &h->sps_array[0];
911 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
913 h->pps = &h->pps_array[0];
914 x264_pps_init( h->pps, h->param.i_sps_id, &h->param, h->sps );
916 x264_validate_levels( h, 1 );
918 h->chroma_qp_table = i_chroma_qp_table + 12 + h->pps->i_chroma_qp_index_offset;
920 if( x264_cqm_init( h ) < 0 )
923 h->mb.i_mb_count = h->sps->i_mb_width * h->sps->i_mb_height;
926 if( h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS )
927 h->frames.i_delay = X264_MAX(h->param.i_bframe,3)*4;
929 h->frames.i_delay = h->param.i_bframe;
930 if( h->param.rc.b_mb_tree || h->param.rc.i_vbv_buffer_size )
931 h->frames.i_delay = X264_MAX( h->frames.i_delay, h->param.rc.i_lookahead );
932 i_slicetype_length = h->frames.i_delay;
933 h->frames.i_delay += h->i_thread_frames - 1;
934 h->frames.i_delay = X264_MIN( h->frames.i_delay, X264_LOOKAHEAD_MAX );
935 h->frames.i_delay += h->param.i_sync_lookahead;
936 h->frames.i_delay += h->param.b_vfr_input && (h->param.rc.i_rc_method == X264_RC_ABR || h->param.rc.b_stat_write
937 || h->param.rc.i_vbv_buffer_size);
938 h->frames.i_bframe_delay = h->param.i_bframe ? (h->param.i_bframe_pyramid ? 2 : 1) : 0;
940 h->frames.i_max_ref0 = h->param.i_frame_reference;
941 h->frames.i_max_ref1 = X264_MIN( h->sps->vui.i_num_reorder_frames, h->param.i_frame_reference );
942 h->frames.i_max_dpb = h->sps->vui.i_max_dec_frame_buffering;
943 h->frames.b_have_lowres = !h->param.rc.b_stat_read
944 && ( h->param.rc.i_rc_method == X264_RC_ABR
945 || h->param.rc.i_rc_method == X264_RC_CRF
946 || h->param.i_bframe_adaptive
947 || h->param.i_scenecut_threshold
948 || h->param.rc.b_mb_tree
949 || h->param.analyse.i_weighted_pred == X264_WEIGHTP_SMART );
950 h->frames.b_have_lowres |= h->param.rc.b_stat_read && h->param.rc.i_vbv_buffer_size > 0;
951 h->frames.b_have_sub8x8_esa = !!(h->param.analyse.inter & X264_ANALYSE_PSUB8x8);
953 h->frames.i_last_keyframe = - h->param.i_keyint_max;
954 h->frames.i_input = 0;
955 h->frames.i_largest_pts = h->frames.i_second_largest_pts = -1;
957 CHECKED_MALLOCZERO( h->frames.unused[0], (h->frames.i_delay + 3) * sizeof(x264_frame_t *) );
958 /* Allocate room for max refs plus a few extra just in case. */
959 CHECKED_MALLOCZERO( h->frames.unused[1], (h->i_thread_frames + 20) * sizeof(x264_frame_t *) );
960 CHECKED_MALLOCZERO( h->frames.current, (h->param.i_sync_lookahead + h->param.i_bframe
961 + h->i_thread_frames + 3) * sizeof(x264_frame_t *) );
962 if( h->param.analyse.i_weighted_pred > 0 )
963 CHECKED_MALLOCZERO( h->frames.blank_unused, h->i_thread_frames * 4 * sizeof(x264_frame_t *) );
966 h->i_cpb_delay = h->i_coded_fields = h->i_disp_fields = h->i_prev_duration = 0;
967 h->i_disp_fields_last_frame = -1;
970 /* init CPU functions */
971 x264_predict_16x16_init( h->param.cpu, h->predict_16x16 );
972 x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c );
973 x264_predict_8x8_init( h->param.cpu, h->predict_8x8, &h->predict_8x8_filter );
974 x264_predict_4x4_init( h->param.cpu, h->predict_4x4 );
975 if( !h->param.b_cabac )
976 x264_init_vlc_tables();
977 x264_pixel_init( h->param.cpu, &h->pixf );
978 x264_dct_init( h->param.cpu, &h->dctf );
979 x264_zigzag_init( h->param.cpu, &h->zigzagf, h->param.b_interlaced );
980 x264_mc_init( h->param.cpu, &h->mc );
981 x264_quant_init( h, h->param.cpu, &h->quantf );
982 x264_deblock_init( h->param.cpu, &h->loopf );
983 x264_dct_init_weights();
987 p = buf + sprintf( buf, "using cpu capabilities:" );
988 for( int i = 0; x264_cpu_names[i].flags; i++ )
990 if( !strcmp(x264_cpu_names[i].name, "SSE2")
991 && h->param.cpu & (X264_CPU_SSE2_IS_FAST|X264_CPU_SSE2_IS_SLOW) )
993 if( !strcmp(x264_cpu_names[i].name, "SSE3")
994 && (h->param.cpu & X264_CPU_SSSE3 || !(h->param.cpu & X264_CPU_CACHELINE_64)) )
996 if( !strcmp(x264_cpu_names[i].name, "SSE4.1")
997 && (h->param.cpu & X264_CPU_SSE42) )
999 if( (h->param.cpu & x264_cpu_names[i].flags) == x264_cpu_names[i].flags
1000 && (!i || x264_cpu_names[i].flags != x264_cpu_names[i-1].flags) )
1001 p += sprintf( p, " %s", x264_cpu_names[i].name );
1004 p += sprintf( p, " none!" );
1005 x264_log( h, X264_LOG_INFO, "%s\n", buf );
1007 for( qp = h->param.rc.i_qp_min; qp <= h->param.rc.i_qp_max; qp++ )
1008 if( x264_analyse_init_costs( h, qp ) )
1010 if( x264_analyse_init_costs( h, X264_LOOKAHEAD_QP ) )
1013 /* Checks for known miscompilation issues. */
1014 if( h->cost_mv[1][2013] != 24 )
1016 x264_log( h, X264_LOG_ERROR, "MV cost test failed: x264 has been miscompiled!\n" );
1020 /* Must be volatile or else GCC will optimize it out. */
1021 volatile int temp = 392;
1022 if( x264_clz( temp ) != 23 )
1024 x264_log( h, X264_LOG_ERROR, "CLZ test failed: x264 has been miscompiled!\n" );
1025 #if defined(ARCH_X86) || defined(ARCH_X86_64)
1026 x264_log( h, X264_LOG_ERROR, "Are you attempting to run an SSE4a-targeted build on a CPU that\n" );
1027 x264_log( h, X264_LOG_ERROR, "doesn't support it?\n" );
1033 h->out.i_bitstream = X264_MAX( 1000000, h->param.i_width * h->param.i_height * 4
1034 * ( h->param.rc.i_rc_method == X264_RC_ABR ? pow( 0.95, h->param.rc.i_qp_min )
1035 : pow( 0.95, h->param.rc.i_qp_constant ) * X264_MAX( 1, h->param.rc.f_ip_factor )));
1037 CHECKED_MALLOC( h->nal_buffer, h->out.i_bitstream * 3/2 + 4 );
1038 h->nal_buffer_size = h->out.i_bitstream * 3/2 + 4;
1041 for( int i = 1; i < h->param.i_threads + !!h->param.i_sync_lookahead; i++ )
1042 CHECKED_MALLOC( h->thread[i], sizeof(x264_t) );
1044 if( x264_lookahead_init( h, i_slicetype_length ) )
1047 for( int i = 0; i < h->param.i_threads; i++ )
1049 int init_nal_count = h->param.i_slice_count + 3;
1050 int allocate_threadlocal_data = !h->param.b_sliced_threads || !i;
1054 if( allocate_threadlocal_data )
1056 h->thread[i]->fdec = x264_frame_pop_unused( h, 1 );
1057 if( !h->thread[i]->fdec )
1061 h->thread[i]->fdec = h->thread[0]->fdec;
1063 CHECKED_MALLOC( h->thread[i]->out.p_bitstream, h->out.i_bitstream );
1064 /* Start each thread with room for init_nal_count NAL units; it'll realloc later if needed. */
1065 CHECKED_MALLOC( h->thread[i]->out.nal, init_nal_count*sizeof(x264_nal_t) );
1066 h->thread[i]->out.i_nals_allocated = init_nal_count;
1068 if( allocate_threadlocal_data && x264_macroblock_cache_allocate( h->thread[i] ) < 0 )
1072 for( int i = 0; i < h->param.i_threads; i++ )
1073 if( x264_macroblock_thread_allocate( h->thread[i], 0 ) < 0 )
1076 if( x264_ratecontrol_new( h ) < 0 )
1079 if( h->param.i_nal_hrd )
1081 x264_log( h, X264_LOG_DEBUG, "HRD bitrate: %i bits/sec\n", h->sps->vui.hrd.i_bit_rate_unscaled );
1082 x264_log( h, X264_LOG_DEBUG, "CPB size: %i bits\n", h->sps->vui.hrd.i_cpb_size_unscaled );
1085 if( h->param.psz_dump_yuv )
1087 /* create or truncate the reconstructed video file */
1088 FILE *f = fopen( h->param.psz_dump_yuv, "w" );
1091 x264_log( h, X264_LOG_ERROR, "dump_yuv: can't write to %s\n", h->param.psz_dump_yuv );
1094 else if( !x264_is_regular_file( f ) )
1096 x264_log( h, X264_LOG_ERROR, "dump_yuv: incompatible with non-regular file %s\n", h->param.psz_dump_yuv );
1102 x264_log( h, X264_LOG_INFO, "profile %s, level %d.%d\n",
1103 h->sps->i_profile_idc == PROFILE_BASELINE ? "Baseline" :
1104 h->sps->i_profile_idc == PROFILE_MAIN ? "Main" :
1105 h->sps->i_profile_idc == PROFILE_HIGH ? "High" :
1106 "High 4:4:4 Predictive", h->sps->i_level_idc/10, h->sps->i_level_idc%10 );
1114 /****************************************************************************
1115 * x264_encoder_reconfig:
1116 ****************************************************************************/
1117 int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
1119 int rc_reconfig = 0;
1120 h = h->thread[h->thread[0]->i_thread_phase];
1121 x264_set_aspect_ratio( h, param, 0 );
1122 #define COPY(var) h->param.var = param->var
1123 COPY( i_frame_reference ); // but never uses more refs than initially specified
1124 COPY( i_bframe_bias );
1125 if( h->param.i_scenecut_threshold )
1126 COPY( i_scenecut_threshold ); // can't turn it on or off, only vary the threshold
1127 COPY( b_deblocking_filter );
1128 COPY( i_deblocking_filter_alphac0 );
1129 COPY( i_deblocking_filter_beta );
1130 COPY( analyse.inter );
1131 COPY( analyse.intra );
1132 COPY( analyse.i_direct_mv_pred );
1133 /* Scratch buffer prevents me_range from being increased for esa/tesa */
1134 if( h->param.analyse.i_me_method < X264_ME_ESA || param->analyse.i_me_range < h->param.analyse.i_me_range )
1135 COPY( analyse.i_me_range );
1136 COPY( analyse.i_noise_reduction );
1137 /* We can't switch out of subme=0 during encoding. */
1138 if( h->param.analyse.i_subpel_refine )
1139 COPY( analyse.i_subpel_refine );
1140 COPY( analyse.i_trellis );
1141 COPY( analyse.b_chroma_me );
1142 COPY( analyse.b_dct_decimate );
1143 COPY( analyse.b_fast_pskip );
1144 COPY( analyse.b_mixed_references );
1145 COPY( analyse.f_psy_rd );
1146 COPY( analyse.f_psy_trellis );
1147 // can only twiddle these if they were enabled to begin with:
1148 if( h->param.analyse.i_me_method >= X264_ME_ESA || param->analyse.i_me_method < X264_ME_ESA )
1149 COPY( analyse.i_me_method );
1150 if( h->param.analyse.i_me_method >= X264_ME_ESA && !h->frames.b_have_sub8x8_esa )
1151 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
1152 if( h->pps->b_transform_8x8_mode )
1153 COPY( analyse.b_transform_8x8 );
1154 if( h->frames.i_max_ref1 > 1 )
1155 COPY( i_bframe_pyramid );
1156 COPY( i_slice_max_size );
1157 COPY( i_slice_max_mbs );
1158 COPY( i_slice_count );
1160 /* VBV can't be turned on if it wasn't on to begin with */
1161 if( h->param.rc.i_vbv_max_bitrate > 0 && h->param.rc.i_vbv_buffer_size > 0 &&
1162 param->rc.i_vbv_max_bitrate > 0 && param->rc.i_vbv_buffer_size > 0 )
1164 COPY( rc.i_vbv_max_bitrate );
1165 COPY( rc.i_vbv_buffer_size );
1166 COPY( rc.i_bitrate );
1169 if( h->param.rc.f_rf_constant != param->rc.f_rf_constant )
1171 COPY( rc.f_rf_constant );
1174 if( h->param.rc.f_rf_constant_max != param->rc.f_rf_constant_max )
1176 COPY( rc.f_rf_constant_max );
1184 int ret = x264_validate_parameters( h );
1186 /* Supported reconfiguration options (1-pass only):
1190 * bitrate (CBR only) */
1191 if( !ret && rc_reconfig )
1192 x264_ratecontrol_init_reconfigurable( h, 0 );
1197 /****************************************************************************
1198 * x264_encoder_parameters:
1199 ****************************************************************************/
1200 void x264_encoder_parameters( x264_t *h, x264_param_t *param )
1202 memcpy( param, &h->thread[h->i_thread_phase]->param, sizeof(x264_param_t) );
1205 /* internal usage */
1206 static void x264_nal_start( x264_t *h, int i_type, int i_ref_idc )
1208 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1210 nal->i_ref_idc = i_ref_idc;
1211 nal->i_type = i_type;
1214 nal->p_payload= &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1217 /* if number of allocated nals is not enough, re-allocate a larger one. */
1218 static int x264_nal_check_buffer( x264_t *h )
1220 if( h->out.i_nal >= h->out.i_nals_allocated )
1222 x264_nal_t *new_out = x264_malloc( sizeof(x264_nal_t) * (h->out.i_nals_allocated*2) );
1225 memcpy( new_out, h->out.nal, sizeof(x264_nal_t) * (h->out.i_nals_allocated) );
1226 x264_free( h->out.nal );
1227 h->out.nal = new_out;
1228 h->out.i_nals_allocated *= 2;
1233 static int x264_nal_end( x264_t *h )
1235 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1236 nal->i_payload = &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8] - nal->p_payload;
1239 return x264_nal_check_buffer( h );
1242 static int x264_encoder_encapsulate_nals( x264_t *h, int start )
1244 int nal_size = 0, previous_nal_size = 0;
1246 for( int i = 0; i < start; i++ )
1247 previous_nal_size += h->out.nal[i].i_payload;
1249 for( int i = start; i < h->out.i_nal; i++ )
1250 nal_size += h->out.nal[i].i_payload;
1252 /* Worst-case NAL unit escaping: reallocate the buffer if it's too small. */
1253 if( h->nal_buffer_size < nal_size * 3/2 + h->out.i_nal * 4 )
1255 uint8_t *buf = x264_malloc( nal_size * 2 + h->out.i_nal * 4 );
1258 if( previous_nal_size )
1259 memcpy( buf, h->nal_buffer, previous_nal_size );
1260 x264_free( h->nal_buffer );
1261 h->nal_buffer = buf;
1264 uint8_t *nal_buffer = h->nal_buffer + previous_nal_size;
1266 for( int i = start; i < h->out.i_nal; i++ )
1268 int long_startcode = !i || h->out.nal[i].i_type == NAL_SPS || h->out.nal[i].i_type == NAL_PPS;
1269 int size = x264_nal_encode( nal_buffer, &h->out.nal[i], h->param.b_annexb, long_startcode );
1270 h->out.nal[i].i_payload = size;
1271 h->out.nal[i].p_payload = nal_buffer;
1275 return nal_buffer - (h->nal_buffer + previous_nal_size);
1278 /****************************************************************************
1279 * x264_encoder_headers:
1280 ****************************************************************************/
1281 int x264_encoder_headers( x264_t *h, x264_nal_t **pp_nal, int *pi_nal )
1284 /* init bitstream context */
1286 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
1288 /* Write SEI, SPS and PPS. */
1290 /* generate sequence parameters */
1291 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
1292 x264_sps_write( &h->out.bs, h->sps );
1293 if( x264_nal_end( h ) )
1296 /* generate picture parameters */
1297 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
1298 x264_pps_write( &h->out.bs, h->pps );
1299 if( x264_nal_end( h ) )
1302 /* identify ourselves */
1303 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
1304 if( x264_sei_version_write( h, &h->out.bs ) )
1306 if( x264_nal_end( h ) )
1309 frame_size = x264_encoder_encapsulate_nals( h, 0 );
1312 *pi_nal = h->out.i_nal;
1313 *pp_nal = &h->out.nal[0];
1319 /* Check to see whether we have chosen a reference list ordering different
1320 * from the standard's default. */
1321 static inline void x264_reference_check_reorder( x264_t *h )
1323 for( int i = 0; i < h->i_ref0 - 1; i++ )
1324 /* P and B-frames use different default orders. */
1325 if( h->sh.i_type == SLICE_TYPE_P ? h->fref0[i]->i_frame_num < h->fref0[i+1]->i_frame_num
1326 : h->fref0[i]->i_poc < h->fref0[i+1]->i_poc )
1328 h->b_ref_reorder[0] = 1;
1333 /* return -1 on failure, else return the index of the new reference frame */
1334 int x264_weighted_reference_duplicate( x264_t *h, int i_ref, const x264_weight_t *w )
1338 x264_frame_t *newframe;
1339 if( i <= 1 ) /* empty list, definitely can't duplicate frame */
1342 newframe = x264_frame_pop_blank_unused( h );
1344 //FIXME: probably don't need to copy everything
1345 *newframe = *h->fref0[i_ref];
1346 newframe->i_reference_count = 1;
1347 newframe->orig = h->fref0[i_ref];
1348 newframe->b_duplicate = 1;
1349 memcpy( h->fenc->weight[j], w, sizeof(h->fenc->weight[i]) );
1351 /* shift the frames to make space for the dupe. */
1352 h->b_ref_reorder[0] = 1;
1353 if( h->i_ref0 < 16 )
1355 h->fref0[15] = NULL;
1356 x264_frame_unshift( &h->fref0[j], newframe );
1361 static void x264_weighted_pred_init( x264_t *h )
1363 /* for now no analysis and set all weights to nothing */
1364 for( int i_ref = 0; i_ref < h->i_ref0; i_ref++ )
1365 h->fenc->weighted[i_ref] = h->fref0[i_ref]->filtered[0];
1367 // FIXME: This only supports weighting of one reference frame
1368 // and duplicates of that frame.
1369 h->fenc->i_lines_weighted = 0;
1371 for( int i_ref = 0; i_ref < (h->i_ref0 << h->sh.b_mbaff); i_ref++ )
1372 for( int i = 0; i < 3; i++ )
1373 h->sh.weight[i_ref][i].weightfn = NULL;
1376 if( h->sh.i_type != SLICE_TYPE_P || h->param.analyse.i_weighted_pred <= 0 )
1379 int i_padv = PADV << h->param.b_interlaced;
1382 int buffer_next = 0;
1383 //FIXME: when chroma support is added, move this into loop
1384 h->sh.weight[0][1].weightfn = h->sh.weight[0][2].weightfn = NULL;
1385 h->sh.weight[0][1].i_denom = h->sh.weight[0][2].i_denom = 0;
1386 for( int j = 0; j < h->i_ref0; j++ )
1388 if( h->fenc->weight[j][0].weightfn )
1390 h->sh.weight[j][0] = h->fenc->weight[j][0];
1391 // if weight is useless, don't write it to stream
1392 if( h->sh.weight[j][0].i_scale == 1<<h->sh.weight[j][0].i_denom && h->sh.weight[j][0].i_offset == 0 )
1393 h->sh.weight[j][0].weightfn = NULL;
1399 h->sh.weight[0][0].i_denom = denom = h->sh.weight[j][0].i_denom;
1400 assert( x264_clip3( denom, 0, 7 ) == denom );
1402 assert( h->sh.weight[j][0].i_denom == denom );
1403 assert( x264_clip3( h->sh.weight[j][0].i_scale, 0, 127 ) == h->sh.weight[j][0].i_scale );
1404 assert( x264_clip3( h->sh.weight[j][0].i_offset, -128, 127 ) == h->sh.weight[j][0].i_offset );
1405 h->fenc->weighted[j] = h->mb.p_weight_buf[buffer_next++] +
1406 h->fenc->i_stride[0] * i_padv + PADH;
1410 //scale full resolution frame
1411 if( h->sh.weight[j][0].weightfn && h->param.i_threads == 1 )
1413 uint8_t *src = h->fref0[j]->filtered[0] - h->fref0[j]->i_stride[0]*i_padv - PADH;
1414 uint8_t *dst = h->fenc->weighted[j] - h->fenc->i_stride[0]*i_padv - PADH;
1415 int stride = h->fenc->i_stride[0];
1416 int width = h->fenc->i_width[0] + PADH*2;
1417 int height = h->fenc->i_lines[0] + i_padv*2;
1418 x264_weight_scale_plane( h, dst, stride, src, stride, width, height, &h->sh.weight[j][0] );
1419 h->fenc->i_lines_weighted = height;
1423 h->sh.weight[0][0].i_denom = 0;
1426 static inline void x264_reference_build_list( x264_t *h, int i_poc )
1430 /* build ref list 0/1 */
1431 h->mb.pic.i_fref[0] = h->i_ref0 = 0;
1432 h->mb.pic.i_fref[1] = h->i_ref1 = 0;
1433 if( h->sh.i_type == SLICE_TYPE_I )
1436 for( int i = 0; h->frames.reference[i]; i++ )
1438 if( h->frames.reference[i]->i_poc < i_poc )
1439 h->fref0[h->i_ref0++] = h->frames.reference[i];
1440 else if( h->frames.reference[i]->i_poc > i_poc )
1441 h->fref1[h->i_ref1++] = h->frames.reference[i];
1444 /* Order ref0 from higher to lower poc */
1448 for( int i = 0; i < h->i_ref0 - 1; i++ )
1450 if( h->fref0[i]->i_poc < h->fref0[i+1]->i_poc )
1452 XCHG( x264_frame_t*, h->fref0[i], h->fref0[i+1] );
1459 if( h->sh.i_mmco_remove_from_end )
1460 for( int i = h->i_ref0-1; i >= h->i_ref0 - h->sh.i_mmco_remove_from_end; i-- )
1462 int diff = h->i_frame_num - h->fref0[i]->i_frame_num;
1463 h->sh.mmco[h->sh.i_mmco_command_count].i_poc = h->fref0[i]->i_poc;
1464 h->sh.mmco[h->sh.i_mmco_command_count++].i_difference_of_pic_nums = diff;
1467 /* Order ref1 from lower to higher poc (bubble sort) for B-frame */
1471 for( int i = 0; i < h->i_ref1 - 1; i++ )
1473 if( h->fref1[i]->i_poc > h->fref1[i+1]->i_poc )
1475 XCHG( x264_frame_t*, h->fref1[i], h->fref1[i+1] );
1482 x264_reference_check_reorder( h );
1484 h->i_ref1 = X264_MIN( h->i_ref1, h->frames.i_max_ref1 );
1485 h->i_ref0 = X264_MIN( h->i_ref0, h->frames.i_max_ref0 );
1486 h->i_ref0 = X264_MIN( h->i_ref0, h->param.i_frame_reference ); // if reconfig() has lowered the limit
1488 /* add duplicates */
1489 if( h->fenc->i_type == X264_TYPE_P )
1492 if( h->param.analyse.i_weighted_pred == X264_WEIGHTP_SMART )
1495 w[1].weightfn = w[2].weightfn = NULL;
1496 if( h->param.rc.b_stat_read )
1497 x264_ratecontrol_set_weights( h, h->fenc );
1499 if( !h->fenc->weight[0][0].weightfn )
1501 h->fenc->weight[0][0].i_denom = 0;
1502 SET_WEIGHT( w[0], 1, 1, 0, -1 );
1503 idx = x264_weighted_reference_duplicate( h, 0, w );
1507 if( h->fenc->weight[0][0].i_scale == 1<<h->fenc->weight[0][0].i_denom )
1509 SET_WEIGHT( h->fenc->weight[0][0], 1, 1, 0, h->fenc->weight[0][0].i_offset );
1511 x264_weighted_reference_duplicate( h, 0, weight_none );
1512 if( h->fenc->weight[0][0].i_offset > -128 )
1514 w[0] = h->fenc->weight[0][0];
1516 h->mc.weight_cache( h, &w[0] );
1517 idx = x264_weighted_reference_duplicate( h, 0, w );
1521 else if( h->param.analyse.i_weighted_pred == X264_WEIGHTP_BLIND )
1523 //weighted offset=-1
1525 SET_WEIGHT( w[0], 1, 1, 0, -1 );
1526 h->fenc->weight[0][0].i_denom = 0;
1527 w[1].weightfn = w[2].weightfn = NULL;
1528 idx = x264_weighted_reference_duplicate( h, 0, w );
1530 h->mb.ref_blind_dupe = idx;
1533 assert( h->i_ref0 + h->i_ref1 <= 16 );
1534 h->mb.pic.i_fref[0] = h->i_ref0;
1535 h->mb.pic.i_fref[1] = h->i_ref1;
1538 static void x264_fdec_filter_row( x264_t *h, int mb_y, int b_inloop )
1540 /* mb_y is the mb to be encoded next, not the mb to be filtered here */
1541 int b_hpel = h->fdec->b_kept_as_ref;
1542 int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
1543 int b_end = mb_y == h->i_threadslice_end;
1544 int b_measure_quality = 1;
1545 int min_y = mb_y - (1 << h->sh.b_mbaff);
1546 int b_start = min_y == h->i_threadslice_start;
1547 int max_y = b_end ? h->i_threadslice_end : mb_y;
1548 b_deblock &= b_hpel || h->param.psz_dump_yuv;
1549 if( h->param.b_sliced_threads && b_start && min_y && !b_inloop )
1551 b_deblock = 0; /* We already deblocked on the inloop pass. */
1552 b_measure_quality = 0; /* We already measured quality on the inloop pass. */
1554 if( mb_y & h->sh.b_mbaff )
1556 if( min_y < h->i_threadslice_start )
1559 if( !b_end && b_inloop )
1560 for( int j = 0; j <= h->sh.b_mbaff; j++ )
1561 for( int i = 0; i < 3; i++ )
1563 memcpy( h->intra_border_backup[j][i],
1564 h->fdec->plane[i] + ((mb_y*16 >> !!i) + j - 1 - h->sh.b_mbaff) * h->fdec->i_stride[i],
1565 h->sps->i_mb_width*16 >> !!i );
1569 for( int y = min_y; y < max_y; y += (1 << h->sh.b_mbaff) )
1570 x264_frame_deblock_row( h, y );
1574 int end = mb_y == h->sps->i_mb_height;
1575 x264_frame_expand_border( h, h->fdec, min_y, end );
1576 if( h->param.analyse.i_subpel_refine )
1578 x264_frame_filter( h, h->fdec, min_y, end );
1579 x264_frame_expand_border_filtered( h, h->fdec, min_y, end );
1583 if( h->i_thread_frames > 1 && h->fdec->b_kept_as_ref )
1584 x264_frame_cond_broadcast( h->fdec, mb_y*16 + (b_end ? 10000 : -(X264_THREAD_HEIGHT << h->sh.b_mbaff)) );
1586 min_y = min_y*16 - 8 * !b_start;
1587 max_y = b_end ? X264_MIN( h->i_threadslice_end*16 , h->param.i_height ) : mb_y*16 - 8;
1589 if( b_measure_quality )
1591 if( h->param.analyse.b_psnr )
1592 for( int i = 0; i < 3; i++ )
1593 h->stat.frame.i_ssd[i] +=
1594 x264_pixel_ssd_wxh( &h->pixf,
1595 h->fdec->plane[i] + (min_y>>!!i) * h->fdec->i_stride[i], h->fdec->i_stride[i],
1596 h->fenc->plane[i] + (min_y>>!!i) * h->fenc->i_stride[i], h->fenc->i_stride[i],
1597 h->param.i_width >> !!i, (max_y-min_y) >> !!i );
1599 if( h->param.analyse.b_ssim )
1602 /* offset by 2 pixels to avoid alignment of ssim blocks with dct blocks,
1603 * and overlap by 4 */
1604 min_y += b_start ? 2 : -6;
1605 h->stat.frame.f_ssim +=
1606 x264_pixel_ssim_wxh( &h->pixf,
1607 h->fdec->plane[0] + 2+min_y*h->fdec->i_stride[0], h->fdec->i_stride[0],
1608 h->fenc->plane[0] + 2+min_y*h->fenc->i_stride[0], h->fenc->i_stride[0],
1609 h->param.i_width-2, max_y-min_y, h->scratch_buffer );
1614 static inline int x264_reference_update( x264_t *h )
1616 if( !h->fdec->b_kept_as_ref )
1618 if( h->i_thread_frames > 1 )
1620 x264_frame_push_unused( h, h->fdec );
1621 h->fdec = x264_frame_pop_unused( h, 1 );
1628 /* apply mmco from previous frame. */
1629 for( int i = 0; i < h->sh.i_mmco_command_count; i++ )
1630 for( int j = 0; h->frames.reference[j]; j++ )
1631 if( h->frames.reference[j]->i_poc == h->sh.mmco[i].i_poc )
1632 x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[j] ) );
1634 /* move frame in the buffer */
1635 x264_frame_push( h->frames.reference, h->fdec );
1636 if( h->frames.reference[h->sps->i_num_ref_frames] )
1637 x264_frame_push_unused( h, x264_frame_shift( h->frames.reference ) );
1638 h->fdec = x264_frame_pop_unused( h, 1 );
1644 static inline void x264_reference_reset( x264_t *h )
1646 while( h->frames.reference[0] )
1647 x264_frame_push_unused( h, x264_frame_pop( h->frames.reference ) );
1652 static inline void x264_reference_hierarchy_reset( x264_t *h )
1655 int b_hasdelayframe = 0;
1656 if( !h->param.i_bframe_pyramid )
1659 /* look for delay frames -- chain must only contain frames that are disposable */
1660 for( int i = 0; h->frames.current[i] && IS_DISPOSABLE( h->frames.current[i]->i_type ); i++ )
1661 b_hasdelayframe |= h->frames.current[i]->i_coded
1662 != h->frames.current[i]->i_frame + h->sps->vui.i_num_reorder_frames;
1664 if( h->param.i_bframe_pyramid != X264_B_PYRAMID_STRICT && !b_hasdelayframe )
1667 /* Remove last BREF. There will never be old BREFs in the
1668 * dpb during a BREF decode when pyramid == STRICT */
1669 for( ref = 0; h->frames.reference[ref]; ref++ )
1671 if( h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT
1672 && h->frames.reference[ref]->i_type == X264_TYPE_BREF )
1674 int diff = h->i_frame_num - h->frames.reference[ref]->i_frame_num;
1675 h->sh.mmco[h->sh.i_mmco_command_count].i_difference_of_pic_nums = diff;
1676 h->sh.mmco[h->sh.i_mmco_command_count++].i_poc = h->frames.reference[ref]->i_poc;
1677 x264_frame_push_unused( h, x264_frame_pop( h->frames.reference ) );
1678 h->b_ref_reorder[0] = 1;
1683 /* Prepare to room in the dpb for the delayed display time of the later b-frame's */
1684 h->sh.i_mmco_remove_from_end = X264_MAX( ref + 2 - h->frames.i_max_dpb, 0 );
1687 static inline void x264_slice_init( x264_t *h, int i_nal_type, int i_global_qp )
1689 /* ------------------------ Create slice header ----------------------- */
1690 if( i_nal_type == NAL_SLICE_IDR )
1692 x264_slice_header_init( h, &h->sh, h->sps, h->pps, h->i_idr_pic_id, h->i_frame_num, i_global_qp );
1695 h->i_idr_pic_id = ( h->i_idr_pic_id + 1 ) % 65536;
1699 x264_slice_header_init( h, &h->sh, h->sps, h->pps, -1, h->i_frame_num, i_global_qp );
1701 h->sh.i_num_ref_idx_l0_active = h->i_ref0 <= 0 ? 1 : h->i_ref0;
1702 h->sh.i_num_ref_idx_l1_active = h->i_ref1 <= 0 ? 1 : h->i_ref1;
1703 if( h->sh.i_num_ref_idx_l0_active != h->pps->i_num_ref_idx_l0_default_active ||
1704 (h->sh.i_type == SLICE_TYPE_B && h->sh.i_num_ref_idx_l1_active != h->pps->i_num_ref_idx_l1_default_active) )
1706 h->sh.b_num_ref_idx_override = 1;
1710 h->fdec->i_frame_num = h->sh.i_frame_num;
1712 if( h->sps->i_poc_type == 0 )
1714 h->sh.i_poc = h->fdec->i_poc;
1715 if( h->param.b_interlaced )
1717 h->sh.i_delta_poc_bottom = h->param.b_tff ? 1 : -1;
1718 if( h->sh.i_delta_poc_bottom == -1 )
1719 h->sh.i_poc = h->fdec->i_poc + 1;
1722 h->sh.i_delta_poc_bottom = 0;
1724 else if( h->sps->i_poc_type == 1 )
1726 /* FIXME TODO FIXME */
1730 /* Nothing to do ? */
1733 x264_macroblock_slice_init( h );
1736 static int x264_slice_write( x264_t *h )
1739 int mb_xy, i_mb_x, i_mb_y;
1740 int i_skip_bak = 0; /* Shut up GCC. */
1742 x264_cabac_t cabac_bak;
1743 uint8_t cabac_prevbyte_bak = 0; /* Shut up GCC. */
1744 /* Assume no more than 3 bytes of NALU escaping.
1745 * NALUs other than the first use a 3-byte startcode. */
1746 int overhead_guess = (NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal)) + 3;
1747 int slice_max_size = h->param.i_slice_max_size > 0 ? (h->param.i_slice_max_size-overhead_guess)*8 : INT_MAX;
1748 int starting_bits = bs_pos(&h->out.bs);
1749 bs_realign( &h->out.bs );
1752 x264_nal_start( h, h->i_nal_type, h->i_nal_ref_idc );
1755 x264_macroblock_thread_init( h );
1757 /* If this isn't the first slice in the threadslice, set the slice QP
1758 * equal to the last QP in the previous slice for more accurate
1759 * CABAC initialization. */
1760 if( h->sh.i_first_mb != h->i_threadslice_start * h->sps->i_mb_width )
1762 h->sh.i_qp = h->mb.i_last_qp;
1763 h->sh.i_qp_delta = h->sh.i_qp - h->pps->i_pic_init_qp;
1766 x264_slice_header_write( &h->out.bs, &h->sh, h->i_nal_ref_idc );
1767 if( h->param.b_cabac )
1769 /* alignment needed */
1770 bs_align_1( &h->out.bs );
1773 x264_cabac_context_init( &h->cabac, h->sh.i_type, h->sh.i_qp, h->sh.i_cabac_init_idc );
1774 x264_cabac_encode_init ( &h->cabac, h->out.bs.p, h->out.bs.p_end );
1776 h->mb.i_last_qp = h->sh.i_qp;
1777 h->mb.i_last_dqp = 0;
1779 i_mb_y = h->sh.i_first_mb / h->sps->i_mb_width;
1780 i_mb_x = h->sh.i_first_mb % h->sps->i_mb_width;
1783 while( (mb_xy = i_mb_x + i_mb_y * h->sps->i_mb_width) <= h->sh.i_last_mb )
1785 int mb_spos = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
1786 if( h->param.i_slice_max_size > 0 )
1788 /* We don't need the contexts because flushing the CABAC encoder has no context
1789 * dependency and macroblocks are only re-encoded in the case where a slice is
1790 * ended (and thus the content of all contexts are thrown away). */
1791 if( h->param.b_cabac )
1793 memcpy( &cabac_bak, &h->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
1794 /* x264's CABAC writer modifies the previous byte during carry, so it has to be
1796 cabac_prevbyte_bak = h->cabac.p[-1];
1801 i_skip_bak = i_skip;
1805 if( i_mb_x == 0 && !h->mb.b_reencode_mb )
1806 x264_fdec_filter_row( h, i_mb_y, 1 );
1809 x264_macroblock_cache_load( h, i_mb_x, i_mb_y );
1811 x264_macroblock_analyse( h );
1813 /* encode this macroblock -> be careful it can change the mb type to P_SKIP if needed */
1814 x264_macroblock_encode( h );
1816 if( x264_bitstream_check_buffer( h ) )
1819 if( h->param.b_cabac )
1821 if( mb_xy > h->sh.i_first_mb && !(h->sh.b_mbaff && (i_mb_y&1)) )
1822 x264_cabac_encode_terminal( &h->cabac );
1824 if( IS_SKIP( h->mb.i_type ) )
1825 x264_cabac_mb_skip( h, 1 );
1828 if( h->sh.i_type != SLICE_TYPE_I )
1829 x264_cabac_mb_skip( h, 0 );
1830 x264_macroblock_write_cabac( h, &h->cabac );
1835 if( IS_SKIP( h->mb.i_type ) )
1839 if( h->sh.i_type != SLICE_TYPE_I )
1841 bs_write_ue( &h->out.bs, i_skip ); /* skip run */
1844 x264_macroblock_write_cavlc( h );
1848 int total_bits = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
1849 int mb_size = total_bits - mb_spos;
1851 /* We'll just re-encode this last macroblock if we go over the max slice size. */
1852 if( total_bits - starting_bits > slice_max_size && !h->mb.b_reencode_mb )
1854 if( mb_xy != h->sh.i_first_mb )
1856 if( h->param.b_cabac )
1858 memcpy( &h->cabac, &cabac_bak, offsetof(x264_cabac_t, f8_bits_encoded) );
1859 h->cabac.p[-1] = cabac_prevbyte_bak;
1864 i_skip = i_skip_bak;
1866 h->mb.b_reencode_mb = 1;
1867 h->sh.i_last_mb = mb_xy-1;
1872 h->sh.i_last_mb = mb_xy;
1873 h->mb.b_reencode_mb = 0;
1877 h->mb.b_reencode_mb = 0;
1879 #ifdef HAVE_VISUALIZE
1880 if( h->param.b_visualize )
1881 x264_visualize_mb( h );
1885 x264_macroblock_cache_save( h );
1887 /* accumulate mb stats */
1889 int b_intra = IS_INTRA( h->mb.i_type );
1890 if( h->param.i_log_level >= X264_LOG_INFO || h->param.rc.b_stat_write )
1892 h->stat.frame.i_mb_count[h->mb.i_type]++;
1893 if( !b_intra && !IS_SKIP( h->mb.i_type ) && !IS_DIRECT( h->mb.i_type ) )
1895 if( h->mb.i_partition != D_8x8 )
1896 h->stat.frame.i_mb_partition[h->mb.i_partition] += 4;
1898 for( int i = 0; i < 4; i++ )
1899 h->stat.frame.i_mb_partition[h->mb.i_sub_partition[i]] ++;
1900 if( h->param.i_frame_reference > 1 )
1901 for( int i_list = 0; i_list <= (h->sh.i_type == SLICE_TYPE_B); i_list++ )
1902 for( int i = 0; i < 4; i++ )
1904 int i_ref = h->mb.cache.ref[i_list][ x264_scan8[4*i] ];
1906 h->stat.frame.i_mb_count_ref[i_list][i_ref] ++;
1911 if( h->param.i_log_level >= X264_LOG_INFO )
1913 if( h->mb.i_cbp_luma | h->mb.i_cbp_chroma )
1915 int cbpsum = (h->mb.i_cbp_luma&1) + ((h->mb.i_cbp_luma>>1)&1)
1916 + ((h->mb.i_cbp_luma>>2)&1) + (h->mb.i_cbp_luma>>3);
1917 h->stat.frame.i_mb_cbp[!b_intra + 0] += cbpsum;
1918 h->stat.frame.i_mb_cbp[!b_intra + 2] += !!h->mb.i_cbp_chroma;
1919 h->stat.frame.i_mb_cbp[!b_intra + 4] += h->mb.i_cbp_chroma >> 1;
1921 if( h->mb.i_cbp_luma && !b_intra )
1923 h->stat.frame.i_mb_count_8x8dct[0] ++;
1924 h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
1926 if( b_intra && h->mb.i_type != I_PCM )
1928 if( h->mb.i_type == I_16x16 )
1929 h->stat.frame.i_mb_pred_mode[0][h->mb.i_intra16x16_pred_mode]++;
1930 else if( h->mb.i_type == I_8x8 )
1931 for( int i = 0; i < 16; i += 4 )
1932 h->stat.frame.i_mb_pred_mode[1][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
1933 else //if( h->mb.i_type == I_4x4 )
1934 for( int i = 0; i < 16; i++ )
1935 h->stat.frame.i_mb_pred_mode[2][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
1936 h->stat.frame.i_mb_pred_mode[3][x264_mb_pred_mode8x8c_fix[h->mb.i_chroma_pred_mode]]++;
1940 x264_ratecontrol_mb( h, mb_size );
1944 i_mb_x += i_mb_y & 1;
1945 i_mb_y ^= i_mb_x < h->sps->i_mb_width;
1949 if( i_mb_x == h->sps->i_mb_width )
1956 if( h->param.b_cabac )
1958 x264_cabac_encode_flush( h, &h->cabac );
1959 h->out.bs.p = h->cabac.p;
1964 bs_write_ue( &h->out.bs, i_skip ); /* last skip run */
1965 /* rbsp_slice_trailing_bits */
1966 bs_rbsp_trailing( &h->out.bs );
1967 bs_flush( &h->out.bs );
1969 if( x264_nal_end( h ) )
1972 if( h->sh.i_last_mb == (h->i_threadslice_end * h->sps->i_mb_width - 1) )
1974 h->stat.frame.i_misc_bits = bs_pos( &h->out.bs )
1975 + (h->out.i_nal*NALU_OVERHEAD * 8)
1976 - h->stat.frame.i_tex_bits
1977 - h->stat.frame.i_mv_bits;
1978 x264_fdec_filter_row( h, h->i_threadslice_end, 1 );
1984 static void x264_thread_sync_context( x264_t *dst, x264_t *src )
1989 // reference counting
1990 for( x264_frame_t **f = src->frames.reference; *f; f++ )
1991 (*f)->i_reference_count++;
1992 for( x264_frame_t **f = dst->frames.reference; *f; f++ )
1993 x264_frame_push_unused( src, *f );
1994 src->fdec->i_reference_count++;
1995 x264_frame_push_unused( src, dst->fdec );
1997 // copy everything except the per-thread pointers and the constants.
1998 memcpy( &dst->i_frame, &src->i_frame, offsetof(x264_t, mb.type) - offsetof(x264_t, i_frame) );
1999 dst->param = src->param;
2000 dst->stat = src->stat;
2003 static void x264_thread_sync_stat( x264_t *dst, x264_t *src )
2007 memcpy( &dst->stat.i_frame_count, &src->stat.i_frame_count, sizeof(dst->stat) - sizeof(dst->stat.frame) );
2010 static void *x264_slices_write( x264_t *h )
2012 int i_slice_num = 0;
2013 int last_thread_mb = h->sh.i_last_mb;
2014 if( h->param.i_sync_lookahead )
2015 x264_lower_thread_priority( 10 );
2018 /* Misalign mask has to be set separately for each thread. */
2019 if( h->param.cpu&X264_CPU_SSE_MISALIGN )
2020 x264_cpu_mask_misalign_sse();
2023 #ifdef HAVE_VISUALIZE
2024 if( h->param.b_visualize )
2025 if( x264_visualize_init( h ) )
2030 memset( &h->stat.frame, 0, sizeof(h->stat.frame) );
2031 h->mb.b_reencode_mb = 0;
2032 while( h->sh.i_first_mb <= last_thread_mb )
2034 h->sh.i_last_mb = last_thread_mb;
2035 if( h->param.i_slice_max_mbs )
2036 h->sh.i_last_mb = h->sh.i_first_mb + h->param.i_slice_max_mbs - 1;
2037 else if( h->param.i_slice_count && !h->param.b_sliced_threads )
2039 int height = h->sps->i_mb_height >> h->param.b_interlaced;
2040 int width = h->sps->i_mb_width << h->param.b_interlaced;
2042 h->sh.i_last_mb = (height * i_slice_num + h->param.i_slice_count/2) / h->param.i_slice_count * width - 1;
2044 h->sh.i_last_mb = X264_MIN( h->sh.i_last_mb, last_thread_mb );
2045 if( x264_stack_align( x264_slice_write, h ) )
2047 h->sh.i_first_mb = h->sh.i_last_mb + 1;
2050 #ifdef HAVE_VISUALIZE
2051 if( h->param.b_visualize )
2053 x264_visualize_show( h );
2054 x264_visualize_close( h );
2061 static int x264_threaded_slices_write( x264_t *h )
2064 /* set first/last mb and sync contexts */
2065 for( int i = 0; i < h->param.i_threads; i++ )
2067 x264_t *t = h->thread[i];
2070 t->param = h->param;
2071 memcpy( &t->i_frame, &h->i_frame, offsetof(x264_t, rc) - offsetof(x264_t, i_frame) );
2073 int height = h->sps->i_mb_height >> h->param.b_interlaced;
2074 t->i_threadslice_start = ((height * i + h->param.i_slice_count/2) / h->param.i_threads) << h->param.b_interlaced;
2075 t->i_threadslice_end = ((height * (i+1) + h->param.i_slice_count/2) / h->param.i_threads) << h->param.b_interlaced;
2076 t->sh.i_first_mb = t->i_threadslice_start * h->sps->i_mb_width;
2077 t->sh.i_last_mb = t->i_threadslice_end * h->sps->i_mb_width - 1;
2080 x264_stack_align( x264_analyse_weight_frame, h, h->sps->i_mb_height*16 + 16 );
2082 x264_threads_distribute_ratecontrol( h );
2085 for( int i = 0; i < h->param.i_threads; i++ )
2087 if( x264_pthread_create( &h->thread[i]->thread_handle, NULL, (void*)x264_slices_write, (void*)h->thread[i] ) )
2089 h->thread[i]->b_thread_active = 1;
2091 for( int i = 0; i < h->param.i_threads; i++ )
2093 x264_pthread_join( h->thread[i]->thread_handle, &ret );
2094 h->thread[i]->b_thread_active = 0;
2096 return (intptr_t)ret;
2099 /* Go back and fix up the hpel on the borders between slices. */
2100 for( int i = 1; i < h->param.i_threads; i++ )
2101 x264_fdec_filter_row( h->thread[i], h->thread[i]->i_threadslice_start + 1, 0 );
2103 x264_threads_merge_ratecontrol( h );
2105 for( int i = 1; i < h->param.i_threads; i++ )
2107 x264_t *t = h->thread[i];
2108 for( int j = 0; j < t->out.i_nal; j++ )
2110 h->out.nal[h->out.i_nal] = t->out.nal[j];
2112 x264_nal_check_buffer( h );
2114 /* All entries in stat.frame are ints except for ssd/ssim. */
2115 for( int j = 0; j < (offsetof(x264_t,stat.frame.i_ssd) - offsetof(x264_t,stat.frame.i_mv_bits)) / sizeof(int); j++ )
2116 ((int*)&h->stat.frame)[j] += ((int*)&t->stat.frame)[j];
2117 for( int j = 0; j < 3; j++ )
2118 h->stat.frame.i_ssd[j] += t->stat.frame.i_ssd[j];
2119 h->stat.frame.f_ssim += t->stat.frame.f_ssim;
2125 /****************************************************************************
2126 * x264_encoder_encode:
2127 * XXX: i_poc : is the poc of the current given picture
2128 * i_frame : is the number of the frame being coded
2129 * ex: type frame poc
2137 ****************************************************************************/
2138 int x264_encoder_encode( x264_t *h,
2139 x264_nal_t **pp_nal, int *pi_nal,
2140 x264_picture_t *pic_in,
2141 x264_picture_t *pic_out )
2143 x264_t *thread_current, *thread_prev, *thread_oldest;
2144 int i_nal_type, i_nal_ref_idc, i_global_qp;
2145 int overhead = NALU_OVERHEAD;
2147 if( h->i_thread_frames > 1 )
2149 thread_prev = h->thread[ h->i_thread_phase ];
2150 h->i_thread_phase = (h->i_thread_phase + 1) % h->i_thread_frames;
2151 thread_current = h->thread[ h->i_thread_phase ];
2152 thread_oldest = h->thread[ (h->i_thread_phase + 1) % h->i_thread_frames ];
2153 x264_thread_sync_context( thread_current, thread_prev );
2154 x264_thread_sync_ratecontrol( thread_current, thread_prev, thread_oldest );
2163 // ok to call this before encoding any frames, since the initial values of fdec have b_kept_as_ref=0
2164 if( x264_reference_update( h ) )
2166 h->fdec->i_lines_completed = -1;
2172 /* ------------------- Setup new frame from picture -------------------- */
2173 if( pic_in != NULL )
2175 /* 1: Copy the picture to a frame and move it to a buffer */
2176 x264_frame_t *fenc = x264_frame_pop_unused( h, 0 );
2180 if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 )
2183 if( h->param.i_width != 16 * h->sps->i_mb_width ||
2184 h->param.i_height != 16 * h->sps->i_mb_height )
2185 x264_frame_expand_border_mod16( h, fenc );
2187 fenc->i_frame = h->frames.i_input++;
2189 if( h->frames.i_bframe_delay && fenc->i_frame == h->frames.i_bframe_delay )
2190 h->frames.i_bframe_delay_time = fenc->i_pts;
2192 if( h->param.b_vfr_input && fenc->i_pts <= h->frames.i_largest_pts )
2193 x264_log( h, X264_LOG_WARNING, "non-strictly-monotonic PTS\n" );
2195 h->frames.i_second_largest_pts = h->frames.i_largest_pts;
2196 h->frames.i_largest_pts = fenc->i_pts;
2198 if( (fenc->i_pic_struct < PIC_STRUCT_AUTO) || (fenc->i_pic_struct > PIC_STRUCT_TRIPLE) )
2199 fenc->i_pic_struct = PIC_STRUCT_AUTO;
2201 if( fenc->i_pic_struct == PIC_STRUCT_AUTO )
2203 int b_interlaced = fenc->param ? fenc->param->b_interlaced : h->param.b_interlaced;
2206 int b_tff = fenc->param ? fenc->param->b_tff : h->param.b_tff;
2207 fenc->i_pic_struct = b_tff ? PIC_STRUCT_TOP_BOTTOM : PIC_STRUCT_BOTTOM_TOP;
2210 fenc->i_pic_struct = PIC_STRUCT_PROGRESSIVE;
2213 if( h->frames.b_have_lowres )
2215 if( h->param.analyse.i_weighted_pred == X264_WEIGHTP_FAKE || h->param.analyse.i_weighted_pred == X264_WEIGHTP_SMART )
2216 x264_weight_plane_analyse( h, fenc );
2217 x264_frame_init_lowres( h, fenc );
2220 if( h->param.rc.b_mb_tree && h->param.rc.b_stat_read )
2222 if( x264_macroblock_tree_read( h, fenc ) )
2225 else if( h->param.rc.i_aq_mode )
2226 x264_adaptive_quant_frame( h, fenc );
2228 /* 2: Place the frame into the queue for its slice type decision */
2229 x264_lookahead_put_frame( h, fenc );
2231 if( h->frames.i_input <= h->frames.i_delay + 1 - h->i_thread_frames )
2233 /* Nothing yet to encode, waiting for filling of buffers */
2234 pic_out->i_type = X264_TYPE_AUTO;
2240 /* signal kills for lookahead thread */
2241 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
2242 h->lookahead->b_exit_thread = 1;
2243 x264_pthread_cond_broadcast( &h->lookahead->ifbuf.cv_fill );
2244 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
2248 /* 3: The picture is analyzed in the lookahead */
2249 if( !h->frames.current[0] )
2250 x264_lookahead_get_frames( h );
2252 if( !h->frames.current[0] && x264_lookahead_is_empty( h ) )
2253 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
2255 /* ------------------- Get frame to be encoded ------------------------- */
2256 /* 4: get picture to encode */
2257 h->fenc = x264_frame_shift( h->frames.current );
2258 if( h->i_frame == 0 )
2259 h->first_pts = h->fenc->i_reordered_pts;
2260 if( h->fenc->param )
2262 x264_encoder_reconfig( h, h->fenc->param );
2263 if( h->fenc->param->param_free )
2264 h->fenc->param->param_free( h->fenc->param );
2267 if( h->fenc->b_keyframe )
2269 h->frames.i_last_keyframe = h->fenc->i_frame;
2270 if( h->fenc->i_type == X264_TYPE_IDR )
2273 h->sh.i_mmco_command_count =
2274 h->sh.i_mmco_remove_from_end = 0;
2275 h->b_ref_reorder[0] =
2276 h->b_ref_reorder[1] = 0;
2278 /* ------------------- Setup frame context ----------------------------- */
2279 /* 5: Init data dependent of frame type */
2280 if( h->fenc->i_type == X264_TYPE_IDR )
2282 /* reset ref pictures */
2283 i_nal_type = NAL_SLICE_IDR;
2284 i_nal_ref_idc = NAL_PRIORITY_HIGHEST;
2285 h->sh.i_type = SLICE_TYPE_I;
2286 x264_reference_reset( h );
2288 else if( h->fenc->i_type == X264_TYPE_I )
2290 i_nal_type = NAL_SLICE;
2291 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
2292 h->sh.i_type = SLICE_TYPE_I;
2293 x264_reference_hierarchy_reset( h );
2295 else if( h->fenc->i_type == X264_TYPE_P )
2297 i_nal_type = NAL_SLICE;
2298 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
2299 h->sh.i_type = SLICE_TYPE_P;
2300 x264_reference_hierarchy_reset( h );
2302 else if( h->fenc->i_type == X264_TYPE_BREF )
2304 i_nal_type = NAL_SLICE;
2305 i_nal_ref_idc = h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT ? NAL_PRIORITY_LOW : NAL_PRIORITY_HIGH;
2306 h->sh.i_type = SLICE_TYPE_B;
2307 x264_reference_hierarchy_reset( h );
2311 i_nal_type = NAL_SLICE;
2312 i_nal_ref_idc = NAL_PRIORITY_DISPOSABLE;
2313 h->sh.i_type = SLICE_TYPE_B;
2317 h->fenc->i_poc = 2 * (h->fenc->i_frame - h->frames.i_last_keyframe);
2318 h->fdec->i_type = h->fenc->i_type;
2319 h->fdec->i_frame = h->fenc->i_frame;
2320 h->fenc->b_kept_as_ref =
2321 h->fdec->b_kept_as_ref = i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE && h->param.i_keyint_max > 1;
2325 /* ------------------- Init ----------------------------- */
2326 /* build ref list 0/1 */
2327 x264_reference_build_list( h, h->fdec->i_poc );
2329 /* ---------------------- Write the bitstream -------------------------- */
2330 /* Init bitstream context */
2331 if( h->param.b_sliced_threads )
2333 for( int i = 0; i < h->param.i_threads; i++ )
2335 bs_init( &h->thread[i]->out.bs, h->thread[i]->out.p_bitstream, h->thread[i]->out.i_bitstream );
2336 h->thread[i]->out.i_nal = 0;
2341 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
2345 if( h->param.b_aud )
2349 if( h->sh.i_type == SLICE_TYPE_I )
2351 else if( h->sh.i_type == SLICE_TYPE_P )
2353 else if( h->sh.i_type == SLICE_TYPE_B )
2358 x264_nal_start( h, NAL_AUD, NAL_PRIORITY_DISPOSABLE );
2359 bs_write( &h->out.bs, 3, pic_type );
2360 bs_rbsp_trailing( &h->out.bs );
2361 if( x264_nal_end( h ) )
2363 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
2366 h->i_nal_type = i_nal_type;
2367 h->i_nal_ref_idc = i_nal_ref_idc;
2369 if( h->param.b_intra_refresh && h->fenc->i_type == X264_TYPE_P )
2371 int pocdiff = (h->fdec->i_poc - h->fref0[0]->i_poc)/2;
2372 float increment = X264_MAX( ((float)h->sps->i_mb_width-1) / h->param.i_keyint_max, 1 );
2373 int max_position = (int)(increment * h->param.i_keyint_max);
2374 if( IS_X264_TYPE_I( h->fref0[0]->i_type ) )
2375 h->fdec->f_pir_position = 0;
2378 h->fdec->f_pir_position = h->fref0[0]->f_pir_position;
2379 if( h->fdec->f_pir_position+0.5 >= max_position )
2381 h->fdec->f_pir_position = 0;
2382 h->fenc->b_keyframe = 1;
2385 h->fdec->i_pir_start_col = h->fdec->f_pir_position+0.5;
2386 h->fdec->f_pir_position += increment * pocdiff;
2387 h->fdec->i_pir_end_col = h->fdec->f_pir_position+0.5;
2390 if( h->fenc->b_keyframe )
2392 /* Write SPS and PPS */
2393 if( h->param.b_repeat_headers )
2395 /* generate sequence parameters */
2396 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
2397 x264_sps_write( &h->out.bs, h->sps );
2398 if( x264_nal_end( h ) )
2400 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
2402 /* generate picture parameters */
2403 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
2404 x264_pps_write( &h->out.bs, h->pps );
2405 if( x264_nal_end( h ) )
2407 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
2410 /* buffering period sei is written in x264_encoder_frame_end */
2412 if( h->param.b_repeat_headers && h->fenc->i_frame == 0 )
2414 /* identify ourself */
2415 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
2416 if( x264_sei_version_write( h, &h->out.bs ) )
2418 if( x264_nal_end( h ) )
2420 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
2423 if( h->fenc->i_type != X264_TYPE_IDR )
2425 int time_to_recovery = X264_MIN( h->sps->i_mb_width - 1, h->param.i_keyint_max ) + h->param.i_bframe;
2426 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
2427 x264_sei_recovery_point_write( h, &h->out.bs, time_to_recovery );
2429 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
2433 /* generate sei pic timing */
2434 if( h->sps->vui.b_pic_struct_present || h->sps->vui.b_nal_hrd_parameters_present )
2436 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
2437 x264_sei_pic_timing_write( h, &h->out.bs );
2438 if( x264_nal_end( h ) )
2440 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
2443 /* Init the rate control */
2444 /* FIXME: Include slice header bit cost. */
2445 x264_ratecontrol_start( h, h->fenc->i_qpplus1, overhead*8 );
2446 i_global_qp = x264_ratecontrol_qp( h );
2448 pic_out->i_qpplus1 =
2449 h->fdec->i_qpplus1 = i_global_qp + 1;
2451 if( h->param.rc.b_stat_read && h->sh.i_type != SLICE_TYPE_I )
2453 x264_reference_build_list_optimal( h );
2454 x264_reference_check_reorder( h );
2458 h->fdec->i_poc_l0ref0 = h->fref0[0]->i_poc;
2460 if( h->sh.i_type == SLICE_TYPE_B )
2461 x264_macroblock_bipred_init( h );
2463 /*------------------------- Weights -------------------------------------*/
2464 x264_weighted_pred_init( h );
2466 /* ------------------------ Create slice header ----------------------- */
2467 x264_slice_init( h, i_nal_type, i_global_qp );
2469 if( i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE )
2473 h->i_threadslice_start = 0;
2474 h->i_threadslice_end = h->sps->i_mb_height;
2475 if( h->i_thread_frames > 1 )
2477 if( x264_pthread_create( &h->thread_handle, NULL, (void*)x264_slices_write, h ) )
2479 h->b_thread_active = 1;
2481 else if( h->param.b_sliced_threads )
2483 if( x264_threaded_slices_write( h ) )
2487 if( (intptr_t)x264_slices_write( h ) )
2490 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
2493 static int x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
2494 x264_nal_t **pp_nal, int *pi_nal,
2495 x264_picture_t *pic_out )
2497 char psz_message[80];
2499 if( h->b_thread_active )
2502 x264_pthread_join( h->thread_handle, &ret );
2503 h->b_thread_active = 0;
2505 return (intptr_t)ret;
2509 pic_out->i_type = X264_TYPE_AUTO;
2514 /* generate sei buffering period and insert it into place */
2515 if( h->fenc->b_keyframe && h->sps->vui.b_nal_hrd_parameters_present )
2517 h->initial_cpb_removal_delay = x264_hrd_fullness( h );
2519 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
2520 x264_sei_buffering_period_write( h, &h->out.bs );
2521 if( x264_nal_end( h ) )
2523 /* buffering period sei must follow AUD, SPS and PPS and precede all other SEIs */
2525 while( h->out.nal[idx].i_type == NAL_AUD ||
2526 h->out.nal[idx].i_type == NAL_SPS ||
2527 h->out.nal[idx].i_type == NAL_PPS )
2529 x264_nal_t nal_tmp = h->out.nal[h->out.i_nal-1];
2530 memmove( &h->out.nal[idx+1], &h->out.nal[idx], (h->out.i_nal-idx-1)*sizeof(x264_nal_t) );
2531 h->out.nal[idx] = nal_tmp;
2534 int frame_size = x264_encoder_encapsulate_nals( h, 0 );
2536 /* Set output picture properties */
2537 pic_out->i_type = h->fenc->i_type;
2539 pic_out->b_keyframe = h->fenc->b_keyframe;
2541 pic_out->i_pts = h->fenc->i_pts *= h->i_dts_compress_multiplier;
2542 if( h->frames.i_bframe_delay )
2544 int64_t *prev_reordered_pts = thread_current->frames.i_prev_reordered_pts;
2545 if( h->i_frame <= h->frames.i_bframe_delay )
2547 if( h->i_dts_compress_multiplier == 1 )
2548 pic_out->i_dts = h->fenc->i_reordered_pts - h->frames.i_bframe_delay_time;
2551 /* DTS compression */
2552 if( h->i_frame == 1 )
2553 thread_current->frames.i_init_delta = h->fenc->i_reordered_pts * h->i_dts_compress_multiplier;
2554 pic_out->i_dts = h->i_frame * thread_current->frames.i_init_delta / h->i_dts_compress_multiplier;
2558 pic_out->i_dts = prev_reordered_pts[ (h->i_frame - h->frames.i_bframe_delay) % h->frames.i_bframe_delay ];
2559 prev_reordered_pts[ h->i_frame % h->frames.i_bframe_delay ] = h->fenc->i_reordered_pts * h->i_dts_compress_multiplier;
2562 pic_out->i_dts = h->fenc->i_reordered_pts;
2563 if( pic_out->i_pts < pic_out->i_dts )
2564 x264_log( h, X264_LOG_WARNING, "invalid DTS: PTS is less than DTS\n" );
2566 pic_out->img.i_plane = h->fdec->i_plane;
2567 for( int i = 0; i < 3; i++ )
2569 pic_out->img.i_stride[i] = h->fdec->i_stride[i];
2570 pic_out->img.plane[i] = h->fdec->plane[i];
2573 x264_frame_push_unused( thread_current, h->fenc );
2575 /* ---------------------- Update encoder state ------------------------- */
2579 if( x264_ratecontrol_end( h, frame_size * 8, &filler ) < 0 )
2582 pic_out->hrd_timing = h->fenc->hrd_timing;
2587 overhead = (FILLER_OVERHEAD - h->param.b_annexb);
2588 if( h->param.i_slice_max_size && filler > h->param.i_slice_max_size )
2590 int next_size = filler - h->param.i_slice_max_size;
2591 int overflow = X264_MAX( overhead - next_size, 0 );
2592 f = h->param.i_slice_max_size - overhead - overflow;
2595 f = X264_MAX( 0, filler - overhead );
2597 x264_nal_start( h, NAL_FILLER, NAL_PRIORITY_DISPOSABLE );
2598 x264_filler_write( h, &h->out.bs, f );
2599 if( x264_nal_end( h ) )
2601 int total_size = x264_encoder_encapsulate_nals( h, h->out.i_nal-1 );
2602 frame_size += total_size;
2603 filler -= total_size;
2606 /* End bitstream, set output */
2607 *pi_nal = h->out.i_nal;
2608 *pp_nal = h->out.nal;
2612 x264_noise_reduction_update( h );
2614 /* ---------------------- Compute/Print statistics --------------------- */
2615 x264_thread_sync_stat( h, h->thread[0] );
2618 h->stat.i_frame_count[h->sh.i_type]++;
2619 h->stat.i_frame_size[h->sh.i_type] += frame_size;
2620 h->stat.f_frame_qp[h->sh.i_type] += h->fdec->f_qp_avg_aq;
2622 for( int i = 0; i < X264_MBTYPE_MAX; i++ )
2623 h->stat.i_mb_count[h->sh.i_type][i] += h->stat.frame.i_mb_count[i];
2624 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
2625 h->stat.i_mb_partition[h->sh.i_type][i] += h->stat.frame.i_mb_partition[i];
2626 for( int i = 0; i < 2; i++ )
2627 h->stat.i_mb_count_8x8dct[i] += h->stat.frame.i_mb_count_8x8dct[i];
2628 for( int i = 0; i < 6; i++ )
2629 h->stat.i_mb_cbp[i] += h->stat.frame.i_mb_cbp[i];
2630 for( int i = 0; i < 4; i++ )
2631 for( int j = 0; j < 13; j++ )
2632 h->stat.i_mb_pred_mode[i][j] += h->stat.frame.i_mb_pred_mode[i][j];
2633 if( h->sh.i_type != SLICE_TYPE_I )
2634 for( int i_list = 0; i_list < 2; i_list++ )
2635 for( int i = 0; i < 32; i++ )
2636 h->stat.i_mb_count_ref[h->sh.i_type][i_list][i] += h->stat.frame.i_mb_count_ref[i_list][i];
2637 if( h->sh.i_type == SLICE_TYPE_P )
2639 h->stat.i_consecutive_bframes[h->fdec->i_frame - h->fref0[0]->i_frame - 1]++;
2640 if( h->param.analyse.i_weighted_pred == X264_WEIGHTP_SMART )
2641 for( int i = 0; i < 3; i++ )
2642 for( int j = 0; j < h->i_ref0; j++ )
2643 if( h->sh.weight[0][i].i_denom != 0 )
2645 h->stat.i_wpred[i]++;
2649 if( h->sh.i_type == SLICE_TYPE_B )
2651 h->stat.i_direct_frames[ h->sh.b_direct_spatial_mv_pred ] ++;
2652 if( h->mb.b_direct_auto_write )
2654 //FIXME somewhat arbitrary time constants
2655 if( h->stat.i_direct_score[0] + h->stat.i_direct_score[1] > h->mb.i_mb_count )
2656 for( int i = 0; i < 2; i++ )
2657 h->stat.i_direct_score[i] = h->stat.i_direct_score[i] * 9/10;
2658 for( int i = 0; i < 2; i++ )
2659 h->stat.i_direct_score[i] += h->stat.frame.i_direct_score[i];
2663 psz_message[0] = '\0';
2664 if( h->param.analyse.b_psnr )
2667 h->stat.frame.i_ssd[0],
2668 h->stat.frame.i_ssd[1],
2669 h->stat.frame.i_ssd[2],
2672 h->stat.i_ssd_global[h->sh.i_type] += ssd[0] + ssd[1] + ssd[2];
2673 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 );
2674 h->stat.f_psnr_mean_y[h->sh.i_type] += x264_psnr( ssd[0], h->param.i_width * h->param.i_height );
2675 h->stat.f_psnr_mean_u[h->sh.i_type] += x264_psnr( ssd[1], h->param.i_width * h->param.i_height / 4 );
2676 h->stat.f_psnr_mean_v[h->sh.i_type] += x264_psnr( ssd[2], h->param.i_width * h->param.i_height / 4 );
2678 snprintf( psz_message, 80, " PSNR Y:%5.2f U:%5.2f V:%5.2f",
2679 x264_psnr( ssd[0], h->param.i_width * h->param.i_height ),
2680 x264_psnr( ssd[1], h->param.i_width * h->param.i_height / 4),
2681 x264_psnr( ssd[2], h->param.i_width * h->param.i_height / 4) );
2684 if( h->param.analyse.b_ssim )
2686 double ssim_y = h->stat.frame.f_ssim
2687 / (((h->param.i_width-6)>>2) * ((h->param.i_height-6)>>2));
2688 h->stat.f_ssim_mean_y[h->sh.i_type] += ssim_y;
2689 snprintf( psz_message + strlen(psz_message), 80 - strlen(psz_message),
2690 " SSIM Y:%.5f", ssim_y );
2692 psz_message[79] = '\0';
2694 x264_log( h, X264_LOG_DEBUG,
2695 "frame=%4d QP=%.2f NAL=%d Slice:%c Poc:%-3d I:%-4d P:%-4d SKIP:%-4d size=%d bytes%s\n",
2697 h->fdec->f_qp_avg_aq,
2699 h->sh.i_type == SLICE_TYPE_I ? 'I' : (h->sh.i_type == SLICE_TYPE_P ? 'P' : 'B' ),
2701 h->stat.frame.i_mb_count_i,
2702 h->stat.frame.i_mb_count_p,
2703 h->stat.frame.i_mb_count_skip,
2707 // keep stats all in one place
2708 x264_thread_sync_stat( h->thread[0], h );
2709 // for the use of the next frame
2710 x264_thread_sync_stat( thread_current, h );
2712 #ifdef DEBUG_MB_TYPE
2714 static const char mb_chars[] = { 'i', 'i', 'I', 'C', 'P', '8', 'S',
2715 'D', '<', 'X', 'B', 'X', '>', 'B', 'B', 'B', 'B', '8', 'S' };
2716 for( int mb_xy = 0; mb_xy < h->sps->i_mb_width * h->sps->i_mb_height; mb_xy++ )
2718 if( h->mb.type[mb_xy] < X264_MBTYPE_MAX && h->mb.type[mb_xy] >= 0 )
2719 fprintf( stderr, "%c ", mb_chars[ h->mb.type[mb_xy] ] );
2721 fprintf( stderr, "? " );
2723 if( (mb_xy+1) % h->sps->i_mb_width == 0 )
2724 fprintf( stderr, "\n" );
2729 /* Remove duplicates, must be done near the end as breaks h->fref0 array
2730 * by freeing some of its pointers. */
2731 for( int i = 0; i < h->i_ref0; i++ )
2732 if( h->fref0[i] && h->fref0[i]->b_duplicate )
2734 x264_frame_push_blank_unused( h, h->fref0[i] );
2738 if( h->param.psz_dump_yuv )
2739 x264_frame_dump( h );
2744 static void x264_print_intra( int64_t *i_mb_count, double i_count, int b_print_pcm, char *intra )
2746 intra += sprintf( intra, "I16..4%s: %4.1f%% %4.1f%% %4.1f%%",
2747 b_print_pcm ? "..PCM" : "",
2748 i_mb_count[I_16x16]/ i_count,
2749 i_mb_count[I_8x8] / i_count,
2750 i_mb_count[I_4x4] / i_count );
2752 sprintf( intra, " %4.1f%%", i_mb_count[I_PCM] / i_count );
2755 /****************************************************************************
2756 * x264_encoder_close:
2757 ****************************************************************************/
2758 void x264_encoder_close ( x264_t *h )
2760 int64_t i_yuv_size = 3 * h->param.i_width * h->param.i_height / 2;
2761 int64_t i_mb_count_size[2][7] = {{0}};
2763 int b_print_pcm = h->stat.i_mb_count[SLICE_TYPE_I][I_PCM]
2764 || h->stat.i_mb_count[SLICE_TYPE_P][I_PCM]
2765 || h->stat.i_mb_count[SLICE_TYPE_B][I_PCM];
2767 x264_lookahead_delete( h );
2769 if( h->param.i_threads > 1 )
2771 // don't strictly have to wait for the other threads, but it's simpler than canceling them
2772 for( int i = 0; i < h->param.i_threads; i++ )
2773 if( h->thread[i]->b_thread_active )
2774 x264_pthread_join( h->thread[i]->thread_handle, NULL );
2775 if( h->i_thread_frames > 1 )
2777 for( int i = 0; i < h->i_thread_frames; i++ )
2778 if( h->thread[i]->b_thread_active )
2780 assert( h->thread[i]->fenc->i_reference_count == 1 );
2781 x264_frame_delete( h->thread[i]->fenc );
2784 x264_t *thread_prev = h->thread[h->i_thread_phase];
2785 x264_thread_sync_ratecontrol( h, thread_prev, h );
2786 x264_thread_sync_ratecontrol( thread_prev, thread_prev, h );
2787 h->i_frame = thread_prev->i_frame + 1 - h->i_thread_frames;
2792 /* Slices used and PSNR */
2793 for( int i = 0; i < 5; i++ )
2795 static const int slice_order[] = { SLICE_TYPE_I, SLICE_TYPE_SI, SLICE_TYPE_P, SLICE_TYPE_SP, SLICE_TYPE_B };
2796 static const char *slice_name[] = { "P", "B", "I", "SP", "SI" };
2797 int i_slice = slice_order[i];
2799 if( h->stat.i_frame_count[i_slice] > 0 )
2801 const int i_count = h->stat.i_frame_count[i_slice];
2802 if( h->param.analyse.b_psnr )
2804 x264_log( h, X264_LOG_INFO,
2805 "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",
2806 slice_name[i_slice],
2808 h->stat.f_frame_qp[i_slice] / i_count,
2809 (double)h->stat.i_frame_size[i_slice] / i_count,
2810 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,
2811 h->stat.f_psnr_average[i_slice] / i_count,
2812 x264_psnr( h->stat.i_ssd_global[i_slice], i_count * i_yuv_size ) );
2816 x264_log( h, X264_LOG_INFO,
2817 "frame %s:%-5d Avg QP:%5.2f size:%6.0f\n",
2818 slice_name[i_slice],
2820 h->stat.f_frame_qp[i_slice] / i_count,
2821 (double)h->stat.i_frame_size[i_slice] / i_count );
2825 if( h->param.i_bframe && h->stat.i_frame_count[SLICE_TYPE_P] )
2829 // weight by number of frames (including the P-frame) that are in a sequence of N B-frames
2830 for( int i = 0; i <= h->param.i_bframe; i++ )
2831 den += (i+1) * h->stat.i_consecutive_bframes[i];
2832 for( int i = 0; i <= h->param.i_bframe; i++ )
2833 p += sprintf( p, " %4.1f%%", 100. * (i+1) * h->stat.i_consecutive_bframes[i] / den );
2834 x264_log( h, X264_LOG_INFO, "consecutive B-frames:%s\n", buf );
2837 for( int i_type = 0; i_type < 2; i_type++ )
2838 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
2840 if( i == D_DIRECT_8x8 ) continue; /* direct is counted as its own type */
2841 i_mb_count_size[i_type][x264_mb_partition_pixel_table[i]] += h->stat.i_mb_partition[i_type][i];
2845 if( h->stat.i_frame_count[SLICE_TYPE_I] > 0 )
2847 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_I];
2848 double i_count = h->stat.i_frame_count[SLICE_TYPE_I] * h->mb.i_mb_count / 100.0;
2849 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
2850 x264_log( h, X264_LOG_INFO, "mb I %s\n", buf );
2852 if( h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
2854 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_P];
2855 double i_count = h->stat.i_frame_count[SLICE_TYPE_P] * h->mb.i_mb_count / 100.0;
2856 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_P];
2857 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
2858 x264_log( h, X264_LOG_INFO,
2859 "mb P %s P16..4: %4.1f%% %4.1f%% %4.1f%% %4.1f%% %4.1f%% skip:%4.1f%%\n",
2861 i_mb_size[PIXEL_16x16] / (i_count*4),
2862 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
2863 i_mb_size[PIXEL_8x8] / (i_count*4),
2864 (i_mb_size[PIXEL_8x4] + i_mb_size[PIXEL_4x8]) / (i_count*4),
2865 i_mb_size[PIXEL_4x4] / (i_count*4),
2866 i_mb_count[P_SKIP] / i_count );
2868 if( h->stat.i_frame_count[SLICE_TYPE_B] > 0 )
2870 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_B];
2871 double i_count = h->stat.i_frame_count[SLICE_TYPE_B] * h->mb.i_mb_count / 100.0;
2872 double i_mb_list_count;
2873 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_B];
2874 int64_t list_count[3] = {0}; /* 0 == L0, 1 == L1, 2 == BI */
2875 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
2876 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
2877 for( int j = 0; j < 2; j++ )
2879 int l0 = x264_mb_type_list_table[i][0][j];
2880 int l1 = x264_mb_type_list_table[i][1][j];
2882 list_count[l1+l0*l1] += h->stat.i_mb_count[SLICE_TYPE_B][i] * 2;
2884 list_count[0] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L0_8x8];
2885 list_count[1] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L1_8x8];
2886 list_count[2] += h->stat.i_mb_partition[SLICE_TYPE_B][D_BI_8x8];
2887 i_mb_count[B_DIRECT] += (h->stat.i_mb_partition[SLICE_TYPE_B][D_DIRECT_8x8]+2)/4;
2888 i_mb_list_count = (list_count[0] + list_count[1] + list_count[2]) / 100.0;
2889 x264_log( h, X264_LOG_INFO,
2890 "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",
2892 i_mb_size[PIXEL_16x16] / (i_count*4),
2893 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
2894 i_mb_size[PIXEL_8x8] / (i_count*4),
2895 i_mb_count[B_DIRECT] / i_count,
2896 i_mb_count[B_SKIP] / i_count,
2897 list_count[0] / i_mb_list_count,
2898 list_count[1] / i_mb_list_count,
2899 list_count[2] / i_mb_list_count );
2902 x264_ratecontrol_summary( h );
2904 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 )
2906 #define SUM3(p) (p[SLICE_TYPE_I] + p[SLICE_TYPE_P] + p[SLICE_TYPE_B])
2907 #define SUM3b(p,o) (p[SLICE_TYPE_I][o] + p[SLICE_TYPE_P][o] + p[SLICE_TYPE_B][o])
2908 int64_t i_i8x8 = SUM3b( h->stat.i_mb_count, I_8x8 );
2909 int64_t i_intra = i_i8x8 + SUM3b( h->stat.i_mb_count, I_4x4 )
2910 + SUM3b( h->stat.i_mb_count, I_16x16 );
2911 int64_t i_all_intra = i_intra + SUM3b( h->stat.i_mb_count, I_PCM);
2912 const int i_count = h->stat.i_frame_count[SLICE_TYPE_I] +
2913 h->stat.i_frame_count[SLICE_TYPE_P] +
2914 h->stat.i_frame_count[SLICE_TYPE_B];
2915 int64_t i_mb_count = i_count * h->mb.i_mb_count;
2916 float fps = (float) h->param.i_fps_num / h->param.i_fps_den;
2918 /* duration algorithm fails with one frame */
2919 if( !h->param.b_vfr_input || i_count == 1 )
2920 f_bitrate = fps * SUM3(h->stat.i_frame_size) / i_count / 125;
2923 float duration = (float)(2 * h->frames.i_largest_pts - h->frames.i_second_largest_pts) * h->param.i_timebase_num / h->param.i_timebase_den;
2924 f_bitrate = SUM3(h->stat.i_frame_size) / duration / 125;
2927 if( h->pps->b_transform_8x8_mode )
2930 if( h->stat.i_mb_count_8x8dct[0] )
2931 sprintf( buf, " inter:%.1f%%", 100. * h->stat.i_mb_count_8x8dct[1] / h->stat.i_mb_count_8x8dct[0] );
2932 x264_log( h, X264_LOG_INFO, "8x8 transform intra:%.1f%%%s\n", 100. * i_i8x8 / i_intra, buf );
2935 if( (h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO ||
2936 (h->stat.i_direct_frames[0] && h->stat.i_direct_frames[1]))
2937 && h->stat.i_frame_count[SLICE_TYPE_B] )
2939 x264_log( h, X264_LOG_INFO, "direct mvs spatial:%.1f%% temporal:%.1f%%\n",
2940 h->stat.i_direct_frames[1] * 100. / h->stat.i_frame_count[SLICE_TYPE_B],
2941 h->stat.i_direct_frames[0] * 100. / h->stat.i_frame_count[SLICE_TYPE_B] );
2945 if( i_mb_count != i_all_intra )
2946 sprintf( buf, " inter: %.1f%% %.1f%% %.1f%%",
2947 h->stat.i_mb_cbp[1] * 100.0 / ((i_mb_count - i_all_intra)*4),
2948 h->stat.i_mb_cbp[3] * 100.0 / ((i_mb_count - i_all_intra) ),
2949 h->stat.i_mb_cbp[5] * 100.0 / ((i_mb_count - i_all_intra)) );
2950 x264_log( h, X264_LOG_INFO, "coded y,uvDC,uvAC intra: %.1f%% %.1f%% %.1f%%%s\n",
2951 h->stat.i_mb_cbp[0] * 100.0 / (i_all_intra*4),
2952 h->stat.i_mb_cbp[2] * 100.0 / (i_all_intra ),
2953 h->stat.i_mb_cbp[4] * 100.0 / (i_all_intra ), buf );
2955 int64_t fixed_pred_modes[4][9] = {{0}};
2956 int64_t sum_pred_modes[4] = {0};
2957 for( int i = 0; i <= I_PRED_16x16_DC_128; i++ )
2959 fixed_pred_modes[0][x264_mb_pred_mode16x16_fix[i]] += h->stat.i_mb_pred_mode[0][i];
2960 sum_pred_modes[0] += h->stat.i_mb_pred_mode[0][i];
2962 if( sum_pred_modes[0] )
2963 x264_log( h, X264_LOG_INFO, "i16 v,h,dc,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
2964 fixed_pred_modes[0][0] * 100.0 / sum_pred_modes[0],
2965 fixed_pred_modes[0][1] * 100.0 / sum_pred_modes[0],
2966 fixed_pred_modes[0][2] * 100.0 / sum_pred_modes[0],
2967 fixed_pred_modes[0][3] * 100.0 / sum_pred_modes[0] );
2968 for( int i = 1; i <= 2; i++ )
2970 for( int j = 0; j <= I_PRED_8x8_DC_128; j++ )
2972 fixed_pred_modes[i][x264_mb_pred_mode4x4_fix(j)] += h->stat.i_mb_pred_mode[i][j];
2973 sum_pred_modes[i] += h->stat.i_mb_pred_mode[i][j];
2975 if( sum_pred_modes[i] )
2976 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,
2977 fixed_pred_modes[i][0] * 100.0 / sum_pred_modes[i],
2978 fixed_pred_modes[i][1] * 100.0 / sum_pred_modes[i],
2979 fixed_pred_modes[i][2] * 100.0 / sum_pred_modes[i],
2980 fixed_pred_modes[i][3] * 100.0 / sum_pred_modes[i],
2981 fixed_pred_modes[i][4] * 100.0 / sum_pred_modes[i],
2982 fixed_pred_modes[i][5] * 100.0 / sum_pred_modes[i],
2983 fixed_pred_modes[i][6] * 100.0 / sum_pred_modes[i],
2984 fixed_pred_modes[i][7] * 100.0 / sum_pred_modes[i],
2985 fixed_pred_modes[i][8] * 100.0 / sum_pred_modes[i] );
2987 for( int i = 0; i <= I_PRED_CHROMA_DC_128; i++ )
2989 fixed_pred_modes[3][x264_mb_pred_mode8x8c_fix[i]] += h->stat.i_mb_pred_mode[3][i];
2990 sum_pred_modes[3] += h->stat.i_mb_pred_mode[3][i];
2992 if( sum_pred_modes[3] )
2993 x264_log( h, X264_LOG_INFO, "i8c dc,h,v,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
2994 fixed_pred_modes[3][0] * 100.0 / sum_pred_modes[3],
2995 fixed_pred_modes[3][1] * 100.0 / sum_pred_modes[3],
2996 fixed_pred_modes[3][2] * 100.0 / sum_pred_modes[3],
2997 fixed_pred_modes[3][3] * 100.0 / sum_pred_modes[3] );
2999 if( h->param.analyse.i_weighted_pred == X264_WEIGHTP_SMART && h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
3000 x264_log( h, X264_LOG_INFO, "Weighted P-Frames: Y:%.1f%%\n",
3001 h->stat.i_wpred[0] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P] );
3003 for( int i_list = 0; i_list < 2; i_list++ )
3004 for( int i_slice = 0; i_slice < 2; i_slice++ )
3009 for( int i = 0; i < 32; i++ )
3010 if( h->stat.i_mb_count_ref[i_slice][i_list][i] )
3012 i_den += h->stat.i_mb_count_ref[i_slice][i_list][i];
3017 for( int i = 0; i <= i_max; i++ )
3018 p += sprintf( p, " %4.1f%%", 100. * h->stat.i_mb_count_ref[i_slice][i_list][i] / i_den );
3019 x264_log( h, X264_LOG_INFO, "ref %c L%d:%s\n", "PB"[i_slice], i_list, buf );
3022 if( h->param.analyse.b_ssim )
3024 x264_log( h, X264_LOG_INFO,
3025 "SSIM Mean Y:%.7f\n",
3026 SUM3( h->stat.f_ssim_mean_y ) / i_count );
3028 if( h->param.analyse.b_psnr )
3030 x264_log( h, X264_LOG_INFO,
3031 "PSNR Mean Y:%6.3f U:%6.3f V:%6.3f Avg:%6.3f Global:%6.3f kb/s:%.2f\n",
3032 SUM3( h->stat.f_psnr_mean_y ) / i_count,
3033 SUM3( h->stat.f_psnr_mean_u ) / i_count,
3034 SUM3( h->stat.f_psnr_mean_v ) / i_count,
3035 SUM3( h->stat.f_psnr_average ) / i_count,
3036 x264_psnr( SUM3( h->stat.i_ssd_global ), i_count * i_yuv_size ),
3040 x264_log( h, X264_LOG_INFO, "kb/s:%.2f\n", f_bitrate );
3044 x264_ratecontrol_delete( h );
3047 if( h->param.rc.psz_stat_out )
3048 free( h->param.rc.psz_stat_out );
3049 if( h->param.rc.psz_stat_in )
3050 free( h->param.rc.psz_stat_in );
3052 x264_cqm_delete( h );
3053 x264_free( h->nal_buffer );
3054 x264_analyse_free_costs( h );
3056 if( h->i_thread_frames > 1)
3057 h = h->thread[h->i_thread_phase];
3060 x264_frame_delete_list( h->frames.unused[0] );
3061 x264_frame_delete_list( h->frames.unused[1] );
3062 x264_frame_delete_list( h->frames.current );
3063 x264_frame_delete_list( h->frames.blank_unused );
3067 for( int i = h->param.i_threads - 1; i >= 0; i-- )
3069 x264_frame_t **frame;
3071 if( !h->param.b_sliced_threads || i == 0 )
3073 for( frame = h->thread[i]->frames.reference; *frame; frame++ )
3075 assert( (*frame)->i_reference_count > 0 );
3076 (*frame)->i_reference_count--;
3077 if( (*frame)->i_reference_count == 0 )
3078 x264_frame_delete( *frame );
3080 frame = &h->thread[i]->fdec;
3081 assert( (*frame)->i_reference_count > 0 );
3082 (*frame)->i_reference_count--;
3083 if( (*frame)->i_reference_count == 0 )
3084 x264_frame_delete( *frame );
3085 x264_macroblock_cache_free( h->thread[i] );
3087 x264_macroblock_thread_free( h->thread[i], 0 );
3088 x264_free( h->thread[i]->out.p_bitstream );
3089 x264_free( h->thread[i]->out.nal);
3090 x264_free( h->thread[i] );
3094 /****************************************************************************
3095 * x264_encoder_delayed_frames:
3096 ****************************************************************************/
3097 int x264_encoder_delayed_frames( x264_t *h )
3099 int delayed_frames = 0;
3100 if( h->i_thread_frames > 1 )
3102 for( int i = 0; i < h->i_thread_frames; i++ )
3103 delayed_frames += h->thread[i]->b_thread_active;
3104 h = h->thread[h->i_thread_phase];
3106 for( int i = 0; h->frames.current[i]; i++ )
3108 x264_pthread_mutex_lock( &h->lookahead->ofbuf.mutex );
3109 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
3110 x264_pthread_mutex_lock( &h->lookahead->next.mutex );
3111 delayed_frames += h->lookahead->ifbuf.i_size + h->lookahead->next.i_size + h->lookahead->ofbuf.i_size;
3112 x264_pthread_mutex_unlock( &h->lookahead->next.mutex );
3113 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
3114 x264_pthread_mutex_unlock( &h->lookahead->ofbuf.mutex );
3115 return delayed_frames;