1 /*****************************************************************************
2 * encoder.c: top-level encoder functions
3 *****************************************************************************
4 * Copyright (C) 2003-2013 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.
24 * This program is also available under a commercial proprietary license.
25 * For more information, contact us at licensing@x264.com.
26 *****************************************************************************/
28 #include "common/common.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 double x264_psnr( double sqe, double size )
55 double mse = sqe / (PIXEL_MAX*PIXEL_MAX * size);
56 if( mse <= 0.0000000001 ) /* Max 100dB */
59 return -10.0 * log10( mse );
62 static double x264_ssim( double ssim )
64 double inv_ssim = 1 - ssim;
65 if( inv_ssim <= 0.0000000001 ) /* Max 100dB */
68 return -10.0 * log10( inv_ssim );
71 static int x264_threadpool_wait_all( x264_t *h )
73 for( int i = 0; i < h->param.i_threads; i++ )
74 if( h->thread[i]->b_thread_active )
76 h->thread[i]->b_thread_active = 0;
77 if( (intptr_t)x264_threadpool_wait( h->threadpool, h->thread[i] ) < 0 )
83 static void x264_frame_dump( x264_t *h )
85 FILE *f = fopen( h->param.psz_dump_yuv, "r+b" );
89 /* Wait for the threads to finish deblocking */
90 if( h->param.b_sliced_threads )
91 x264_threadpool_wait_all( h );
93 /* Write the frame in display order */
94 int frame_size = FRAME_SIZE( h->param.i_height * h->param.i_width * sizeof(pixel) );
95 fseek( f, (uint64_t)h->fdec->i_frame * frame_size, SEEK_SET );
96 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
97 for( int y = 0; y < h->param.i_height; y++ )
98 fwrite( &h->fdec->plane[p][y*h->fdec->i_stride[p]], sizeof(pixel), h->param.i_width, f );
101 int cw = h->param.i_width>>1;
102 int ch = h->param.i_height>>CHROMA_V_SHIFT;
103 pixel *planeu = x264_malloc( (cw*ch*2+32)*sizeof(pixel) );
104 pixel *planev = planeu + cw*ch + 16;
105 h->mc.plane_copy_deinterleave( planeu, cw, planev, cw, h->fdec->plane[1], h->fdec->i_stride[1], cw, ch );
106 fwrite( planeu, 1, cw*ch*sizeof(pixel), f );
107 fwrite( planev, 1, cw*ch*sizeof(pixel), f );
113 /* Fill "default" values */
114 static void x264_slice_header_init( x264_t *h, x264_slice_header_t *sh,
115 x264_sps_t *sps, x264_pps_t *pps,
116 int i_idr_pic_id, int i_frame, int i_qp )
118 x264_param_t *param = &h->param;
120 /* First we fill all fields */
125 sh->i_last_mb = h->mb.i_mb_count - 1;
126 sh->i_pps_id = pps->i_id;
128 sh->i_frame_num = i_frame;
130 sh->b_mbaff = PARAM_INTERLACED;
131 sh->b_field_pic = 0; /* no field support for now */
132 sh->b_bottom_field = 0; /* not yet used */
134 sh->i_idr_pic_id = i_idr_pic_id;
136 /* poc stuff, fixed later */
138 sh->i_delta_poc_bottom = 0;
139 sh->i_delta_poc[0] = 0;
140 sh->i_delta_poc[1] = 0;
142 sh->i_redundant_pic_cnt = 0;
144 h->mb.b_direct_auto_write = h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO
146 && ( h->param.rc.b_stat_write || !h->param.rc.b_stat_read );
148 if( !h->mb.b_direct_auto_read && sh->i_type == SLICE_TYPE_B )
150 if( h->fref[1][0]->i_poc_l0ref0 == h->fref[0][0]->i_poc )
152 if( h->mb.b_direct_auto_write )
153 sh->b_direct_spatial_mv_pred = ( h->stat.i_direct_score[1] > h->stat.i_direct_score[0] );
155 sh->b_direct_spatial_mv_pred = ( param->analyse.i_direct_mv_pred == X264_DIRECT_PRED_SPATIAL );
159 h->mb.b_direct_auto_write = 0;
160 sh->b_direct_spatial_mv_pred = 1;
163 /* else b_direct_spatial_mv_pred was read from the 2pass statsfile */
165 sh->b_num_ref_idx_override = 0;
166 sh->i_num_ref_idx_l0_active = 1;
167 sh->i_num_ref_idx_l1_active = 1;
169 sh->b_ref_pic_list_reordering[0] = h->b_ref_reorder[0];
170 sh->b_ref_pic_list_reordering[1] = h->b_ref_reorder[1];
172 /* If the ref list isn't in the default order, construct reordering header */
173 for( int list = 0; list < 2; list++ )
175 if( sh->b_ref_pic_list_reordering[list] )
177 int pred_frame_num = i_frame;
178 for( int i = 0; i < h->i_ref[list]; i++ )
180 int diff = h->fref[list][i]->i_frame_num - pred_frame_num;
181 sh->ref_pic_list_order[list][i].idc = ( diff > 0 );
182 sh->ref_pic_list_order[list][i].arg = (abs(diff) - 1) & ((1 << sps->i_log2_max_frame_num) - 1);
183 pred_frame_num = h->fref[list][i]->i_frame_num;
188 sh->i_cabac_init_idc = param->i_cabac_init_idc;
190 sh->i_qp = SPEC_QP(i_qp);
191 sh->i_qp_delta = sh->i_qp - pps->i_pic_init_qp;
192 sh->b_sp_for_swidth = 0;
195 int deblock_thresh = i_qp + 2 * X264_MIN(param->i_deblocking_filter_alphac0, param->i_deblocking_filter_beta);
196 /* If effective qp <= 15, deblocking would have no effect anyway */
197 if( param->b_deblocking_filter && (h->mb.b_variable_qp || 15 < deblock_thresh ) )
198 sh->i_disable_deblocking_filter_idc = param->b_sliced_threads ? 2 : 0;
200 sh->i_disable_deblocking_filter_idc = 1;
201 sh->i_alpha_c0_offset = param->i_deblocking_filter_alphac0 << 1;
202 sh->i_beta_offset = param->i_deblocking_filter_beta << 1;
205 static void x264_slice_header_write( bs_t *s, x264_slice_header_t *sh, int i_nal_ref_idc )
209 int first_x = sh->i_first_mb % sh->sps->i_mb_width;
210 int first_y = sh->i_first_mb / sh->sps->i_mb_width;
211 assert( (first_y&1) == 0 );
212 bs_write_ue( s, (2*first_x + sh->sps->i_mb_width*(first_y&~1) + (first_y&1)) >> 1 );
215 bs_write_ue( s, sh->i_first_mb );
217 bs_write_ue( s, sh->i_type + 5 ); /* same type things */
218 bs_write_ue( s, sh->i_pps_id );
219 bs_write( s, sh->sps->i_log2_max_frame_num, sh->i_frame_num & ((1<<sh->sps->i_log2_max_frame_num)-1) );
221 if( !sh->sps->b_frame_mbs_only )
223 bs_write1( s, sh->b_field_pic );
224 if( sh->b_field_pic )
225 bs_write1( s, sh->b_bottom_field );
228 if( sh->i_idr_pic_id >= 0 ) /* NAL IDR */
229 bs_write_ue( s, sh->i_idr_pic_id );
231 if( sh->sps->i_poc_type == 0 )
233 bs_write( s, sh->sps->i_log2_max_poc_lsb, sh->i_poc & ((1<<sh->sps->i_log2_max_poc_lsb)-1) );
234 if( sh->pps->b_pic_order && !sh->b_field_pic )
235 bs_write_se( s, sh->i_delta_poc_bottom );
238 if( sh->pps->b_redundant_pic_cnt )
239 bs_write_ue( s, sh->i_redundant_pic_cnt );
241 if( sh->i_type == SLICE_TYPE_B )
242 bs_write1( s, sh->b_direct_spatial_mv_pred );
244 if( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_B )
246 bs_write1( s, sh->b_num_ref_idx_override );
247 if( sh->b_num_ref_idx_override )
249 bs_write_ue( s, sh->i_num_ref_idx_l0_active - 1 );
250 if( sh->i_type == SLICE_TYPE_B )
251 bs_write_ue( s, sh->i_num_ref_idx_l1_active - 1 );
255 /* ref pic list reordering */
256 if( sh->i_type != SLICE_TYPE_I )
258 bs_write1( s, sh->b_ref_pic_list_reordering[0] );
259 if( sh->b_ref_pic_list_reordering[0] )
261 for( int i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
263 bs_write_ue( s, sh->ref_pic_list_order[0][i].idc );
264 bs_write_ue( s, sh->ref_pic_list_order[0][i].arg );
269 if( sh->i_type == SLICE_TYPE_B )
271 bs_write1( s, sh->b_ref_pic_list_reordering[1] );
272 if( sh->b_ref_pic_list_reordering[1] )
274 for( int i = 0; i < sh->i_num_ref_idx_l1_active; i++ )
276 bs_write_ue( s, sh->ref_pic_list_order[1][i].idc );
277 bs_write_ue( s, sh->ref_pic_list_order[1][i].arg );
283 sh->b_weighted_pred = 0;
284 if( sh->pps->b_weighted_pred && sh->i_type == SLICE_TYPE_P )
286 sh->b_weighted_pred = sh->weight[0][0].weightfn || sh->weight[0][1].weightfn || sh->weight[0][2].weightfn;
287 /* pred_weight_table() */
288 bs_write_ue( s, sh->weight[0][0].i_denom );
289 bs_write_ue( s, sh->weight[0][1].i_denom );
290 for( int i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
292 int luma_weight_l0_flag = !!sh->weight[i][0].weightfn;
293 int chroma_weight_l0_flag = !!sh->weight[i][1].weightfn || !!sh->weight[i][2].weightfn;
294 bs_write1( s, luma_weight_l0_flag );
295 if( luma_weight_l0_flag )
297 bs_write_se( s, sh->weight[i][0].i_scale );
298 bs_write_se( s, sh->weight[i][0].i_offset );
300 bs_write1( s, chroma_weight_l0_flag );
301 if( chroma_weight_l0_flag )
303 for( int j = 1; j < 3; j++ )
305 bs_write_se( s, sh->weight[i][j].i_scale );
306 bs_write_se( s, sh->weight[i][j].i_offset );
311 else if( sh->pps->b_weighted_bipred == 1 && sh->i_type == SLICE_TYPE_B )
316 if( i_nal_ref_idc != 0 )
318 if( sh->i_idr_pic_id >= 0 )
320 bs_write1( s, 0 ); /* no output of prior pics flag */
321 bs_write1( s, 0 ); /* long term reference flag */
325 bs_write1( s, sh->i_mmco_command_count > 0 ); /* adaptive_ref_pic_marking_mode_flag */
326 if( sh->i_mmco_command_count > 0 )
328 for( int i = 0; i < sh->i_mmco_command_count; i++ )
330 bs_write_ue( s, 1 ); /* mark short term ref as unused */
331 bs_write_ue( s, sh->mmco[i].i_difference_of_pic_nums - 1 );
333 bs_write_ue( s, 0 ); /* end command list */
338 if( sh->pps->b_cabac && sh->i_type != SLICE_TYPE_I )
339 bs_write_ue( s, sh->i_cabac_init_idc );
341 bs_write_se( s, sh->i_qp_delta ); /* slice qp delta */
343 if( sh->pps->b_deblocking_filter_control )
345 bs_write_ue( s, sh->i_disable_deblocking_filter_idc );
346 if( sh->i_disable_deblocking_filter_idc != 1 )
348 bs_write_se( s, sh->i_alpha_c0_offset >> 1 );
349 bs_write_se( s, sh->i_beta_offset >> 1 );
354 /* If we are within a reasonable distance of the end of the memory allocated for the bitstream, */
355 /* reallocate, adding an arbitrary amount of space. */
356 static int x264_bitstream_check_buffer_internal( x264_t *h, int size, int b_cabac, int i_nal )
358 if( (b_cabac && (h->cabac.p_end - h->cabac.p < size)) ||
359 (h->out.bs.p_end - h->out.bs.p < size) )
361 int buf_size = h->out.i_bitstream + size;
362 uint8_t *buf = x264_malloc( buf_size );
365 int aligned_size = h->out.i_bitstream & ~15;
366 h->mc.memcpy_aligned( buf, h->out.p_bitstream, aligned_size );
367 memcpy( buf + aligned_size, h->out.p_bitstream + aligned_size, h->out.i_bitstream - aligned_size );
369 intptr_t delta = buf - h->out.p_bitstream;
371 h->out.bs.p_start += delta;
372 h->out.bs.p += delta;
373 h->out.bs.p_end = buf + buf_size;
375 h->cabac.p_start += delta;
377 h->cabac.p_end = buf + buf_size;
379 for( int i = 0; i <= i_nal; i++ )
380 h->out.nal[i].p_payload += delta;
382 x264_free( h->out.p_bitstream );
383 h->out.p_bitstream = buf;
384 h->out.i_bitstream = buf_size;
389 static int x264_bitstream_check_buffer( x264_t *h )
391 int max_row_size = (2500 << SLICE_MBAFF) * h->mb.i_mb_width;
392 return x264_bitstream_check_buffer_internal( h, max_row_size, h->param.b_cabac, h->out.i_nal );
395 static int x264_bitstream_check_buffer_filler( x264_t *h, int filler )
397 filler += 32; // add padding for safety
398 return x264_bitstream_check_buffer_internal( h, filler, 0, -1 );
402 static void x264_encoder_thread_init( x264_t *h )
404 if( h->param.i_sync_lookahead )
405 x264_lower_thread_priority( 10 );
409 /****************************************************************************
411 ****************************************************************************
412 ****************************** External API*********************************
413 ****************************************************************************
415 ****************************************************************************/
417 static int x264_validate_parameters( x264_t *h, int b_open )
422 int cpuflags = x264_cpu_detect();
425 if( !(cpuflags & X264_CPU_SSE) )
427 x264_log( h, X264_LOG_ERROR, "your cpu does not support SSE1, but x264 was compiled with asm\n");
431 if( !(cpuflags & X264_CPU_MMX2) )
433 x264_log( h, X264_LOG_ERROR, "your cpu does not support MMXEXT, but x264 was compiled with asm\n");
437 if( !fail && !(cpuflags & X264_CPU_CMOV) )
439 x264_log( h, X264_LOG_ERROR, "your cpu does not support CMOV, but x264 was compiled with asm\n");
444 x264_log( h, X264_LOG_ERROR, "to run x264, recompile without asm (configure --disable-asm)\n");
451 h->param.b_interlaced = !!PARAM_INTERLACED;
453 if( h->param.b_interlaced )
455 x264_log( h, X264_LOG_ERROR, "not compiled with interlaced support\n" );
460 if( h->param.i_width <= 0 || h->param.i_height <= 0 )
462 x264_log( h, X264_LOG_ERROR, "invalid width x height (%dx%d)\n",
463 h->param.i_width, h->param.i_height );
467 int i_csp = h->param.i_csp & X264_CSP_MASK;
468 #if X264_CHROMA_FORMAT
469 if( CHROMA_FORMAT != CHROMA_420 && i_csp >= X264_CSP_I420 && i_csp <= X264_CSP_NV12 )
471 x264_log( h, X264_LOG_ERROR, "not compiled with 4:2:0 support\n" );
474 else if( CHROMA_FORMAT != CHROMA_422 && i_csp >= X264_CSP_I422 && i_csp <= X264_CSP_NV16 )
476 x264_log( h, X264_LOG_ERROR, "not compiled with 4:2:2 support\n" );
479 else if( CHROMA_FORMAT != CHROMA_444 && i_csp >= X264_CSP_I444 && i_csp <= X264_CSP_RGB )
481 x264_log( h, X264_LOG_ERROR, "not compiled with 4:4:4 support\n" );
485 if( i_csp <= X264_CSP_NONE || i_csp >= X264_CSP_MAX )
487 x264_log( h, X264_LOG_ERROR, "invalid CSP (only I420/YV12/NV12/I422/YV16/NV16/I444/YV24/BGR/BGRA/RGB supported)\n" );
491 if( i_csp < X264_CSP_I444 && h->param.i_width % 2 )
493 x264_log( h, X264_LOG_ERROR, "width not divisible by 2 (%dx%d)\n",
494 h->param.i_width, h->param.i_height );
498 if( i_csp < X264_CSP_I422 && PARAM_INTERLACED && h->param.i_height % 4 )
500 x264_log( h, X264_LOG_ERROR, "height not divisible by 4 (%dx%d)\n",
501 h->param.i_width, h->param.i_height );
505 if( (i_csp < X264_CSP_I422 || PARAM_INTERLACED) && h->param.i_height % 2 )
507 x264_log( h, X264_LOG_ERROR, "height not divisible by 2 (%dx%d)\n",
508 h->param.i_width, h->param.i_height );
512 if( (h->param.crop_rect.i_left + h->param.crop_rect.i_right ) >= h->param.i_width ||
513 (h->param.crop_rect.i_top + h->param.crop_rect.i_bottom) >= h->param.i_height )
515 x264_log( h, X264_LOG_ERROR, "invalid crop-rect %u,%u,%u,%u\n", h->param.crop_rect.i_left,
516 h->param.crop_rect.i_top, h->param.crop_rect.i_right, h->param.crop_rect.i_bottom );
520 if( h->param.i_threads == X264_THREADS_AUTO )
521 h->param.i_threads = x264_cpu_num_processors() * (h->param.b_sliced_threads?2:3)/2;
522 int max_sliced_threads = X264_MAX( 1, (h->param.i_height+15)/16 / 4 );
523 if( h->param.i_threads > 1 )
526 x264_log( h, X264_LOG_WARNING, "not compiled with thread support!\n");
527 h->param.i_threads = 1;
529 /* Avoid absurdly small thread slices as they can reduce performance
530 * and VBV compliance. Capped at an arbitrary 4 rows per thread. */
531 if( h->param.b_sliced_threads )
532 h->param.i_threads = X264_MIN( h->param.i_threads, max_sliced_threads );
534 h->param.i_threads = x264_clip3( h->param.i_threads, 1, X264_THREAD_MAX );
535 if( h->param.i_threads == 1 )
537 h->param.b_sliced_threads = 0;
538 h->param.i_lookahead_threads = 1;
540 h->i_thread_frames = h->param.b_sliced_threads ? 1 : h->param.i_threads;
541 if( h->i_thread_frames > 1 )
542 h->param.nalu_process = NULL;
544 if( h->param.b_opencl )
547 x264_log( h, X264_LOG_WARNING, "OpenCL: not compiled with OpenCL support, disabling\n" );
548 h->param.b_opencl = 0;
550 x264_log( h, X264_LOG_WARNING, "OpenCL lookahead does not support high bit depth, disabling opencl\n" );
551 h->param.b_opencl = 0;
553 if( h->param.i_width < 32 || h->param.i_height < 32 )
555 x264_log( h, X264_LOG_WARNING, "OpenCL: frame size is too small, disabling opencl\n" );
556 h->param.b_opencl = 0;
559 if( h->param.opencl_device_id && h->param.i_opencl_device )
561 x264_log( h, X264_LOG_WARNING, "OpenCL: device id and device skip count configured; dropping skip\n" );
562 h->param.i_opencl_device = 0;
566 h->param.i_keyint_max = x264_clip3( h->param.i_keyint_max, 1, X264_KEYINT_MAX_INFINITE );
567 if( h->param.i_keyint_max == 1 )
569 h->param.b_intra_refresh = 0;
570 h->param.analyse.i_weighted_pred = 0;
573 h->param.i_frame_packing = x264_clip3( h->param.i_frame_packing, -1, 5 );
575 /* Detect default ffmpeg settings and terminate with an error. */
579 score += h->param.analyse.i_me_range == 0;
580 score += h->param.rc.i_qp_step == 3;
581 score += h->param.i_keyint_max == 12;
582 score += h->param.rc.i_qp_min == 2;
583 score += h->param.rc.i_qp_max == 31;
584 score += h->param.rc.f_qcompress == 0.5;
585 score += fabs(h->param.rc.f_ip_factor - 1.25) < 0.01;
586 score += fabs(h->param.rc.f_pb_factor - 1.25) < 0.01;
587 score += h->param.analyse.inter == 0 && h->param.analyse.i_subpel_refine == 8;
590 x264_log( h, X264_LOG_ERROR, "broken ffmpeg default settings detected\n" );
591 x264_log( h, X264_LOG_ERROR, "use an encoding preset (e.g. -vpre medium)\n" );
592 x264_log( h, X264_LOG_ERROR, "preset usage: -vpre <speed> -vpre <profile>\n" );
593 x264_log( h, X264_LOG_ERROR, "speed presets are listed in x264 --help\n" );
594 x264_log( h, X264_LOG_ERROR, "profile is optional; x264 defaults to high\n" );
599 if( h->param.rc.i_rc_method < 0 || h->param.rc.i_rc_method > 2 )
601 x264_log( h, X264_LOG_ERROR, "no ratecontrol method specified\n" );
604 h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, -QP_BD_OFFSET, 51 );
605 h->param.rc.f_rf_constant_max = x264_clip3f( h->param.rc.f_rf_constant_max, -QP_BD_OFFSET, 51 );
606 h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, 0, QP_MAX );
607 h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 11 );
608 h->param.rc.f_ip_factor = X264_MAX( h->param.rc.f_ip_factor, 0.01f );
609 h->param.rc.f_pb_factor = X264_MAX( h->param.rc.f_pb_factor, 0.01f );
610 if( h->param.rc.i_rc_method == X264_RC_CRF )
612 h->param.rc.i_qp_constant = h->param.rc.f_rf_constant + QP_BD_OFFSET;
613 h->param.rc.i_bitrate = 0;
615 if( b_open && (h->param.rc.i_rc_method == X264_RC_CQP || h->param.rc.i_rc_method == X264_RC_CRF)
616 && h->param.rc.i_qp_constant == 0 )
618 h->mb.b_lossless = 1;
619 h->param.i_cqm_preset = X264_CQM_FLAT;
620 h->param.psz_cqm_file = NULL;
621 h->param.rc.i_rc_method = X264_RC_CQP;
622 h->param.rc.f_ip_factor = 1;
623 h->param.rc.f_pb_factor = 1;
624 h->param.analyse.b_psnr = 0;
625 h->param.analyse.b_ssim = 0;
626 h->param.analyse.i_chroma_qp_offset = 0;
627 h->param.analyse.i_trellis = 0;
628 h->param.analyse.b_fast_pskip = 0;
629 h->param.analyse.i_noise_reduction = 0;
630 h->param.analyse.b_psy = 0;
631 h->param.i_bframe = 0;
632 /* 8x8dct is not useful without RD in CAVLC lossless */
633 if( !h->param.b_cabac && h->param.analyse.i_subpel_refine < 6 )
634 h->param.analyse.b_transform_8x8 = 0;
636 if( h->param.rc.i_rc_method == X264_RC_CQP )
638 float qp_p = h->param.rc.i_qp_constant;
639 float qp_i = qp_p - 6*log2f( h->param.rc.f_ip_factor );
640 float qp_b = qp_p + 6*log2f( h->param.rc.f_pb_factor );
641 h->param.rc.i_qp_min = x264_clip3( (int)(X264_MIN3( qp_p, qp_i, qp_b )), 0, QP_MAX );
642 h->param.rc.i_qp_max = x264_clip3( (int)(X264_MAX3( qp_p, qp_i, qp_b ) + .999), 0, QP_MAX );
643 h->param.rc.i_aq_mode = 0;
644 h->param.rc.b_mb_tree = 0;
645 h->param.rc.i_bitrate = 0;
647 h->param.rc.i_qp_max = x264_clip3( h->param.rc.i_qp_max, 0, QP_MAX );
648 h->param.rc.i_qp_min = x264_clip3( h->param.rc.i_qp_min, 0, h->param.rc.i_qp_max );
649 h->param.rc.i_qp_step = x264_clip3( h->param.rc.i_qp_step, 2, QP_MAX );
650 h->param.rc.i_bitrate = x264_clip3( h->param.rc.i_bitrate, 0, 2000000 );
651 if( h->param.rc.i_rc_method == X264_RC_ABR && !h->param.rc.i_bitrate )
653 x264_log( h, X264_LOG_ERROR, "bitrate not specified\n" );
656 h->param.rc.i_vbv_buffer_size = x264_clip3( h->param.rc.i_vbv_buffer_size, 0, 2000000 );
657 h->param.rc.i_vbv_max_bitrate = x264_clip3( h->param.rc.i_vbv_max_bitrate, 0, 2000000 );
658 h->param.rc.f_vbv_buffer_init = x264_clip3f( h->param.rc.f_vbv_buffer_init, 0, 2000000 );
659 if( h->param.rc.i_vbv_buffer_size )
661 if( h->param.rc.i_rc_method == X264_RC_CQP )
663 x264_log( h, X264_LOG_WARNING, "VBV is incompatible with constant QP, ignored.\n" );
664 h->param.rc.i_vbv_max_bitrate = 0;
665 h->param.rc.i_vbv_buffer_size = 0;
667 else if( h->param.rc.i_vbv_max_bitrate == 0 )
669 if( h->param.rc.i_rc_method == X264_RC_ABR )
671 x264_log( h, X264_LOG_WARNING, "VBV maxrate unspecified, assuming CBR\n" );
672 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
676 x264_log( h, X264_LOG_WARNING, "VBV bufsize set but maxrate unspecified, ignored\n" );
677 h->param.rc.i_vbv_buffer_size = 0;
680 else if( h->param.rc.i_vbv_max_bitrate < h->param.rc.i_bitrate &&
681 h->param.rc.i_rc_method == X264_RC_ABR )
683 x264_log( h, X264_LOG_WARNING, "max bitrate less than average bitrate, assuming CBR\n" );
684 h->param.rc.i_bitrate = h->param.rc.i_vbv_max_bitrate;
687 else if( h->param.rc.i_vbv_max_bitrate )
689 x264_log( h, X264_LOG_WARNING, "VBV maxrate specified, but no bufsize, ignored\n" );
690 h->param.rc.i_vbv_max_bitrate = 0;
693 h->param.i_slice_max_size = X264_MAX( h->param.i_slice_max_size, 0 );
694 h->param.i_slice_max_mbs = X264_MAX( h->param.i_slice_max_mbs, 0 );
695 h->param.i_slice_min_mbs = X264_MAX( h->param.i_slice_min_mbs, 0 );
696 if( h->param.i_slice_max_mbs )
697 h->param.i_slice_min_mbs = X264_MIN( h->param.i_slice_min_mbs, h->param.i_slice_max_mbs/2 );
698 else if( !h->param.i_slice_max_size )
699 h->param.i_slice_min_mbs = 0;
700 if( PARAM_INTERLACED && h->param.i_slice_min_mbs )
702 x264_log( h, X264_LOG_WARNING, "interlace + slice-min-mbs is not implemented\n" );
703 h->param.i_slice_min_mbs = 0;
705 int mb_width = (h->param.i_width+15)/16;
706 if( h->param.i_slice_min_mbs > mb_width )
708 x264_log( h, X264_LOG_WARNING, "slice-min-mbs > row mb size (%d) not implemented\n", mb_width );
709 h->param.i_slice_min_mbs = mb_width;
712 int max_slices = (h->param.i_height+((16<<PARAM_INTERLACED)-1))/(16<<PARAM_INTERLACED);
713 if( h->param.b_sliced_threads )
714 h->param.i_slice_count = x264_clip3( h->param.i_threads, 0, max_slices );
717 h->param.i_slice_count = x264_clip3( h->param.i_slice_count, 0, max_slices );
718 if( h->param.i_slice_max_mbs || h->param.i_slice_max_size )
719 h->param.i_slice_count = 0;
721 if( h->param.i_slice_count_max > 0 )
722 h->param.i_slice_count_max = X264_MAX( h->param.i_slice_count, h->param.i_slice_count_max );
724 if( h->param.b_bluray_compat )
726 h->param.i_bframe_pyramid = X264_MIN( X264_B_PYRAMID_STRICT, h->param.i_bframe_pyramid );
727 h->param.i_bframe = X264_MIN( h->param.i_bframe, 3 );
729 h->param.i_nal_hrd = X264_MAX( h->param.i_nal_hrd, X264_NAL_HRD_VBR );
730 h->param.i_slice_max_size = 0;
731 h->param.i_slice_max_mbs = 0;
732 h->param.b_intra_refresh = 0;
733 h->param.i_frame_reference = X264_MIN( h->param.i_frame_reference, 6 );
734 h->param.i_dpb_size = X264_MIN( h->param.i_dpb_size, 6 );
735 /* Don't use I-frames, because Blu-ray treats them the same as IDR. */
736 h->param.i_keyint_min = 1;
737 /* Due to the proliferation of broken players that don't handle dupes properly. */
738 h->param.analyse.i_weighted_pred = X264_MIN( h->param.analyse.i_weighted_pred, X264_WEIGHTP_SIMPLE );
739 if( h->param.b_fake_interlaced )
740 h->param.b_pic_struct = 1;
743 h->param.i_frame_reference = x264_clip3( h->param.i_frame_reference, 1, X264_REF_MAX );
744 h->param.i_dpb_size = x264_clip3( h->param.i_dpb_size, 1, X264_REF_MAX );
745 if( h->param.i_scenecut_threshold < 0 )
746 h->param.i_scenecut_threshold = 0;
747 h->param.analyse.i_direct_mv_pred = x264_clip3( h->param.analyse.i_direct_mv_pred, X264_DIRECT_PRED_NONE, X264_DIRECT_PRED_AUTO );
748 if( !h->param.analyse.i_subpel_refine && h->param.analyse.i_direct_mv_pred > X264_DIRECT_PRED_SPATIAL )
750 x264_log( h, X264_LOG_WARNING, "subme=0 + direct=temporal is not supported\n" );
751 h->param.analyse.i_direct_mv_pred = X264_DIRECT_PRED_SPATIAL;
753 h->param.i_bframe = x264_clip3( h->param.i_bframe, 0, X264_MIN( X264_BFRAME_MAX, h->param.i_keyint_max-1 ) );
754 h->param.i_bframe_bias = x264_clip3( h->param.i_bframe_bias, -90, 100 );
755 if( h->param.i_bframe <= 1 )
756 h->param.i_bframe_pyramid = X264_B_PYRAMID_NONE;
757 h->param.i_bframe_pyramid = x264_clip3( h->param.i_bframe_pyramid, X264_B_PYRAMID_NONE, X264_B_PYRAMID_NORMAL );
758 h->param.i_bframe_adaptive = x264_clip3( h->param.i_bframe_adaptive, X264_B_ADAPT_NONE, X264_B_ADAPT_TRELLIS );
759 if( !h->param.i_bframe )
761 h->param.i_bframe_adaptive = X264_B_ADAPT_NONE;
762 h->param.analyse.i_direct_mv_pred = 0;
763 h->param.analyse.b_weighted_bipred = 0;
764 h->param.b_open_gop = 0;
766 if( h->param.b_intra_refresh && h->param.i_bframe_pyramid == X264_B_PYRAMID_NORMAL )
768 x264_log( h, X264_LOG_WARNING, "b-pyramid normal + intra-refresh is not supported\n" );
769 h->param.i_bframe_pyramid = X264_B_PYRAMID_STRICT;
771 if( h->param.b_intra_refresh && (h->param.i_frame_reference > 1 || h->param.i_dpb_size > 1) )
773 x264_log( h, X264_LOG_WARNING, "ref > 1 + intra-refresh is not supported\n" );
774 h->param.i_frame_reference = 1;
775 h->param.i_dpb_size = 1;
777 if( h->param.b_intra_refresh && h->param.b_open_gop )
779 x264_log( h, X264_LOG_WARNING, "intra-refresh is not compatible with open-gop\n" );
780 h->param.b_open_gop = 0;
782 if( !h->param.i_fps_num || !h->param.i_fps_den )
784 h->param.i_fps_num = 25;
785 h->param.i_fps_den = 1;
787 float fps = (float) h->param.i_fps_num / h->param.i_fps_den;
788 if( h->param.i_keyint_min == X264_KEYINT_MIN_AUTO )
789 h->param.i_keyint_min = X264_MIN( h->param.i_keyint_max / 10, fps );
790 h->param.i_keyint_min = x264_clip3( h->param.i_keyint_min, 1, h->param.i_keyint_max/2+1 );
791 h->param.rc.i_lookahead = x264_clip3( h->param.rc.i_lookahead, 0, X264_LOOKAHEAD_MAX );
793 int maxrate = X264_MAX( h->param.rc.i_vbv_max_bitrate, h->param.rc.i_bitrate );
794 float bufsize = maxrate ? (float)h->param.rc.i_vbv_buffer_size / maxrate : 0;
795 h->param.rc.i_lookahead = X264_MIN( h->param.rc.i_lookahead, X264_MAX( h->param.i_keyint_max, bufsize*fps ) );
798 if( !h->param.i_timebase_num || !h->param.i_timebase_den || !(h->param.b_vfr_input || h->param.b_pulldown) )
800 h->param.i_timebase_num = h->param.i_fps_den;
801 h->param.i_timebase_den = h->param.i_fps_num;
804 h->param.rc.f_qcompress = x264_clip3f( h->param.rc.f_qcompress, 0.0, 1.0 );
805 if( h->param.i_keyint_max == 1 || h->param.rc.f_qcompress == 1 )
806 h->param.rc.b_mb_tree = 0;
807 if( (!h->param.b_intra_refresh && h->param.i_keyint_max != X264_KEYINT_MAX_INFINITE) &&
808 !h->param.rc.i_lookahead && h->param.rc.b_mb_tree )
810 x264_log( h, X264_LOG_WARNING, "lookaheadless mb-tree requires intra refresh or infinite keyint\n" );
811 h->param.rc.b_mb_tree = 0;
813 if( b_open && h->param.rc.b_stat_read )
814 h->param.rc.i_lookahead = 0;
816 if( h->param.i_sync_lookahead < 0 )
817 h->param.i_sync_lookahead = h->param.i_bframe + 1;
818 h->param.i_sync_lookahead = X264_MIN( h->param.i_sync_lookahead, X264_LOOKAHEAD_MAX );
819 if( h->param.rc.b_stat_read || h->i_thread_frames == 1 )
820 h->param.i_sync_lookahead = 0;
822 h->param.i_sync_lookahead = 0;
825 h->param.i_deblocking_filter_alphac0 = x264_clip3( h->param.i_deblocking_filter_alphac0, -6, 6 );
826 h->param.i_deblocking_filter_beta = x264_clip3( h->param.i_deblocking_filter_beta, -6, 6 );
827 h->param.analyse.i_luma_deadzone[0] = x264_clip3( h->param.analyse.i_luma_deadzone[0], 0, 32 );
828 h->param.analyse.i_luma_deadzone[1] = x264_clip3( h->param.analyse.i_luma_deadzone[1], 0, 32 );
830 h->param.i_cabac_init_idc = x264_clip3( h->param.i_cabac_init_idc, 0, 2 );
832 if( h->param.i_cqm_preset < X264_CQM_FLAT || h->param.i_cqm_preset > X264_CQM_CUSTOM )
833 h->param.i_cqm_preset = X264_CQM_FLAT;
835 if( h->param.analyse.i_me_method < X264_ME_DIA ||
836 h->param.analyse.i_me_method > X264_ME_TESA )
837 h->param.analyse.i_me_method = X264_ME_HEX;
838 h->param.analyse.i_me_range = x264_clip3( h->param.analyse.i_me_range, 4, 1024 );
839 if( h->param.analyse.i_me_range > 16 && h->param.analyse.i_me_method <= X264_ME_HEX )
840 h->param.analyse.i_me_range = 16;
841 if( h->param.analyse.i_me_method == X264_ME_TESA &&
842 (h->mb.b_lossless || h->param.analyse.i_subpel_refine <= 1) )
843 h->param.analyse.i_me_method = X264_ME_ESA;
844 h->param.analyse.b_mixed_references = h->param.analyse.b_mixed_references && h->param.i_frame_reference > 1;
845 h->param.analyse.inter &= X264_ANALYSE_PSUB16x16|X264_ANALYSE_PSUB8x8|X264_ANALYSE_BSUB16x16|
846 X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
847 h->param.analyse.intra &= X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
848 if( !(h->param.analyse.inter & X264_ANALYSE_PSUB16x16) )
849 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
850 if( !h->param.analyse.b_transform_8x8 )
852 h->param.analyse.inter &= ~X264_ANALYSE_I8x8;
853 h->param.analyse.intra &= ~X264_ANALYSE_I8x8;
855 h->param.analyse.i_trellis = x264_clip3( h->param.analyse.i_trellis, 0, 2 );
856 h->param.rc.i_aq_mode = x264_clip3( h->param.rc.i_aq_mode, 0, 2 );
857 h->param.rc.f_aq_strength = x264_clip3f( h->param.rc.f_aq_strength, 0, 3 );
858 if( h->param.rc.f_aq_strength == 0 )
859 h->param.rc.i_aq_mode = 0;
861 if( h->param.i_log_level < X264_LOG_INFO )
863 h->param.analyse.b_psnr = 0;
864 h->param.analyse.b_ssim = 0;
866 /* Warn users trying to measure PSNR/SSIM with psy opts on. */
867 if( b_open && (h->param.analyse.b_psnr || h->param.analyse.b_ssim) )
871 if( h->param.analyse.b_psy )
873 s = h->param.analyse.b_psnr ? "psnr" : "ssim";
874 x264_log( h, X264_LOG_WARNING, "--%s used with psy on: results will be invalid!\n", s );
876 else if( !h->param.rc.i_aq_mode && h->param.analyse.b_ssim )
878 x264_log( h, X264_LOG_WARNING, "--ssim used with AQ off: results will be invalid!\n" );
881 else if( h->param.rc.i_aq_mode && h->param.analyse.b_psnr )
883 x264_log( h, X264_LOG_WARNING, "--psnr used with AQ on: results will be invalid!\n" );
887 x264_log( h, X264_LOG_WARNING, "--tune %s should be used if attempting to benchmark %s!\n", s, s );
890 if( !h->param.analyse.b_psy )
892 h->param.analyse.f_psy_rd = 0;
893 h->param.analyse.f_psy_trellis = 0;
895 h->param.analyse.f_psy_rd = x264_clip3f( h->param.analyse.f_psy_rd, 0, 10 );
896 h->param.analyse.f_psy_trellis = x264_clip3f( h->param.analyse.f_psy_trellis, 0, 10 );
897 h->mb.i_psy_rd = h->param.analyse.i_subpel_refine >= 6 ? FIX8( h->param.analyse.f_psy_rd ) : 0;
898 h->mb.i_psy_trellis = h->param.analyse.i_trellis ? FIX8( h->param.analyse.f_psy_trellis / 4 ) : 0;
899 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -32, 32);
900 /* In 4:4:4 mode, chroma gets twice as much resolution, so we can halve its quality. */
901 if( b_open && i_csp >= X264_CSP_I444 && i_csp < X264_CSP_BGR && h->param.analyse.b_psy )
902 h->param.analyse.i_chroma_qp_offset += 6;
903 /* Psy RDO increases overall quantizers to improve the quality of luma--this indirectly hurts chroma quality */
904 /* so we lower the chroma QP offset to compensate */
905 if( b_open && h->mb.i_psy_rd )
906 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_rd < 0.25 ? 1 : 2;
907 /* Psy trellis has a similar effect. */
908 if( b_open && h->mb.i_psy_trellis )
909 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_trellis < 0.25 ? 1 : 2;
910 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
911 /* MB-tree requires AQ to be on, even if the strength is zero. */
912 if( !h->param.rc.i_aq_mode && h->param.rc.b_mb_tree )
914 h->param.rc.i_aq_mode = 1;
915 h->param.rc.f_aq_strength = 0;
917 h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
918 if( h->param.analyse.i_subpel_refine >= 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
919 h->param.analyse.i_subpel_refine = 9;
922 const x264_level_t *l = x264_levels;
923 if( h->param.i_level_idc < 0 )
925 int maxrate_bak = h->param.rc.i_vbv_max_bitrate;
926 if( h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.i_vbv_buffer_size <= 0 )
927 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate * 2;
928 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
929 do h->param.i_level_idc = l->level_idc;
930 while( l[1].level_idc && x264_validate_levels( h, 0 ) && l++ );
931 h->param.rc.i_vbv_max_bitrate = maxrate_bak;
935 while( l->level_idc && l->level_idc != h->param.i_level_idc )
937 if( l->level_idc == 0 )
939 x264_log( h, X264_LOG_ERROR, "invalid level_idc: %d\n", h->param.i_level_idc );
943 if( h->param.analyse.i_mv_range <= 0 )
944 h->param.analyse.i_mv_range = l->mv_range >> PARAM_INTERLACED;
946 h->param.analyse.i_mv_range = x264_clip3(h->param.analyse.i_mv_range, 32, 512 >> PARAM_INTERLACED);
949 h->param.analyse.i_weighted_pred = x264_clip3( h->param.analyse.i_weighted_pred, X264_WEIGHTP_NONE, X264_WEIGHTP_SMART );
951 if( h->param.i_lookahead_threads == X264_THREADS_AUTO )
953 if( h->param.b_sliced_threads )
954 h->param.i_lookahead_threads = h->param.i_threads;
957 /* If we're using much slower lookahead settings than encoding settings, it helps a lot to use
958 * more lookahead threads. This typically happens in the first pass of a two-pass encode, so
959 * try to guess at this sort of case.
961 * Tuned by a little bit of real encoding with the various presets. */
962 int badapt = h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS;
963 int subme = X264_MIN( h->param.analyse.i_subpel_refine / 3, 3 ) + (h->param.analyse.i_subpel_refine > 1);
964 int bframes = X264_MIN( (h->param.i_bframe - 1) / 3, 3 );
966 /* [b-adapt 0/1 vs 2][quantized subme][quantized bframes] */
967 static const uint8_t lookahead_thread_div[2][5][4] =
968 {{{6,6,6,6}, {3,3,3,3}, {4,4,4,4}, {6,6,6,6}, {12,12,12,12}},
969 {{3,2,1,1}, {2,1,1,1}, {4,3,2,1}, {6,4,3,2}, {12, 9, 6, 4}}};
971 h->param.i_lookahead_threads = h->param.i_threads / lookahead_thread_div[badapt][subme][bframes];
972 /* Since too many lookahead threads significantly degrades lookahead accuracy, limit auto
973 * lookahead threads to about 8 macroblock rows high each at worst. This number is chosen
974 * pretty much arbitrarily. */
975 h->param.i_lookahead_threads = X264_MIN( h->param.i_lookahead_threads, h->param.i_height / 128 );
978 h->param.i_lookahead_threads = x264_clip3( h->param.i_lookahead_threads, 1, X264_MIN( max_sliced_threads, X264_LOOKAHEAD_THREAD_MAX ) );
980 if( PARAM_INTERLACED )
982 if( h->param.analyse.i_me_method >= X264_ME_ESA )
984 x264_log( h, X264_LOG_WARNING, "interlace + me=esa is not implemented\n" );
985 h->param.analyse.i_me_method = X264_ME_UMH;
987 if( h->param.analyse.i_weighted_pred > 0 )
989 x264_log( h, X264_LOG_WARNING, "interlace + weightp is not implemented\n" );
990 h->param.analyse.i_weighted_pred = X264_WEIGHTP_NONE;
994 if( !h->param.analyse.i_weighted_pred && h->param.rc.b_mb_tree && h->param.analyse.b_psy )
995 h->param.analyse.i_weighted_pred = X264_WEIGHTP_FAKE;
997 if( h->i_thread_frames > 1 )
999 int r = h->param.analyse.i_mv_range_thread;
1003 // half of the available space is reserved and divided evenly among the threads,
1004 // the rest is allocated to whichever thread is far enough ahead to use it.
1005 // reserving more space increases quality for some videos, but costs more time
1006 // in thread synchronization.
1007 int max_range = (h->param.i_height + X264_THREAD_HEIGHT) / h->i_thread_frames - X264_THREAD_HEIGHT;
1010 r = X264_MAX( r, h->param.analyse.i_me_range );
1011 r = X264_MIN( r, h->param.analyse.i_mv_range );
1012 // round up to use the whole mb row
1013 r2 = (r & ~15) + ((-X264_THREAD_HEIGHT) & 15);
1016 x264_log( h, X264_LOG_DEBUG, "using mv_range_thread = %d\n", r2 );
1017 h->param.analyse.i_mv_range_thread = r2;
1020 if( h->param.rc.f_rate_tolerance < 0 )
1021 h->param.rc.f_rate_tolerance = 0;
1022 if( h->param.rc.f_qblur < 0 )
1023 h->param.rc.f_qblur = 0;
1024 if( h->param.rc.f_complexity_blur < 0 )
1025 h->param.rc.f_complexity_blur = 0;
1027 h->param.i_sps_id &= 31;
1029 if( PARAM_INTERLACED )
1030 h->param.b_pic_struct = 1;
1032 h->param.i_nal_hrd = x264_clip3( h->param.i_nal_hrd, X264_NAL_HRD_NONE, X264_NAL_HRD_CBR );
1034 if( h->param.i_nal_hrd && !h->param.rc.i_vbv_buffer_size )
1036 x264_log( h, X264_LOG_WARNING, "NAL HRD parameters require VBV parameters\n" );
1037 h->param.i_nal_hrd = X264_NAL_HRD_NONE;
1040 if( h->param.i_nal_hrd == X264_NAL_HRD_CBR &&
1041 (h->param.rc.i_bitrate != h->param.rc.i_vbv_max_bitrate || !h->param.rc.i_vbv_max_bitrate) )
1043 x264_log( h, X264_LOG_WARNING, "CBR HRD requires constant bitrate\n" );
1044 h->param.i_nal_hrd = X264_NAL_HRD_VBR;
1047 /* ensure the booleans are 0 or 1 so they can be used in math */
1048 #define BOOLIFY(x) h->param.x = !!h->param.x
1050 BOOLIFY( b_constrained_intra );
1051 BOOLIFY( b_deblocking_filter );
1052 BOOLIFY( b_deterministic );
1053 BOOLIFY( b_sliced_threads );
1054 BOOLIFY( b_interlaced );
1055 BOOLIFY( b_intra_refresh );
1056 BOOLIFY( b_visualize );
1058 BOOLIFY( b_repeat_headers );
1059 BOOLIFY( b_annexb );
1060 BOOLIFY( b_vfr_input );
1061 BOOLIFY( b_pulldown );
1063 BOOLIFY( b_pic_struct );
1064 BOOLIFY( b_fake_interlaced );
1065 BOOLIFY( b_open_gop );
1066 BOOLIFY( b_bluray_compat );
1067 BOOLIFY( b_stitchable );
1068 BOOLIFY( b_full_recon );
1069 BOOLIFY( b_opencl );
1070 BOOLIFY( analyse.b_transform_8x8 );
1071 BOOLIFY( analyse.b_weighted_bipred );
1072 BOOLIFY( analyse.b_chroma_me );
1073 BOOLIFY( analyse.b_mixed_references );
1074 BOOLIFY( analyse.b_fast_pskip );
1075 BOOLIFY( analyse.b_dct_decimate );
1076 BOOLIFY( analyse.b_psy );
1077 BOOLIFY( analyse.b_psnr );
1078 BOOLIFY( analyse.b_ssim );
1079 BOOLIFY( rc.b_stat_write );
1080 BOOLIFY( rc.b_stat_read );
1081 BOOLIFY( rc.b_mb_tree );
1087 static void mbcmp_init( x264_t *h )
1089 int satd = !h->mb.b_lossless && h->param.analyse.i_subpel_refine > 1;
1090 memcpy( h->pixf.mbcmp, satd ? h->pixf.satd : h->pixf.sad_aligned, sizeof(h->pixf.mbcmp) );
1091 memcpy( h->pixf.mbcmp_unaligned, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.mbcmp_unaligned) );
1092 h->pixf.intra_mbcmp_x3_16x16 = satd ? h->pixf.intra_satd_x3_16x16 : h->pixf.intra_sad_x3_16x16;
1093 h->pixf.intra_mbcmp_x3_8x16c = satd ? h->pixf.intra_satd_x3_8x16c : h->pixf.intra_sad_x3_8x16c;
1094 h->pixf.intra_mbcmp_x3_8x8c = satd ? h->pixf.intra_satd_x3_8x8c : h->pixf.intra_sad_x3_8x8c;
1095 h->pixf.intra_mbcmp_x3_8x8 = satd ? h->pixf.intra_sa8d_x3_8x8 : h->pixf.intra_sad_x3_8x8;
1096 h->pixf.intra_mbcmp_x3_4x4 = satd ? h->pixf.intra_satd_x3_4x4 : h->pixf.intra_sad_x3_4x4;
1097 h->pixf.intra_mbcmp_x9_4x4 = h->param.b_cpu_independent || h->mb.b_lossless ? NULL
1098 : satd ? h->pixf.intra_satd_x9_4x4 : h->pixf.intra_sad_x9_4x4;
1099 h->pixf.intra_mbcmp_x9_8x8 = h->param.b_cpu_independent || h->mb.b_lossless ? NULL
1100 : satd ? h->pixf.intra_sa8d_x9_8x8 : h->pixf.intra_sad_x9_8x8;
1101 satd &= h->param.analyse.i_me_method == X264_ME_TESA;
1102 memcpy( h->pixf.fpelcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.fpelcmp) );
1103 memcpy( h->pixf.fpelcmp_x3, satd ? h->pixf.satd_x3 : h->pixf.sad_x3, sizeof(h->pixf.fpelcmp_x3) );
1104 memcpy( h->pixf.fpelcmp_x4, satd ? h->pixf.satd_x4 : h->pixf.sad_x4, sizeof(h->pixf.fpelcmp_x4) );
1107 static void chroma_dsp_init( x264_t *h )
1109 memcpy( h->luma2chroma_pixel, x264_luma2chroma_pixel[CHROMA_FORMAT], sizeof(h->luma2chroma_pixel) );
1111 switch( CHROMA_FORMAT )
1114 memcpy( h->predict_chroma, h->predict_8x8c, sizeof(h->predict_chroma) );
1115 h->mc.prefetch_fenc = h->mc.prefetch_fenc_420;
1116 h->loopf.deblock_chroma[0] = h->loopf.deblock_h_chroma_420;
1117 h->loopf.deblock_chroma_intra[0] = h->loopf.deblock_h_chroma_420_intra;
1118 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_chroma_420_mbaff;
1119 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_chroma_420_intra_mbaff;
1120 h->pixf.intra_mbcmp_x3_chroma = h->pixf.intra_mbcmp_x3_8x8c;
1121 h->quantf.coeff_last[DCT_CHROMA_DC] = h->quantf.coeff_last4;
1122 h->quantf.coeff_level_run[DCT_CHROMA_DC] = h->quantf.coeff_level_run4;
1125 memcpy( h->predict_chroma, h->predict_8x16c, sizeof(h->predict_chroma) );
1126 h->mc.prefetch_fenc = h->mc.prefetch_fenc_422;
1127 h->loopf.deblock_chroma[0] = h->loopf.deblock_h_chroma_422;
1128 h->loopf.deblock_chroma_intra[0] = h->loopf.deblock_h_chroma_422_intra;
1129 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_chroma_422_mbaff;
1130 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_chroma_422_intra_mbaff;
1131 h->pixf.intra_mbcmp_x3_chroma = h->pixf.intra_mbcmp_x3_8x16c;
1132 h->quantf.coeff_last[DCT_CHROMA_DC] = h->quantf.coeff_last8;
1133 h->quantf.coeff_level_run[DCT_CHROMA_DC] = h->quantf.coeff_level_run8;
1136 h->mc.prefetch_fenc = h->mc.prefetch_fenc_422; /* FIXME: doesn't cover V plane */
1137 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_luma_mbaff;
1138 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_luma_intra_mbaff;
1143 static void x264_set_aspect_ratio( x264_t *h, x264_param_t *param, int initial )
1146 if( param->vui.i_sar_width > 0 && param->vui.i_sar_height > 0 )
1148 uint32_t i_w = param->vui.i_sar_width;
1149 uint32_t i_h = param->vui.i_sar_height;
1150 uint32_t old_w = h->param.vui.i_sar_width;
1151 uint32_t old_h = h->param.vui.i_sar_height;
1153 x264_reduce_fraction( &i_w, &i_h );
1155 while( i_w > 65535 || i_h > 65535 )
1161 x264_reduce_fraction( &i_w, &i_h );
1163 if( i_w != old_w || i_h != old_h || initial )
1165 h->param.vui.i_sar_width = 0;
1166 h->param.vui.i_sar_height = 0;
1167 if( i_w == 0 || i_h == 0 )
1168 x264_log( h, X264_LOG_WARNING, "cannot create valid sample aspect ratio\n" );
1171 x264_log( h, initial?X264_LOG_INFO:X264_LOG_DEBUG, "using SAR=%d/%d\n", i_w, i_h );
1172 h->param.vui.i_sar_width = i_w;
1173 h->param.vui.i_sar_height = i_h;
1175 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1180 /****************************************************************************
1181 * x264_encoder_open:
1182 ****************************************************************************/
1183 x264_t *x264_encoder_open( x264_param_t *param )
1187 int qp, i_slicetype_length;
1189 CHECKED_MALLOCZERO( h, sizeof(x264_t) );
1191 /* Create a copy of param */
1192 memcpy( &h->param, param, sizeof(x264_param_t) );
1194 if( param->param_free )
1195 param->param_free( param );
1197 if( x264_threading_init() )
1199 x264_log( h, X264_LOG_ERROR, "unable to initialize threading\n" );
1203 if( x264_validate_parameters( h, 1 ) < 0 )
1206 if( h->param.psz_cqm_file )
1207 if( x264_cqm_parse_file( h, h->param.psz_cqm_file ) < 0 )
1210 if( h->param.rc.psz_stat_out )
1211 h->param.rc.psz_stat_out = strdup( h->param.rc.psz_stat_out );
1212 if( h->param.rc.psz_stat_in )
1213 h->param.rc.psz_stat_in = strdup( h->param.rc.psz_stat_in );
1215 x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den );
1216 x264_reduce_fraction( &h->param.i_timebase_num, &h->param.i_timebase_den );
1221 h->i_idr_pic_id = 0;
1223 if( (uint64_t)h->param.i_timebase_den * 2 > UINT32_MAX )
1225 x264_log( h, X264_LOG_ERROR, "Effective timebase denominator %u exceeds H.264 maximum\n", h->param.i_timebase_den );
1229 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1230 x264_pps_init( h->pps, h->param.i_sps_id, &h->param, h->sps );
1232 x264_set_aspect_ratio( h, &h->param, 1 );
1234 x264_validate_levels( h, 1 );
1236 h->chroma_qp_table = i_chroma_qp_table + 12 + h->pps->i_chroma_qp_index_offset;
1238 if( x264_cqm_init( h ) < 0 )
1241 h->mb.i_mb_width = h->sps->i_mb_width;
1242 h->mb.i_mb_height = h->sps->i_mb_height;
1243 h->mb.i_mb_count = h->mb.i_mb_width * h->mb.i_mb_height;
1245 h->mb.chroma_h_shift = CHROMA_FORMAT == CHROMA_420 || CHROMA_FORMAT == CHROMA_422;
1246 h->mb.chroma_v_shift = CHROMA_FORMAT == CHROMA_420;
1248 /* Adaptive MBAFF and subme 0 are not supported as we require halving motion
1249 * vectors during prediction, resulting in hpel mvs.
1250 * The chosen solution is to make MBAFF non-adaptive in this case. */
1251 h->mb.b_adaptive_mbaff = PARAM_INTERLACED && h->param.analyse.i_subpel_refine;
1254 if( h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS && !h->param.rc.b_stat_read )
1255 h->frames.i_delay = X264_MAX(h->param.i_bframe,3)*4;
1257 h->frames.i_delay = h->param.i_bframe;
1258 if( h->param.rc.b_mb_tree || h->param.rc.i_vbv_buffer_size )
1259 h->frames.i_delay = X264_MAX( h->frames.i_delay, h->param.rc.i_lookahead );
1260 i_slicetype_length = h->frames.i_delay;
1261 h->frames.i_delay += h->i_thread_frames - 1;
1262 h->frames.i_delay += h->param.i_sync_lookahead;
1263 h->frames.i_delay += h->param.b_vfr_input;
1264 h->frames.i_bframe_delay = h->param.i_bframe ? (h->param.i_bframe_pyramid ? 2 : 1) : 0;
1266 h->frames.i_max_ref0 = h->param.i_frame_reference;
1267 h->frames.i_max_ref1 = X264_MIN( h->sps->vui.i_num_reorder_frames, h->param.i_frame_reference );
1268 h->frames.i_max_dpb = h->sps->vui.i_max_dec_frame_buffering;
1269 h->frames.b_have_lowres = !h->param.rc.b_stat_read
1270 && ( h->param.rc.i_rc_method == X264_RC_ABR
1271 || h->param.rc.i_rc_method == X264_RC_CRF
1272 || h->param.i_bframe_adaptive
1273 || h->param.i_scenecut_threshold
1274 || h->param.rc.b_mb_tree
1275 || h->param.analyse.i_weighted_pred );
1276 h->frames.b_have_lowres |= h->param.rc.b_stat_read && h->param.rc.i_vbv_buffer_size > 0;
1277 h->frames.b_have_sub8x8_esa = !!(h->param.analyse.inter & X264_ANALYSE_PSUB8x8);
1279 h->frames.i_last_idr =
1280 h->frames.i_last_keyframe = - h->param.i_keyint_max;
1281 h->frames.i_input = 0;
1282 h->frames.i_largest_pts = h->frames.i_second_largest_pts = -1;
1283 h->frames.i_poc_last_open_gop = -1;
1285 CHECKED_MALLOCZERO( h->frames.unused[0], (h->frames.i_delay + 3) * sizeof(x264_frame_t *) );
1286 /* Allocate room for max refs plus a few extra just in case. */
1287 CHECKED_MALLOCZERO( h->frames.unused[1], (h->i_thread_frames + X264_REF_MAX + 4) * sizeof(x264_frame_t *) );
1288 CHECKED_MALLOCZERO( h->frames.current, (h->param.i_sync_lookahead + h->param.i_bframe
1289 + h->i_thread_frames + 3) * sizeof(x264_frame_t *) );
1290 if( h->param.analyse.i_weighted_pred > 0 )
1291 CHECKED_MALLOCZERO( h->frames.blank_unused, h->i_thread_frames * 4 * sizeof(x264_frame_t *) );
1292 h->i_ref[0] = h->i_ref[1] = 0;
1293 h->i_cpb_delay = h->i_coded_fields = h->i_disp_fields = 0;
1294 h->i_prev_duration = ((uint64_t)h->param.i_fps_den * h->sps->vui.i_time_scale) / ((uint64_t)h->param.i_fps_num * h->sps->vui.i_num_units_in_tick);
1295 h->i_disp_fields_last_frame = -1;
1298 /* init CPU functions */
1299 x264_predict_16x16_init( h->param.cpu, h->predict_16x16 );
1300 x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c );
1301 x264_predict_8x16c_init( h->param.cpu, h->predict_8x16c );
1302 x264_predict_8x8_init( h->param.cpu, h->predict_8x8, &h->predict_8x8_filter );
1303 x264_predict_4x4_init( h->param.cpu, h->predict_4x4 );
1304 x264_pixel_init( h->param.cpu, &h->pixf );
1305 x264_dct_init( h->param.cpu, &h->dctf );
1306 x264_zigzag_init( h->param.cpu, &h->zigzagf_progressive, &h->zigzagf_interlaced );
1307 memcpy( &h->zigzagf, PARAM_INTERLACED ? &h->zigzagf_interlaced : &h->zigzagf_progressive, sizeof(h->zigzagf) );
1308 x264_mc_init( h->param.cpu, &h->mc, h->param.b_cpu_independent );
1309 x264_quant_init( h, h->param.cpu, &h->quantf );
1310 x264_deblock_init( h->param.cpu, &h->loopf, PARAM_INTERLACED );
1311 x264_bitstream_init( h->param.cpu, &h->bsf );
1312 if( h->param.b_cabac )
1313 x264_cabac_init( h );
1315 x264_stack_align( x264_cavlc_init, h );
1318 chroma_dsp_init( h );
1320 p = buf + sprintf( buf, "using cpu capabilities:" );
1321 for( int i = 0; x264_cpu_names[i].flags; i++ )
1323 if( !strcmp(x264_cpu_names[i].name, "SSE")
1324 && h->param.cpu & (X264_CPU_SSE2) )
1326 if( !strcmp(x264_cpu_names[i].name, "SSE2")
1327 && h->param.cpu & (X264_CPU_SSE2_IS_FAST|X264_CPU_SSE2_IS_SLOW) )
1329 if( !strcmp(x264_cpu_names[i].name, "SSE3")
1330 && (h->param.cpu & X264_CPU_SSSE3 || !(h->param.cpu & X264_CPU_CACHELINE_64)) )
1332 if( !strcmp(x264_cpu_names[i].name, "SSE4.1")
1333 && (h->param.cpu & X264_CPU_SSE42) )
1335 if( !strcmp(x264_cpu_names[i].name, "BMI1")
1336 && (h->param.cpu & X264_CPU_BMI2) )
1338 if( (h->param.cpu & x264_cpu_names[i].flags) == x264_cpu_names[i].flags
1339 && (!i || x264_cpu_names[i].flags != x264_cpu_names[i-1].flags) )
1340 p += sprintf( p, " %s", x264_cpu_names[i].name );
1343 p += sprintf( p, " none!" );
1344 x264_log( h, X264_LOG_INFO, "%s\n", buf );
1346 float *logs = x264_analyse_prepare_costs( h );
1349 for( qp = X264_MIN( h->param.rc.i_qp_min, QP_MAX_SPEC ); qp <= h->param.rc.i_qp_max; qp++ )
1350 if( x264_analyse_init_costs( h, logs, qp ) )
1352 if( x264_analyse_init_costs( h, logs, X264_LOOKAHEAD_QP ) )
1356 static const uint16_t cost_mv_correct[7] = { 24, 47, 95, 189, 379, 757, 1515 };
1357 /* Checks for known miscompilation issues. */
1358 if( h->cost_mv[X264_LOOKAHEAD_QP][2013] != cost_mv_correct[BIT_DEPTH-8] )
1360 x264_log( h, X264_LOG_ERROR, "MV cost test failed: x264 has been miscompiled!\n" );
1364 /* Must be volatile or else GCC will optimize it out. */
1365 volatile int temp = 392;
1366 if( x264_clz( temp ) != 23 )
1368 x264_log( h, X264_LOG_ERROR, "CLZ test failed: x264 has been miscompiled!\n" );
1369 #if ARCH_X86 || ARCH_X86_64
1370 x264_log( h, X264_LOG_ERROR, "Are you attempting to run an SSE4a/LZCNT-targeted build on a CPU that\n" );
1371 x264_log( h, X264_LOG_ERROR, "doesn't support it?\n" );
1377 h->out.i_bitstream = X264_MAX( 1000000, h->param.i_width * h->param.i_height * 4
1378 * ( h->param.rc.i_rc_method == X264_RC_ABR ? pow( 0.95, h->param.rc.i_qp_min )
1379 : pow( 0.95, h->param.rc.i_qp_constant ) * X264_MAX( 1, h->param.rc.f_ip_factor )));
1381 h->nal_buffer_size = h->out.i_bitstream * 3/2 + 4 + 64; /* +4 for startcode, +64 for nal_escape assembly padding */
1382 CHECKED_MALLOC( h->nal_buffer, h->nal_buffer_size );
1384 if( h->param.i_threads > 1 &&
1385 x264_threadpool_init( &h->threadpool, h->param.i_threads, (void*)x264_encoder_thread_init, h ) )
1387 if( h->param.i_lookahead_threads > 1 &&
1388 x264_threadpool_init( &h->lookaheadpool, h->param.i_lookahead_threads, NULL, NULL ) )
1392 if( h->param.b_opencl )
1394 h->opencl.ocl = x264_opencl_load_library();
1395 if( !h->opencl.ocl )
1397 x264_log( h, X264_LOG_WARNING, "failed to load OpenCL\n" );
1398 h->param.b_opencl = 0;
1404 for( int i = 1; i < h->param.i_threads + !!h->param.i_sync_lookahead; i++ )
1405 CHECKED_MALLOC( h->thread[i], sizeof(x264_t) );
1406 if( h->param.i_lookahead_threads > 1 )
1407 for( int i = 0; i < h->param.i_lookahead_threads; i++ )
1409 CHECKED_MALLOC( h->lookahead_thread[i], sizeof(x264_t) );
1410 *h->lookahead_thread[i] = *h;
1413 for( int i = 0; i < h->param.i_threads; i++ )
1415 int init_nal_count = h->param.i_slice_count + 3;
1416 int allocate_threadlocal_data = !h->param.b_sliced_threads || !i;
1420 if( x264_pthread_mutex_init( &h->thread[i]->mutex, NULL ) )
1422 if( x264_pthread_cond_init( &h->thread[i]->cv, NULL ) )
1425 if( allocate_threadlocal_data )
1427 h->thread[i]->fdec = x264_frame_pop_unused( h, 1 );
1428 if( !h->thread[i]->fdec )
1432 h->thread[i]->fdec = h->thread[0]->fdec;
1434 CHECKED_MALLOC( h->thread[i]->out.p_bitstream, h->out.i_bitstream );
1435 /* Start each thread with room for init_nal_count NAL units; it'll realloc later if needed. */
1436 CHECKED_MALLOC( h->thread[i]->out.nal, init_nal_count*sizeof(x264_nal_t) );
1437 h->thread[i]->out.i_nals_allocated = init_nal_count;
1439 if( allocate_threadlocal_data && x264_macroblock_cache_allocate( h->thread[i] ) < 0 )
1444 if( h->param.b_opencl && x264_opencl_lookahead_init( h ) < 0 )
1445 h->param.b_opencl = 0;
1448 if( x264_lookahead_init( h, i_slicetype_length ) )
1451 for( int i = 0; i < h->param.i_threads; i++ )
1452 if( x264_macroblock_thread_allocate( h->thread[i], 0 ) < 0 )
1455 if( x264_ratecontrol_new( h ) < 0 )
1458 if( h->param.i_nal_hrd )
1460 x264_log( h, X264_LOG_DEBUG, "HRD bitrate: %i bits/sec\n", h->sps->vui.hrd.i_bit_rate_unscaled );
1461 x264_log( h, X264_LOG_DEBUG, "CPB size: %i bits\n", h->sps->vui.hrd.i_cpb_size_unscaled );
1464 if( h->param.psz_dump_yuv )
1466 /* create or truncate the reconstructed video file */
1467 FILE *f = fopen( h->param.psz_dump_yuv, "w" );
1470 x264_log( h, X264_LOG_ERROR, "dump_yuv: can't write to %s\n", h->param.psz_dump_yuv );
1473 else if( !x264_is_regular_file( f ) )
1475 x264_log( h, X264_LOG_ERROR, "dump_yuv: incompatible with non-regular file %s\n", h->param.psz_dump_yuv );
1481 const char *profile = h->sps->i_profile_idc == PROFILE_BASELINE ? "Constrained Baseline" :
1482 h->sps->i_profile_idc == PROFILE_MAIN ? "Main" :
1483 h->sps->i_profile_idc == PROFILE_HIGH ? "High" :
1484 h->sps->i_profile_idc == PROFILE_HIGH10 ? (h->sps->b_constraint_set3 == 1 ? "High 10 Intra" : "High 10") :
1485 h->sps->i_profile_idc == PROFILE_HIGH422 ? (h->sps->b_constraint_set3 == 1 ? "High 4:2:2 Intra" : "High 4:2:2") :
1486 h->sps->b_constraint_set3 == 1 ? "High 4:4:4 Intra" : "High 4:4:4 Predictive";
1488 snprintf( level, sizeof(level), "%d.%d", h->sps->i_level_idc/10, h->sps->i_level_idc%10 );
1489 if( h->sps->i_level_idc == 9 || ( h->sps->i_level_idc == 11 && h->sps->b_constraint_set3 &&
1490 (h->sps->i_profile_idc == PROFILE_BASELINE || h->sps->i_profile_idc == PROFILE_MAIN) ) )
1491 strcpy( level, "1b" );
1493 if( h->sps->i_profile_idc < PROFILE_HIGH10 )
1495 x264_log( h, X264_LOG_INFO, "profile %s, level %s\n",
1500 static const char * const subsampling[4] = { "4:0:0", "4:2:0", "4:2:2", "4:4:4" };
1501 x264_log( h, X264_LOG_INFO, "profile %s, level %s, %s %d-bit\n",
1502 profile, level, subsampling[CHROMA_FORMAT], BIT_DEPTH );
1511 /****************************************************************************
1512 * x264_encoder_reconfig:
1513 ****************************************************************************/
1514 int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
1516 /* If the previous frame isn't done encoding, reconfiguring is probably dangerous. */
1517 if( h->param.b_sliced_threads )
1518 if( x264_threadpool_wait_all( h ) < 0 )
1521 int rc_reconfig = 0;
1522 h = h->thread[h->thread[0]->i_thread_phase];
1523 x264_set_aspect_ratio( h, param, 0 );
1524 #define COPY(var) h->param.var = param->var
1525 COPY( i_frame_reference ); // but never uses more refs than initially specified
1526 COPY( i_bframe_bias );
1527 if( h->param.i_scenecut_threshold )
1528 COPY( i_scenecut_threshold ); // can't turn it on or off, only vary the threshold
1529 COPY( b_deblocking_filter );
1530 COPY( i_deblocking_filter_alphac0 );
1531 COPY( i_deblocking_filter_beta );
1532 COPY( i_frame_packing );
1533 COPY( analyse.inter );
1534 COPY( analyse.intra );
1535 COPY( analyse.i_direct_mv_pred );
1536 /* Scratch buffer prevents me_range from being increased for esa/tesa */
1537 if( h->param.analyse.i_me_method < X264_ME_ESA || param->analyse.i_me_range < h->param.analyse.i_me_range )
1538 COPY( analyse.i_me_range );
1539 COPY( analyse.i_noise_reduction );
1540 /* We can't switch out of subme=0 during encoding. */
1541 if( h->param.analyse.i_subpel_refine )
1542 COPY( analyse.i_subpel_refine );
1543 COPY( analyse.i_trellis );
1544 COPY( analyse.b_chroma_me );
1545 COPY( analyse.b_dct_decimate );
1546 COPY( analyse.b_fast_pskip );
1547 COPY( analyse.b_mixed_references );
1548 COPY( analyse.f_psy_rd );
1549 COPY( analyse.f_psy_trellis );
1551 // can only twiddle these if they were enabled to begin with:
1552 if( h->param.analyse.i_me_method >= X264_ME_ESA || param->analyse.i_me_method < X264_ME_ESA )
1553 COPY( analyse.i_me_method );
1554 if( h->param.analyse.i_me_method >= X264_ME_ESA && !h->frames.b_have_sub8x8_esa )
1555 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
1556 if( h->pps->b_transform_8x8_mode )
1557 COPY( analyse.b_transform_8x8 );
1558 if( h->frames.i_max_ref1 > 1 )
1559 COPY( i_bframe_pyramid );
1560 COPY( i_slice_max_size );
1561 COPY( i_slice_max_mbs );
1562 COPY( i_slice_min_mbs );
1563 COPY( i_slice_count );
1564 COPY( i_slice_count_max );
1567 /* VBV can't be turned on if it wasn't on to begin with */
1568 if( h->param.rc.i_vbv_max_bitrate > 0 && h->param.rc.i_vbv_buffer_size > 0 &&
1569 param->rc.i_vbv_max_bitrate > 0 && param->rc.i_vbv_buffer_size > 0 )
1571 rc_reconfig |= h->param.rc.i_vbv_max_bitrate != param->rc.i_vbv_max_bitrate;
1572 rc_reconfig |= h->param.rc.i_vbv_buffer_size != param->rc.i_vbv_buffer_size;
1573 rc_reconfig |= h->param.rc.i_bitrate != param->rc.i_bitrate;
1574 COPY( rc.i_vbv_max_bitrate );
1575 COPY( rc.i_vbv_buffer_size );
1576 COPY( rc.i_bitrate );
1578 rc_reconfig |= h->param.rc.f_rf_constant != param->rc.f_rf_constant;
1579 rc_reconfig |= h->param.rc.f_rf_constant_max != param->rc.f_rf_constant_max;
1580 COPY( rc.f_rf_constant );
1581 COPY( rc.f_rf_constant_max );
1586 int ret = x264_validate_parameters( h, 0 );
1588 /* Supported reconfiguration options (1-pass only):
1592 * bitrate (CBR only) */
1593 if( !ret && rc_reconfig )
1594 x264_ratecontrol_init_reconfigurable( h, 0 );
1599 /****************************************************************************
1600 * x264_encoder_parameters:
1601 ****************************************************************************/
1602 void x264_encoder_parameters( x264_t *h, x264_param_t *param )
1604 memcpy( param, &h->thread[h->i_thread_phase]->param, sizeof(x264_param_t) );
1607 /* internal usage */
1608 static void x264_nal_start( x264_t *h, int i_type, int i_ref_idc )
1610 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1612 nal->i_ref_idc = i_ref_idc;
1613 nal->i_type = i_type;
1614 nal->b_long_startcode = 1;
1617 nal->p_payload= &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1620 /* if number of allocated nals is not enough, re-allocate a larger one. */
1621 static int x264_nal_check_buffer( x264_t *h )
1623 if( h->out.i_nal >= h->out.i_nals_allocated )
1625 x264_nal_t *new_out = x264_malloc( sizeof(x264_nal_t) * (h->out.i_nals_allocated*2) );
1628 memcpy( new_out, h->out.nal, sizeof(x264_nal_t) * (h->out.i_nals_allocated) );
1629 x264_free( h->out.nal );
1630 h->out.nal = new_out;
1631 h->out.i_nals_allocated *= 2;
1636 static int x264_nal_end( x264_t *h )
1638 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1639 uint8_t *end = &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1640 nal->i_payload = end - nal->p_payload;
1641 /* Assembly implementation of nal_escape reads past the end of the input.
1642 * While undefined padding wouldn't actually affect the output, it makes valgrind unhappy. */
1643 memset( end, 0xff, 64 );
1644 if( h->param.nalu_process )
1645 h->param.nalu_process( h, nal, h->fenc->opaque );
1648 return x264_nal_check_buffer( h );
1651 static int x264_encoder_encapsulate_nals( x264_t *h, int start )
1653 x264_t *h0 = h->thread[0];
1654 int nal_size = 0, previous_nal_size = 0;
1656 if( h->param.nalu_process )
1658 for( int i = start; i < h->out.i_nal; i++ )
1659 nal_size += h->out.nal[i].i_payload;
1663 for( int i = 0; i < start; i++ )
1664 previous_nal_size += h->out.nal[i].i_payload;
1666 for( int i = start; i < h->out.i_nal; i++ )
1667 nal_size += h->out.nal[i].i_payload;
1669 /* Worst-case NAL unit escaping: reallocate the buffer if it's too small. */
1670 int necessary_size = previous_nal_size + nal_size * 3/2 + h->out.i_nal * 4 + 4 + 64;
1671 if( h0->nal_buffer_size < necessary_size )
1673 necessary_size *= 2;
1674 uint8_t *buf = x264_malloc( necessary_size );
1677 if( previous_nal_size )
1678 memcpy( buf, h0->nal_buffer, previous_nal_size );
1680 intptr_t delta = buf - h0->nal_buffer;
1681 for( int i = 0; i < start; i++ )
1682 h->out.nal[i].p_payload += delta;
1684 x264_free( h0->nal_buffer );
1685 h0->nal_buffer = buf;
1686 h0->nal_buffer_size = necessary_size;
1689 uint8_t *nal_buffer = h0->nal_buffer + previous_nal_size;
1691 for( int i = start; i < h->out.i_nal; i++ )
1693 h->out.nal[i].b_long_startcode = !i || h->out.nal[i].i_type == NAL_SPS || h->out.nal[i].i_type == NAL_PPS;
1694 x264_nal_encode( h, nal_buffer, &h->out.nal[i] );
1695 nal_buffer += h->out.nal[i].i_payload;
1700 return nal_buffer - (h0->nal_buffer + previous_nal_size);
1703 /****************************************************************************
1704 * x264_encoder_headers:
1705 ****************************************************************************/
1706 int x264_encoder_headers( x264_t *h, x264_nal_t **pp_nal, int *pi_nal )
1709 /* init bitstream context */
1711 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
1713 /* Write SEI, SPS and PPS. */
1715 /* generate sequence parameters */
1716 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
1717 x264_sps_write( &h->out.bs, h->sps );
1718 if( x264_nal_end( h ) )
1721 /* generate picture parameters */
1722 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
1723 x264_pps_write( &h->out.bs, h->sps, h->pps );
1724 if( x264_nal_end( h ) )
1727 /* identify ourselves */
1728 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
1729 if( x264_sei_version_write( h, &h->out.bs ) )
1731 if( x264_nal_end( h ) )
1734 frame_size = x264_encoder_encapsulate_nals( h, 0 );
1735 if( frame_size < 0 )
1739 *pi_nal = h->out.i_nal;
1740 *pp_nal = &h->out.nal[0];
1746 /* Check to see whether we have chosen a reference list ordering different
1747 * from the standard's default. */
1748 static inline void x264_reference_check_reorder( x264_t *h )
1750 /* The reorder check doesn't check for missing frames, so just
1751 * force a reorder if one of the reference list is corrupt. */
1752 for( int i = 0; h->frames.reference[i]; i++ )
1753 if( h->frames.reference[i]->b_corrupt )
1755 h->b_ref_reorder[0] = 1;
1758 for( int list = 0; list <= (h->sh.i_type == SLICE_TYPE_B); list++ )
1759 for( int i = 0; i < h->i_ref[list] - 1; i++ )
1761 int framenum_diff = h->fref[list][i+1]->i_frame_num - h->fref[list][i]->i_frame_num;
1762 int poc_diff = h->fref[list][i+1]->i_poc - h->fref[list][i]->i_poc;
1763 /* P and B-frames use different default orders. */
1764 if( h->sh.i_type == SLICE_TYPE_P ? framenum_diff > 0 : list == 1 ? poc_diff < 0 : poc_diff > 0 )
1766 h->b_ref_reorder[list] = 1;
1772 /* return -1 on failure, else return the index of the new reference frame */
1773 int x264_weighted_reference_duplicate( x264_t *h, int i_ref, const x264_weight_t *w )
1775 int i = h->i_ref[0];
1777 x264_frame_t *newframe;
1778 if( i <= 1 ) /* empty list, definitely can't duplicate frame */
1781 //Duplication is only used in X264_WEIGHTP_SMART
1782 if( h->param.analyse.i_weighted_pred != X264_WEIGHTP_SMART )
1785 /* Duplication is a hack to compensate for crappy rounding in motion compensation.
1786 * With high bit depth, it's not worth doing, so turn it off except in the case of
1787 * unweighted dupes. */
1788 if( BIT_DEPTH > 8 && w != x264_weight_none )
1791 newframe = x264_frame_pop_blank_unused( h );
1795 //FIXME: probably don't need to copy everything
1796 *newframe = *h->fref[0][i_ref];
1797 newframe->i_reference_count = 1;
1798 newframe->orig = h->fref[0][i_ref];
1799 newframe->b_duplicate = 1;
1800 memcpy( h->fenc->weight[j], w, sizeof(h->fenc->weight[i]) );
1802 /* shift the frames to make space for the dupe. */
1803 h->b_ref_reorder[0] = 1;
1804 if( h->i_ref[0] < X264_REF_MAX )
1806 h->fref[0][X264_REF_MAX-1] = NULL;
1807 x264_frame_unshift( &h->fref[0][j], newframe );
1812 static void x264_weighted_pred_init( x264_t *h )
1814 /* for now no analysis and set all weights to nothing */
1815 for( int i_ref = 0; i_ref < h->i_ref[0]; i_ref++ )
1816 h->fenc->weighted[i_ref] = h->fref[0][i_ref]->filtered[0][0];
1818 // FIXME: This only supports weighting of one reference frame
1819 // and duplicates of that frame.
1820 h->fenc->i_lines_weighted = 0;
1822 for( int i_ref = 0; i_ref < (h->i_ref[0] << SLICE_MBAFF); i_ref++ )
1823 for( int i = 0; i < 3; i++ )
1824 h->sh.weight[i_ref][i].weightfn = NULL;
1827 if( h->sh.i_type != SLICE_TYPE_P || h->param.analyse.i_weighted_pred <= 0 )
1830 int i_padv = PADV << PARAM_INTERLACED;
1832 int weightplane[2] = { 0, 0 };
1833 int buffer_next = 0;
1834 for( int i = 0; i < 3; i++ )
1836 for( int j = 0; j < h->i_ref[0]; j++ )
1838 if( h->fenc->weight[j][i].weightfn )
1840 h->sh.weight[j][i] = h->fenc->weight[j][i];
1841 // if weight is useless, don't write it to stream
1842 if( h->sh.weight[j][i].i_scale == 1<<h->sh.weight[j][i].i_denom && h->sh.weight[j][i].i_offset == 0 )
1843 h->sh.weight[j][i].weightfn = NULL;
1846 if( !weightplane[!!i] )
1848 weightplane[!!i] = 1;
1849 h->sh.weight[0][!!i].i_denom = denom = h->sh.weight[j][i].i_denom;
1850 assert( x264_clip3( denom, 0, 7 ) == denom );
1853 assert( h->sh.weight[j][i].i_denom == denom );
1856 h->fenc->weighted[j] = h->mb.p_weight_buf[buffer_next++] + h->fenc->i_stride[0] * i_padv + PADH;
1857 //scale full resolution frame
1858 if( h->param.i_threads == 1 )
1860 pixel *src = h->fref[0][j]->filtered[0][0] - h->fref[0][j]->i_stride[0]*i_padv - PADH;
1861 pixel *dst = h->fenc->weighted[j] - h->fenc->i_stride[0]*i_padv - PADH;
1862 int stride = h->fenc->i_stride[0];
1863 int width = h->fenc->i_width[0] + PADH*2;
1864 int height = h->fenc->i_lines[0] + i_padv*2;
1865 x264_weight_scale_plane( h, dst, stride, src, stride, width, height, &h->sh.weight[j][0] );
1866 h->fenc->i_lines_weighted = height;
1874 if( weightplane[1] )
1875 for( int i = 0; i < h->i_ref[0]; i++ )
1877 if( h->sh.weight[i][1].weightfn && !h->sh.weight[i][2].weightfn )
1879 h->sh.weight[i][2].i_scale = 1 << h->sh.weight[0][1].i_denom;
1880 h->sh.weight[i][2].i_offset = 0;
1882 else if( h->sh.weight[i][2].weightfn && !h->sh.weight[i][1].weightfn )
1884 h->sh.weight[i][1].i_scale = 1 << h->sh.weight[0][1].i_denom;
1885 h->sh.weight[i][1].i_offset = 0;
1889 if( !weightplane[0] )
1890 h->sh.weight[0][0].i_denom = 0;
1891 if( !weightplane[1] )
1892 h->sh.weight[0][1].i_denom = 0;
1893 h->sh.weight[0][2].i_denom = h->sh.weight[0][1].i_denom;
1896 static inline int x264_reference_distance( x264_t *h, x264_frame_t *frame )
1898 if( h->param.i_frame_packing == 5 )
1899 return abs((h->fenc->i_frame&~1) - (frame->i_frame&~1)) +
1900 ((h->fenc->i_frame&1) != (frame->i_frame&1));
1902 return abs(h->fenc->i_frame - frame->i_frame);
1905 static inline void x264_reference_build_list( x264_t *h, int i_poc )
1909 /* build ref list 0/1 */
1910 h->mb.pic.i_fref[0] = h->i_ref[0] = 0;
1911 h->mb.pic.i_fref[1] = h->i_ref[1] = 0;
1912 if( h->sh.i_type == SLICE_TYPE_I )
1915 for( int i = 0; h->frames.reference[i]; i++ )
1917 if( h->frames.reference[i]->b_corrupt )
1919 if( h->frames.reference[i]->i_poc < i_poc )
1920 h->fref[0][h->i_ref[0]++] = h->frames.reference[i];
1921 else if( h->frames.reference[i]->i_poc > i_poc )
1922 h->fref[1][h->i_ref[1]++] = h->frames.reference[i];
1925 /* Order reference lists by distance from the current frame. */
1926 for( int list = 0; list < 2; list++ )
1928 h->fref_nearest[list] = h->fref[list][0];
1932 for( int i = 0; i < h->i_ref[list] - 1; i++ )
1934 if( list ? h->fref[list][i+1]->i_poc < h->fref_nearest[list]->i_poc
1935 : h->fref[list][i+1]->i_poc > h->fref_nearest[list]->i_poc )
1936 h->fref_nearest[list] = h->fref[list][i+1];
1937 if( x264_reference_distance( h, h->fref[list][i] ) > x264_reference_distance( h, h->fref[list][i+1] ) )
1939 XCHG( x264_frame_t*, h->fref[list][i], h->fref[list][i+1] );
1947 if( h->sh.i_mmco_remove_from_end )
1948 for( int i = h->i_ref[0]-1; i >= h->i_ref[0] - h->sh.i_mmco_remove_from_end; i-- )
1950 int diff = h->i_frame_num - h->fref[0][i]->i_frame_num;
1951 h->sh.mmco[h->sh.i_mmco_command_count].i_poc = h->fref[0][i]->i_poc;
1952 h->sh.mmco[h->sh.i_mmco_command_count++].i_difference_of_pic_nums = diff;
1955 x264_reference_check_reorder( h );
1957 h->i_ref[1] = X264_MIN( h->i_ref[1], h->frames.i_max_ref1 );
1958 h->i_ref[0] = X264_MIN( h->i_ref[0], h->frames.i_max_ref0 );
1959 h->i_ref[0] = X264_MIN( h->i_ref[0], h->param.i_frame_reference ); // if reconfig() has lowered the limit
1961 /* For Blu-ray compliance, don't reference frames outside of the minigop. */
1962 if( IS_X264_TYPE_B( h->fenc->i_type ) && h->param.b_bluray_compat )
1963 h->i_ref[0] = X264_MIN( h->i_ref[0], IS_X264_TYPE_B( h->fref[0][0]->i_type ) + 1 );
1965 /* add duplicates */
1966 if( h->fenc->i_type == X264_TYPE_P )
1969 if( h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE )
1972 w[1].weightfn = w[2].weightfn = NULL;
1973 if( h->param.rc.b_stat_read )
1974 x264_ratecontrol_set_weights( h, h->fenc );
1976 if( !h->fenc->weight[0][0].weightfn )
1978 h->fenc->weight[0][0].i_denom = 0;
1979 SET_WEIGHT( w[0], 1, 1, 0, -1 );
1980 idx = x264_weighted_reference_duplicate( h, 0, w );
1984 if( h->fenc->weight[0][0].i_scale == 1<<h->fenc->weight[0][0].i_denom )
1986 SET_WEIGHT( h->fenc->weight[0][0], 1, 1, 0, h->fenc->weight[0][0].i_offset );
1988 x264_weighted_reference_duplicate( h, 0, x264_weight_none );
1989 if( h->fenc->weight[0][0].i_offset > -128 )
1991 w[0] = h->fenc->weight[0][0];
1993 h->mc.weight_cache( h, &w[0] );
1994 idx = x264_weighted_reference_duplicate( h, 0, w );
1998 h->mb.ref_blind_dupe = idx;
2001 assert( h->i_ref[0] + h->i_ref[1] <= X264_REF_MAX );
2002 h->mb.pic.i_fref[0] = h->i_ref[0];
2003 h->mb.pic.i_fref[1] = h->i_ref[1];
2006 static void x264_fdec_filter_row( x264_t *h, int mb_y, int pass )
2008 /* mb_y is the mb to be encoded next, not the mb to be filtered here */
2009 int b_hpel = h->fdec->b_kept_as_ref;
2010 int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
2011 int b_end = mb_y == h->i_threadslice_end;
2012 int b_measure_quality = 1;
2013 int min_y = mb_y - (1 << SLICE_MBAFF);
2014 int b_start = min_y == h->i_threadslice_start;
2015 /* Even in interlaced mode, deblocking never modifies more than 4 pixels
2016 * above each MB, as bS=4 doesn't happen for the top of interlaced mbpairs. */
2017 int minpix_y = min_y*16 - 4 * !b_start;
2018 int maxpix_y = mb_y*16 - 4 * !b_end;
2019 b_deblock &= b_hpel || h->param.b_full_recon || h->param.psz_dump_yuv;
2020 if( h->param.b_sliced_threads )
2024 /* During encode: only do deblock if asked for */
2027 b_deblock &= h->param.b_full_recon;
2030 /* During post-encode pass: do deblock if not done yet, do hpel for all
2031 * rows except those between slices. */
2033 b_deblock &= !h->param.b_full_recon;
2034 b_hpel &= !(b_start && min_y > 0);
2035 b_measure_quality = 0;
2037 /* Final pass: do the rows between slices in sequence. */
2040 b_measure_quality = 0;
2044 if( mb_y & SLICE_MBAFF )
2046 if( min_y < h->i_threadslice_start )
2050 for( int y = min_y; y < mb_y; y += (1 << SLICE_MBAFF) )
2051 x264_frame_deblock_row( h, y );
2053 /* FIXME: Prediction requires different borders for interlaced/progressive mc,
2054 * but the actual image data is equivalent. For now, maintain this
2055 * consistency by copying deblocked pixels between planes. */
2056 if( PARAM_INTERLACED && (!h->param.b_sliced_threads || pass == 1) )
2057 for( int p = 0; p < h->fdec->i_plane; p++ )
2058 for( int i = minpix_y>>(CHROMA_V_SHIFT && p); i < maxpix_y>>(CHROMA_V_SHIFT && p); i++ )
2059 memcpy( h->fdec->plane_fld[p] + i*h->fdec->i_stride[p],
2060 h->fdec->plane[p] + i*h->fdec->i_stride[p],
2061 h->mb.i_mb_width*16*sizeof(pixel) );
2063 if( h->fdec->b_kept_as_ref && (!h->param.b_sliced_threads || pass == 1) )
2064 x264_frame_expand_border( h, h->fdec, min_y );
2067 int end = mb_y == h->mb.i_mb_height;
2068 /* Can't do hpel until the previous slice is done encoding. */
2069 if( h->param.analyse.i_subpel_refine )
2071 x264_frame_filter( h, h->fdec, min_y, end );
2072 x264_frame_expand_border_filtered( h, h->fdec, min_y, end );
2076 if( SLICE_MBAFF && pass == 0 )
2077 for( int i = 0; i < 3; i++ )
2079 XCHG( pixel *, h->intra_border_backup[0][i], h->intra_border_backup[3][i] );
2080 XCHG( pixel *, h->intra_border_backup[1][i], h->intra_border_backup[4][i] );
2083 if( h->i_thread_frames > 1 && h->fdec->b_kept_as_ref )
2084 x264_frame_cond_broadcast( h->fdec, mb_y*16 + (b_end ? 10000 : -(X264_THREAD_HEIGHT << SLICE_MBAFF)) );
2086 if( b_measure_quality )
2088 maxpix_y = X264_MIN( maxpix_y, h->param.i_height );
2089 if( h->param.analyse.b_psnr )
2091 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
2092 h->stat.frame.i_ssd[p] += x264_pixel_ssd_wxh( &h->pixf,
2093 h->fdec->plane[p] + minpix_y * h->fdec->i_stride[p], h->fdec->i_stride[p],
2094 h->fenc->plane[p] + minpix_y * h->fenc->i_stride[p], h->fenc->i_stride[p],
2095 h->param.i_width, maxpix_y-minpix_y );
2098 uint64_t ssd_u, ssd_v;
2099 int v_shift = CHROMA_V_SHIFT;
2100 x264_pixel_ssd_nv12( &h->pixf,
2101 h->fdec->plane[1] + (minpix_y>>v_shift) * h->fdec->i_stride[1], h->fdec->i_stride[1],
2102 h->fenc->plane[1] + (minpix_y>>v_shift) * h->fenc->i_stride[1], h->fenc->i_stride[1],
2103 h->param.i_width>>1, (maxpix_y-minpix_y)>>v_shift, &ssd_u, &ssd_v );
2104 h->stat.frame.i_ssd[1] += ssd_u;
2105 h->stat.frame.i_ssd[2] += ssd_v;
2109 if( h->param.analyse.b_ssim )
2113 /* offset by 2 pixels to avoid alignment of ssim blocks with dct blocks,
2114 * and overlap by 4 */
2115 minpix_y += b_start ? 2 : -6;
2116 h->stat.frame.f_ssim +=
2117 x264_pixel_ssim_wxh( &h->pixf,
2118 h->fdec->plane[0] + 2+minpix_y*h->fdec->i_stride[0], h->fdec->i_stride[0],
2119 h->fenc->plane[0] + 2+minpix_y*h->fenc->i_stride[0], h->fenc->i_stride[0],
2120 h->param.i_width-2, maxpix_y-minpix_y, h->scratch_buffer, &ssim_cnt );
2121 h->stat.frame.i_ssim_cnt += ssim_cnt;
2126 static inline int x264_reference_update( x264_t *h )
2128 if( !h->fdec->b_kept_as_ref )
2130 if( h->i_thread_frames > 1 )
2132 x264_frame_push_unused( h, h->fdec );
2133 h->fdec = x264_frame_pop_unused( h, 1 );
2140 /* apply mmco from previous frame. */
2141 for( int i = 0; i < h->sh.i_mmco_command_count; i++ )
2142 for( int j = 0; h->frames.reference[j]; j++ )
2143 if( h->frames.reference[j]->i_poc == h->sh.mmco[i].i_poc )
2144 x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[j] ) );
2146 /* move frame in the buffer */
2147 x264_frame_push( h->frames.reference, h->fdec );
2148 if( h->frames.reference[h->sps->i_num_ref_frames] )
2149 x264_frame_push_unused( h, x264_frame_shift( h->frames.reference ) );
2150 h->fdec = x264_frame_pop_unused( h, 1 );
2156 static inline void x264_reference_reset( x264_t *h )
2158 while( h->frames.reference[0] )
2159 x264_frame_push_unused( h, x264_frame_pop( h->frames.reference ) );
2164 static inline void x264_reference_hierarchy_reset( x264_t *h )
2167 int b_hasdelayframe = 0;
2169 /* look for delay frames -- chain must only contain frames that are disposable */
2170 for( int i = 0; h->frames.current[i] && IS_DISPOSABLE( h->frames.current[i]->i_type ); i++ )
2171 b_hasdelayframe |= h->frames.current[i]->i_coded
2172 != h->frames.current[i]->i_frame + h->sps->vui.i_num_reorder_frames;
2174 /* This function must handle b-pyramid and clear frames for open-gop */
2175 if( h->param.i_bframe_pyramid != X264_B_PYRAMID_STRICT && !b_hasdelayframe && h->frames.i_poc_last_open_gop == -1 )
2178 /* Remove last BREF. There will never be old BREFs in the
2179 * dpb during a BREF decode when pyramid == STRICT */
2180 for( ref = 0; h->frames.reference[ref]; ref++ )
2182 if( ( h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT
2183 && h->frames.reference[ref]->i_type == X264_TYPE_BREF )
2184 || ( h->frames.reference[ref]->i_poc < h->frames.i_poc_last_open_gop
2185 && h->sh.i_type != SLICE_TYPE_B ) )
2187 int diff = h->i_frame_num - h->frames.reference[ref]->i_frame_num;
2188 h->sh.mmco[h->sh.i_mmco_command_count].i_difference_of_pic_nums = diff;
2189 h->sh.mmco[h->sh.i_mmco_command_count++].i_poc = h->frames.reference[ref]->i_poc;
2190 x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[ref] ) );
2191 h->b_ref_reorder[0] = 1;
2196 /* Prepare room in the dpb for the delayed display time of the later b-frame's */
2197 if( h->param.i_bframe_pyramid )
2198 h->sh.i_mmco_remove_from_end = X264_MAX( ref + 2 - h->frames.i_max_dpb, 0 );
2201 static inline void x264_slice_init( x264_t *h, int i_nal_type, int i_global_qp )
2203 /* ------------------------ Create slice header ----------------------- */
2204 if( i_nal_type == NAL_SLICE_IDR )
2206 x264_slice_header_init( h, &h->sh, h->sps, h->pps, h->i_idr_pic_id, h->i_frame_num, i_global_qp );
2209 h->i_idr_pic_id ^= 1;
2213 x264_slice_header_init( h, &h->sh, h->sps, h->pps, -1, h->i_frame_num, i_global_qp );
2215 h->sh.i_num_ref_idx_l0_active = h->i_ref[0] <= 0 ? 1 : h->i_ref[0];
2216 h->sh.i_num_ref_idx_l1_active = h->i_ref[1] <= 0 ? 1 : h->i_ref[1];
2217 if( h->sh.i_num_ref_idx_l0_active != h->pps->i_num_ref_idx_l0_default_active ||
2218 (h->sh.i_type == SLICE_TYPE_B && h->sh.i_num_ref_idx_l1_active != h->pps->i_num_ref_idx_l1_default_active) )
2220 h->sh.b_num_ref_idx_override = 1;
2224 if( h->fenc->i_type == X264_TYPE_BREF && h->param.b_bluray_compat && h->sh.i_mmco_command_count )
2227 h->sh_backup = h->sh;
2230 h->fdec->i_frame_num = h->sh.i_frame_num;
2232 if( h->sps->i_poc_type == 0 )
2234 h->sh.i_poc = h->fdec->i_poc;
2235 if( PARAM_INTERLACED )
2237 h->sh.i_delta_poc_bottom = h->param.b_tff ? 1 : -1;
2238 h->sh.i_poc += h->sh.i_delta_poc_bottom == -1;
2241 h->sh.i_delta_poc_bottom = 0;
2242 h->fdec->i_delta_poc[0] = h->sh.i_delta_poc_bottom == -1;
2243 h->fdec->i_delta_poc[1] = h->sh.i_delta_poc_bottom == 1;
2247 /* Nothing to do ? */
2250 x264_macroblock_slice_init( h );
2256 uint8_t cabac_prevbyte;
2259 x264_frame_stat_t stat;
2262 int field_decoding_flag;
2265 static ALWAYS_INLINE void x264_bitstream_backup( x264_t *h, x264_bs_bak_t *bak, int i_skip, int full )
2269 bak->stat = h->stat.frame;
2270 bak->last_qp = h->mb.i_last_qp;
2271 bak->last_dqp = h->mb.i_last_dqp;
2272 bak->field_decoding_flag = h->mb.field_decoding_flag;
2276 bak->stat.i_mv_bits = h->stat.frame.i_mv_bits;
2277 bak->stat.i_tex_bits = h->stat.frame.i_tex_bits;
2279 /* In the per-MB backup, we don't need the contexts because flushing the CABAC
2280 * encoder has no context dependency and in this case, a slice is ended (and
2281 * thus the content of all contexts are thrown away). */
2282 if( h->param.b_cabac )
2285 memcpy( &bak->cabac, &h->cabac, sizeof(x264_cabac_t) );
2287 memcpy( &bak->cabac, &h->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
2288 /* x264's CABAC writer modifies the previous byte during carry, so it has to be
2290 bak->cabac_prevbyte = h->cabac.p[-1];
2294 bak->bs = h->out.bs;
2299 static ALWAYS_INLINE void x264_bitstream_restore( x264_t *h, x264_bs_bak_t *bak, int *skip, int full )
2303 h->stat.frame = bak->stat;
2304 h->mb.i_last_qp = bak->last_qp;
2305 h->mb.i_last_dqp = bak->last_dqp;
2306 h->mb.field_decoding_flag = bak->field_decoding_flag;
2310 h->stat.frame.i_mv_bits = bak->stat.i_mv_bits;
2311 h->stat.frame.i_tex_bits = bak->stat.i_tex_bits;
2313 if( h->param.b_cabac )
2316 memcpy( &h->cabac, &bak->cabac, sizeof(x264_cabac_t) );
2318 memcpy( &h->cabac, &bak->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
2319 h->cabac.p[-1] = bak->cabac_prevbyte;
2323 h->out.bs = bak->bs;
2328 static int x264_slice_write( x264_t *h )
2331 int mb_xy, i_mb_x, i_mb_y;
2332 /* NALUs other than the first use a 3-byte startcode.
2333 * Add one extra byte for the rbsp, and one more for the final CABAC putbyte.
2334 * Then add an extra 5 bytes just in case, to account for random NAL escapes and
2335 * other inaccuracies. */
2336 int overhead_guess = (NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal)) + 1 + h->param.b_cabac + 5;
2337 int slice_max_size = h->param.i_slice_max_size > 0 ? (h->param.i_slice_max_size-overhead_guess)*8 : 0;
2338 int back_up_bitstream = slice_max_size || (!h->param.b_cabac && h->sps->i_profile_idc < PROFILE_HIGH);
2339 int starting_bits = bs_pos(&h->out.bs);
2340 int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
2341 int b_hpel = h->fdec->b_kept_as_ref;
2342 int orig_last_mb = h->sh.i_last_mb;
2343 int thread_last_mb = h->i_threadslice_end * h->mb.i_mb_width - 1;
2344 uint8_t *last_emu_check;
2345 #define BS_BAK_SLICE_MAX_SIZE 0
2346 #define BS_BAK_SLICE_MIN_MBS 1
2347 #define BS_BAK_ROW_VBV 2
2348 x264_bs_bak_t bs_bak[3];
2349 b_deblock &= b_hpel || h->param.b_full_recon || h->param.psz_dump_yuv;
2350 bs_realign( &h->out.bs );
2353 x264_nal_start( h, h->i_nal_type, h->i_nal_ref_idc );
2354 h->out.nal[h->out.i_nal].i_first_mb = h->sh.i_first_mb;
2357 x264_macroblock_thread_init( h );
2359 /* Set the QP equal to the first QP in the slice for more accurate CABAC initialization. */
2360 h->mb.i_mb_xy = h->sh.i_first_mb;
2361 h->sh.i_qp = x264_ratecontrol_mb_qp( h );
2362 h->sh.i_qp = SPEC_QP( h->sh.i_qp );
2363 h->sh.i_qp_delta = h->sh.i_qp - h->pps->i_pic_init_qp;
2365 x264_slice_header_write( &h->out.bs, &h->sh, h->i_nal_ref_idc );
2366 if( h->param.b_cabac )
2368 /* alignment needed */
2369 bs_align_1( &h->out.bs );
2372 x264_cabac_context_init( h, &h->cabac, h->sh.i_type, x264_clip3( h->sh.i_qp-QP_BD_OFFSET, 0, 51 ), h->sh.i_cabac_init_idc );
2373 x264_cabac_encode_init ( &h->cabac, h->out.bs.p, h->out.bs.p_end );
2374 last_emu_check = h->cabac.p;
2377 last_emu_check = h->out.bs.p;
2378 h->mb.i_last_qp = h->sh.i_qp;
2379 h->mb.i_last_dqp = 0;
2380 h->mb.field_decoding_flag = 0;
2382 i_mb_y = h->sh.i_first_mb / h->mb.i_mb_width;
2383 i_mb_x = h->sh.i_first_mb % h->mb.i_mb_width;
2388 mb_xy = i_mb_x + i_mb_y * h->mb.i_mb_width;
2389 int mb_spos = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
2393 if( x264_bitstream_check_buffer( h ) )
2395 if( !(i_mb_y & SLICE_MBAFF) && h->param.rc.i_vbv_buffer_size )
2396 x264_bitstream_backup( h, &bs_bak[BS_BAK_ROW_VBV], i_skip, 1 );
2397 if( !h->mb.b_reencode_mb )
2398 x264_fdec_filter_row( h, i_mb_y, 0 );
2401 if( !(i_mb_y & SLICE_MBAFF) && back_up_bitstream )
2403 x264_bitstream_backup( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], i_skip, 0 );
2404 if( slice_max_size && (thread_last_mb+1-mb_xy) == h->param.i_slice_min_mbs )
2405 x264_bitstream_backup( h, &bs_bak[BS_BAK_SLICE_MIN_MBS], i_skip, 0 );
2408 if( PARAM_INTERLACED )
2410 if( h->mb.b_adaptive_mbaff )
2414 /* FIXME: VSAD is fast but fairly poor at choosing the best interlace type. */
2415 h->mb.b_interlaced = x264_field_vsad( h, i_mb_x, i_mb_y );
2416 memcpy( &h->zigzagf, MB_INTERLACED ? &h->zigzagf_interlaced : &h->zigzagf_progressive, sizeof(h->zigzagf) );
2417 if( !MB_INTERLACED && (i_mb_y+2) == h->mb.i_mb_height )
2418 x264_expand_border_mbpair( h, i_mb_x, i_mb_y );
2421 h->mb.field[mb_xy] = MB_INTERLACED;
2426 x264_macroblock_cache_load_interlaced( h, i_mb_x, i_mb_y );
2428 x264_macroblock_cache_load_progressive( h, i_mb_x, i_mb_y );
2430 x264_macroblock_analyse( h );
2432 /* encode this macroblock -> be careful it can change the mb type to P_SKIP if needed */
2434 x264_macroblock_encode( h );
2436 if( h->param.b_cabac )
2438 if( mb_xy > h->sh.i_first_mb && !(SLICE_MBAFF && (i_mb_y&1)) )
2439 x264_cabac_encode_terminal( &h->cabac );
2441 if( IS_SKIP( h->mb.i_type ) )
2442 x264_cabac_mb_skip( h, 1 );
2445 if( h->sh.i_type != SLICE_TYPE_I )
2446 x264_cabac_mb_skip( h, 0 );
2447 x264_macroblock_write_cabac( h, &h->cabac );
2452 if( IS_SKIP( h->mb.i_type ) )
2456 if( h->sh.i_type != SLICE_TYPE_I )
2458 bs_write_ue( &h->out.bs, i_skip ); /* skip run */
2461 x264_macroblock_write_cavlc( h );
2462 /* If there was a CAVLC level code overflow, try again at a higher QP. */
2463 if( h->mb.b_overflow )
2465 h->mb.i_chroma_qp = h->chroma_qp_table[++h->mb.i_qp];
2466 h->mb.i_skip_intra = 0;
2467 h->mb.b_skip_mc = 0;
2468 h->mb.b_overflow = 0;
2469 x264_bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], &i_skip, 0 );
2475 int total_bits = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
2476 int mb_size = total_bits - mb_spos;
2478 if( slice_max_size && (!SLICE_MBAFF || (i_mb_y&1)) )
2480 /* Count the skip run, just in case. */
2481 if( !h->param.b_cabac )
2482 total_bits += bs_size_ue_big( i_skip );
2483 /* Check for escape bytes. */
2484 uint8_t *end = h->param.b_cabac ? h->cabac.p : h->out.bs.p;
2485 for( ; last_emu_check < end - 2; last_emu_check++ )
2486 if( last_emu_check[0] == 0 && last_emu_check[1] == 0 && last_emu_check[2] <= 3 )
2488 slice_max_size -= 8;
2491 /* We'll just re-encode this last macroblock if we go over the max slice size. */
2492 if( total_bits - starting_bits > slice_max_size && !h->mb.b_reencode_mb )
2494 if( !x264_frame_new_slice( h, h->fdec ) )
2496 /* Handle the most obnoxious slice-min-mbs edge case: we need to end the slice
2497 * because it's gone over the maximum size, but doing so would violate slice-min-mbs.
2498 * If possible, roll back to the last checkpoint and try again.
2499 * We could try raising QP, but that would break in the case where a slice spans multiple
2500 * rows, which the re-encoding infrastructure can't currently handle. */
2501 if( mb_xy <= thread_last_mb && (thread_last_mb+1-mb_xy) < h->param.i_slice_min_mbs )
2503 if( thread_last_mb-h->param.i_slice_min_mbs < h->sh.i_first_mb+h->param.i_slice_min_mbs )
2505 x264_log( h, X264_LOG_WARNING, "slice-max-size violated (frame %d, cause: slice-min-mbs)\n", h->i_frame );
2509 x264_bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MIN_MBS], &i_skip, 0 );
2510 h->mb.b_reencode_mb = 1;
2511 h->sh.i_last_mb = thread_last_mb-h->param.i_slice_min_mbs;
2514 if( mb_xy-SLICE_MBAFF*h->mb.i_mb_stride != h->sh.i_first_mb )
2516 x264_bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], &i_skip, 0 );
2517 h->mb.b_reencode_mb = 1;
2520 // set to bottom of previous mbpair
2522 h->sh.i_last_mb = mb_xy-1+h->mb.i_mb_stride*(!(i_mb_y&1));
2524 h->sh.i_last_mb = (i_mb_y-2+!(i_mb_y&1))*h->mb.i_mb_stride + h->mb.i_mb_width - 1;
2527 h->sh.i_last_mb = mb_xy-1;
2531 h->sh.i_last_mb = mb_xy;
2538 h->mb.b_reencode_mb = 0;
2541 if( h->param.b_visualize )
2542 x264_visualize_mb( h );
2546 x264_macroblock_cache_save( h );
2548 if( x264_ratecontrol_mb( h, mb_size ) < 0 )
2550 x264_bitstream_restore( h, &bs_bak[BS_BAK_ROW_VBV], &i_skip, 1 );
2551 h->mb.b_reencode_mb = 1;
2553 i_mb_y = i_mb_y - SLICE_MBAFF;
2554 h->mb.i_mb_prev_xy = i_mb_y * h->mb.i_mb_stride - 1;
2555 h->sh.i_last_mb = orig_last_mb;
2559 /* accumulate mb stats */
2560 h->stat.frame.i_mb_count[h->mb.i_type]++;
2562 int b_intra = IS_INTRA( h->mb.i_type );
2563 int b_skip = IS_SKIP( h->mb.i_type );
2564 if( h->param.i_log_level >= X264_LOG_INFO || h->param.rc.b_stat_write )
2566 if( !b_intra && !b_skip && !IS_DIRECT( h->mb.i_type ) )
2568 if( h->mb.i_partition != D_8x8 )
2569 h->stat.frame.i_mb_partition[h->mb.i_partition] += 4;
2571 for( int i = 0; i < 4; i++ )
2572 h->stat.frame.i_mb_partition[h->mb.i_sub_partition[i]] ++;
2573 if( h->param.i_frame_reference > 1 )
2574 for( int i_list = 0; i_list <= (h->sh.i_type == SLICE_TYPE_B); i_list++ )
2575 for( int i = 0; i < 4; i++ )
2577 int i_ref = h->mb.cache.ref[i_list][ x264_scan8[4*i] ];
2579 h->stat.frame.i_mb_count_ref[i_list][i_ref] ++;
2584 if( h->param.i_log_level >= X264_LOG_INFO )
2586 if( h->mb.i_cbp_luma | h->mb.i_cbp_chroma )
2590 for( int i = 0; i < 4; i++ )
2591 if( h->mb.i_cbp_luma & (1 << i) )
2592 for( int p = 0; p < 3; p++ )
2595 int nnz8x8 = M16( &h->mb.cache.non_zero_count[x264_scan8[s8]+0] )
2596 | M16( &h->mb.cache.non_zero_count[x264_scan8[s8]+8] );
2597 h->stat.frame.i_mb_cbp[!b_intra + p*2] += !!nnz8x8;
2602 int cbpsum = (h->mb.i_cbp_luma&1) + ((h->mb.i_cbp_luma>>1)&1)
2603 + ((h->mb.i_cbp_luma>>2)&1) + (h->mb.i_cbp_luma>>3);
2604 h->stat.frame.i_mb_cbp[!b_intra + 0] += cbpsum;
2605 h->stat.frame.i_mb_cbp[!b_intra + 2] += !!h->mb.i_cbp_chroma;
2606 h->stat.frame.i_mb_cbp[!b_intra + 4] += h->mb.i_cbp_chroma >> 1;
2609 if( h->mb.i_cbp_luma && !b_intra )
2611 h->stat.frame.i_mb_count_8x8dct[0] ++;
2612 h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
2614 if( b_intra && h->mb.i_type != I_PCM )
2616 if( h->mb.i_type == I_16x16 )
2617 h->stat.frame.i_mb_pred_mode[0][h->mb.i_intra16x16_pred_mode]++;
2618 else if( h->mb.i_type == I_8x8 )
2619 for( int i = 0; i < 16; i += 4 )
2620 h->stat.frame.i_mb_pred_mode[1][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
2621 else //if( h->mb.i_type == I_4x4 )
2622 for( int i = 0; i < 16; i++ )
2623 h->stat.frame.i_mb_pred_mode[2][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
2624 h->stat.frame.i_mb_pred_mode[3][x264_mb_chroma_pred_mode_fix[h->mb.i_chroma_pred_mode]]++;
2626 h->stat.frame.i_mb_field[b_intra?0:b_skip?2:1] += MB_INTERLACED;
2629 /* calculate deblock strength values (actual deblocking is done per-row along with hpel) */
2631 x264_macroblock_deblock_strength( h );
2633 if( mb_xy == h->sh.i_last_mb )
2638 i_mb_x += i_mb_y & 1;
2639 i_mb_y ^= i_mb_x < h->mb.i_mb_width;
2643 if( i_mb_x == h->mb.i_mb_width )
2649 if( h->sh.i_last_mb < h->sh.i_first_mb )
2652 h->out.nal[h->out.i_nal].i_last_mb = h->sh.i_last_mb;
2654 if( h->param.b_cabac )
2656 x264_cabac_encode_flush( h, &h->cabac );
2657 h->out.bs.p = h->cabac.p;
2662 bs_write_ue( &h->out.bs, i_skip ); /* last skip run */
2663 /* rbsp_slice_trailing_bits */
2664 bs_rbsp_trailing( &h->out.bs );
2665 bs_flush( &h->out.bs );
2667 if( x264_nal_end( h ) )
2670 if( h->sh.i_last_mb == (h->i_threadslice_end * h->mb.i_mb_width - 1) )
2672 h->stat.frame.i_misc_bits = bs_pos( &h->out.bs )
2673 + (h->out.i_nal*NALU_OVERHEAD * 8)
2674 - h->stat.frame.i_tex_bits
2675 - h->stat.frame.i_mv_bits;
2676 x264_fdec_filter_row( h, h->i_threadslice_end, 0 );
2678 if( h->param.b_sliced_threads )
2680 /* Tell the main thread we're done. */
2681 x264_threadslice_cond_broadcast( h, 1 );
2683 for( int mb_y = h->i_threadslice_start; mb_y <= h->i_threadslice_end; mb_y++ )
2684 x264_fdec_filter_row( h, mb_y, 1 );
2685 x264_threadslice_cond_broadcast( h, 2 );
2686 /* Do the first row of hpel, now that the previous slice is done */
2687 if( h->i_thread_idx > 0 )
2689 x264_threadslice_cond_wait( h->thread[h->i_thread_idx-1], 2 );
2690 x264_fdec_filter_row( h, h->i_threadslice_start + (1 << SLICE_MBAFF), 2 );
2694 /* Free mb info after the last thread's done using it */
2695 if( h->fdec->mb_info_free && (!h->param.b_sliced_threads || h->i_thread_idx == (h->param.i_threads-1)) )
2697 h->fdec->mb_info_free( h->fdec->mb_info );
2698 h->fdec->mb_info = NULL;
2699 h->fdec->mb_info_free = NULL;
2706 static void x264_thread_sync_context( x264_t *dst, x264_t *src )
2711 // reference counting
2712 for( x264_frame_t **f = src->frames.reference; *f; f++ )
2713 (*f)->i_reference_count++;
2714 for( x264_frame_t **f = dst->frames.reference; *f; f++ )
2715 x264_frame_push_unused( src, *f );
2716 src->fdec->i_reference_count++;
2717 x264_frame_push_unused( src, dst->fdec );
2719 // copy everything except the per-thread pointers and the constants.
2720 memcpy( &dst->i_frame, &src->i_frame, offsetof(x264_t, mb.base) - offsetof(x264_t, i_frame) );
2721 dst->param = src->param;
2722 dst->stat = src->stat;
2723 dst->pixf = src->pixf;
2726 static void x264_thread_sync_stat( x264_t *dst, x264_t *src )
2730 memcpy( &dst->stat.i_frame_count, &src->stat.i_frame_count, sizeof(dst->stat) - sizeof(dst->stat.frame) );
2733 static void *x264_slices_write( x264_t *h )
2735 int i_slice_num = 0;
2736 int last_thread_mb = h->sh.i_last_mb;
2739 if( h->param.b_visualize )
2740 if( x264_visualize_init( h ) )
2745 memset( &h->stat.frame, 0, sizeof(h->stat.frame) );
2746 h->mb.b_reencode_mb = 0;
2747 while( h->sh.i_first_mb + SLICE_MBAFF*h->mb.i_mb_stride <= last_thread_mb )
2749 h->sh.i_last_mb = last_thread_mb;
2750 if( !i_slice_num || !x264_frame_new_slice( h, h->fdec ) )
2752 if( h->param.i_slice_max_mbs )
2756 // convert first to mbaff form, add slice-max-mbs, then convert back to normal form
2757 int last_mbaff = 2*(h->sh.i_first_mb % h->mb.i_mb_width)
2758 + h->mb.i_mb_width*(h->sh.i_first_mb / h->mb.i_mb_width)
2759 + h->param.i_slice_max_mbs - 1;
2760 int last_x = (last_mbaff % (2*h->mb.i_mb_width))/2;
2761 int last_y = (last_mbaff / (2*h->mb.i_mb_width))*2 + 1;
2762 h->sh.i_last_mb = last_x + h->mb.i_mb_stride*last_y;
2766 h->sh.i_last_mb = h->sh.i_first_mb + h->param.i_slice_max_mbs - 1;
2767 if( h->sh.i_last_mb < last_thread_mb && last_thread_mb - h->sh.i_last_mb < h->param.i_slice_min_mbs )
2768 h->sh.i_last_mb = last_thread_mb - h->param.i_slice_min_mbs;
2772 else if( h->param.i_slice_count && !h->param.b_sliced_threads )
2774 int height = h->mb.i_mb_height >> PARAM_INTERLACED;
2775 int width = h->mb.i_mb_width << PARAM_INTERLACED;
2777 h->sh.i_last_mb = (height * i_slice_num + h->param.i_slice_count/2) / h->param.i_slice_count * width - 1;
2780 h->sh.i_last_mb = X264_MIN( h->sh.i_last_mb, last_thread_mb );
2781 if( x264_stack_align( x264_slice_write, h ) )
2783 h->sh.i_first_mb = h->sh.i_last_mb + 1;
2784 // if i_first_mb is not the last mb in a row then go to the next mb in MBAFF order
2785 if( SLICE_MBAFF && h->sh.i_first_mb % h->mb.i_mb_width )
2786 h->sh.i_first_mb -= h->mb.i_mb_stride;
2790 if( h->param.b_visualize )
2792 x264_visualize_show( h );
2793 x264_visualize_close( h );
2800 /* Tell other threads we're done, so they wouldn't wait for it */
2801 if( h->param.b_sliced_threads )
2802 x264_threadslice_cond_broadcast( h, 2 );
2806 static int x264_threaded_slices_write( x264_t *h )
2808 /* set first/last mb and sync contexts */
2809 for( int i = 0; i < h->param.i_threads; i++ )
2811 x264_t *t = h->thread[i];
2814 t->param = h->param;
2815 memcpy( &t->i_frame, &h->i_frame, offsetof(x264_t, rc) - offsetof(x264_t, i_frame) );
2817 int height = h->mb.i_mb_height >> PARAM_INTERLACED;
2818 t->i_threadslice_start = ((height * i + h->param.i_slice_count/2) / h->param.i_threads) << PARAM_INTERLACED;
2819 t->i_threadslice_end = ((height * (i+1) + h->param.i_slice_count/2) / h->param.i_threads) << PARAM_INTERLACED;
2820 t->sh.i_first_mb = t->i_threadslice_start * h->mb.i_mb_width;
2821 t->sh.i_last_mb = t->i_threadslice_end * h->mb.i_mb_width - 1;
2824 x264_stack_align( x264_analyse_weight_frame, h, h->mb.i_mb_height*16 + 16 );
2826 x264_threads_distribute_ratecontrol( h );
2829 for( int i = 0; i < h->param.i_threads; i++ )
2831 h->thread[i]->i_thread_idx = i;
2832 h->thread[i]->b_thread_active = 1;
2833 x264_threadslice_cond_broadcast( h->thread[i], 0 );
2836 for( int i = 0; i < h->param.i_threads; i++ )
2837 x264_threadpool_run( h->threadpool, (void*)x264_slices_write, h->thread[i] );
2839 for( int i = 0; i < h->param.i_threads; i++ )
2840 x264_threadslice_cond_wait( h->thread[i], 1 );
2842 x264_threads_merge_ratecontrol( h );
2844 for( int i = 1; i < h->param.i_threads; i++ )
2846 x264_t *t = h->thread[i];
2847 for( int j = 0; j < t->out.i_nal; j++ )
2849 h->out.nal[h->out.i_nal] = t->out.nal[j];
2851 x264_nal_check_buffer( h );
2853 /* All entries in stat.frame are ints except for ssd/ssim. */
2854 for( int j = 0; j < (offsetof(x264_t,stat.frame.i_ssd) - offsetof(x264_t,stat.frame.i_mv_bits)) / sizeof(int); j++ )
2855 ((int*)&h->stat.frame)[j] += ((int*)&t->stat.frame)[j];
2856 for( int j = 0; j < 3; j++ )
2857 h->stat.frame.i_ssd[j] += t->stat.frame.i_ssd[j];
2858 h->stat.frame.f_ssim += t->stat.frame.f_ssim;
2859 h->stat.frame.i_ssim_cnt += t->stat.frame.i_ssim_cnt;
2865 void x264_encoder_intra_refresh( x264_t *h )
2867 h = h->thread[h->i_thread_phase];
2868 h->b_queued_intra_refresh = 1;
2871 int x264_encoder_invalidate_reference( x264_t *h, int64_t pts )
2873 if( h->param.i_bframe )
2875 x264_log( h, X264_LOG_ERROR, "x264_encoder_invalidate_reference is not supported with B-frames enabled\n" );
2878 if( h->param.b_intra_refresh )
2880 x264_log( h, X264_LOG_ERROR, "x264_encoder_invalidate_reference is not supported with intra refresh enabled\n" );
2883 h = h->thread[h->i_thread_phase];
2884 if( pts >= h->i_last_idr_pts )
2886 for( int i = 0; h->frames.reference[i]; i++ )
2887 if( pts <= h->frames.reference[i]->i_pts )
2888 h->frames.reference[i]->b_corrupt = 1;
2889 if( pts <= h->fdec->i_pts )
2890 h->fdec->b_corrupt = 1;
2895 /****************************************************************************
2896 * x264_encoder_encode:
2897 * XXX: i_poc : is the poc of the current given picture
2898 * i_frame : is the number of the frame being coded
2899 * ex: type frame poc
2907 ****************************************************************************/
2908 int x264_encoder_encode( x264_t *h,
2909 x264_nal_t **pp_nal, int *pi_nal,
2910 x264_picture_t *pic_in,
2911 x264_picture_t *pic_out )
2913 x264_t *thread_current, *thread_prev, *thread_oldest;
2914 int i_nal_type, i_nal_ref_idc, i_global_qp;
2915 int overhead = NALU_OVERHEAD;
2918 if( h->opencl.b_fatal_error )
2922 if( h->i_thread_frames > 1 )
2924 thread_prev = h->thread[ h->i_thread_phase ];
2925 h->i_thread_phase = (h->i_thread_phase + 1) % h->i_thread_frames;
2926 thread_current = h->thread[ h->i_thread_phase ];
2927 thread_oldest = h->thread[ (h->i_thread_phase + 1) % h->i_thread_frames ];
2928 x264_thread_sync_context( thread_current, thread_prev );
2929 x264_thread_sync_ratecontrol( thread_current, thread_prev, thread_oldest );
2937 h->i_cpb_delay_pir_offset = h->i_cpb_delay_pir_offset_next;
2943 /* ------------------- Setup new frame from picture -------------------- */
2944 if( pic_in != NULL )
2946 /* 1: Copy the picture to a frame and move it to a buffer */
2947 x264_frame_t *fenc = x264_frame_pop_unused( h, 0 );
2951 if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 )
2954 if( h->param.i_width != 16 * h->mb.i_mb_width ||
2955 h->param.i_height != 16 * h->mb.i_mb_height )
2956 x264_frame_expand_border_mod16( h, fenc );
2958 fenc->i_frame = h->frames.i_input++;
2960 if( fenc->i_frame == 0 )
2961 h->frames.i_first_pts = fenc->i_pts;
2962 if( h->frames.i_bframe_delay && fenc->i_frame == h->frames.i_bframe_delay )
2963 h->frames.i_bframe_delay_time = fenc->i_pts - h->frames.i_first_pts;
2965 if( h->param.b_vfr_input && fenc->i_pts <= h->frames.i_largest_pts )
2966 x264_log( h, X264_LOG_WARNING, "non-strictly-monotonic PTS\n" );
2968 h->frames.i_second_largest_pts = h->frames.i_largest_pts;
2969 h->frames.i_largest_pts = fenc->i_pts;
2971 if( (fenc->i_pic_struct < PIC_STRUCT_AUTO) || (fenc->i_pic_struct > PIC_STRUCT_TRIPLE) )
2972 fenc->i_pic_struct = PIC_STRUCT_AUTO;
2974 if( fenc->i_pic_struct == PIC_STRUCT_AUTO )
2977 int b_interlaced = fenc->param ? fenc->param->b_interlaced : h->param.b_interlaced;
2979 int b_interlaced = 0;
2983 int b_tff = fenc->param ? fenc->param->b_tff : h->param.b_tff;
2984 fenc->i_pic_struct = b_tff ? PIC_STRUCT_TOP_BOTTOM : PIC_STRUCT_BOTTOM_TOP;
2987 fenc->i_pic_struct = PIC_STRUCT_PROGRESSIVE;
2990 if( h->param.rc.b_mb_tree && h->param.rc.b_stat_read )
2992 if( x264_macroblock_tree_read( h, fenc, pic_in->prop.quant_offsets ) )
2996 x264_stack_align( x264_adaptive_quant_frame, h, fenc, pic_in->prop.quant_offsets );
2998 if( pic_in->prop.quant_offsets_free )
2999 pic_in->prop.quant_offsets_free( pic_in->prop.quant_offsets );
3001 if( h->frames.b_have_lowres )
3002 x264_frame_init_lowres( h, fenc );
3004 /* 2: Place the frame into the queue for its slice type decision */
3005 x264_lookahead_put_frame( h, fenc );
3007 if( h->frames.i_input <= h->frames.i_delay + 1 - h->i_thread_frames )
3009 /* Nothing yet to encode, waiting for filling of buffers */
3010 pic_out->i_type = X264_TYPE_AUTO;
3016 /* signal kills for lookahead thread */
3017 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
3018 h->lookahead->b_exit_thread = 1;
3019 x264_pthread_cond_broadcast( &h->lookahead->ifbuf.cv_fill );
3020 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
3024 /* 3: The picture is analyzed in the lookahead */
3025 if( !h->frames.current[0] )
3026 x264_lookahead_get_frames( h );
3028 if( !h->frames.current[0] && x264_lookahead_is_empty( h ) )
3029 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
3031 /* ------------------- Get frame to be encoded ------------------------- */
3032 /* 4: get picture to encode */
3033 h->fenc = x264_frame_shift( h->frames.current );
3035 /* If applicable, wait for previous frame reconstruction to finish */
3036 if( h->param.b_sliced_threads )
3037 if( x264_threadpool_wait_all( h ) < 0 )
3040 if( h->i_frame == h->i_thread_frames - 1 )
3041 h->i_reordered_pts_delay = h->fenc->i_reordered_pts;
3042 if( h->fenc->param )
3044 x264_encoder_reconfig( h, h->fenc->param );
3045 if( h->fenc->param->param_free )
3047 h->fenc->param->param_free( h->fenc->param );
3048 h->fenc->param = NULL;
3052 // ok to call this before encoding any frames, since the initial values of fdec have b_kept_as_ref=0
3053 if( x264_reference_update( h ) )
3055 h->fdec->i_lines_completed = -1;
3057 if( !IS_X264_TYPE_I( h->fenc->i_type ) )
3059 int valid_refs_left = 0;
3060 for( int i = 0; h->frames.reference[i]; i++ )
3061 if( !h->frames.reference[i]->b_corrupt )
3063 /* No valid reference frames left: force an IDR. */
3064 if( !valid_refs_left )
3066 h->fenc->b_keyframe = 1;
3067 h->fenc->i_type = X264_TYPE_IDR;
3071 if( h->fenc->b_keyframe )
3073 h->frames.i_last_keyframe = h->fenc->i_frame;
3074 if( h->fenc->i_type == X264_TYPE_IDR )
3077 h->frames.i_last_idr = h->fenc->i_frame;
3080 h->sh.i_mmco_command_count =
3081 h->sh.i_mmco_remove_from_end = 0;
3082 h->b_ref_reorder[0] =
3083 h->b_ref_reorder[1] = 0;
3085 h->fenc->i_poc = 2 * ( h->fenc->i_frame - X264_MAX( h->frames.i_last_idr, 0 ) );
3087 /* ------------------- Setup frame context ----------------------------- */
3088 /* 5: Init data dependent of frame type */
3089 if( h->fenc->i_type == X264_TYPE_IDR )
3091 /* reset ref pictures */
3092 i_nal_type = NAL_SLICE_IDR;
3093 i_nal_ref_idc = NAL_PRIORITY_HIGHEST;
3094 h->sh.i_type = SLICE_TYPE_I;
3095 x264_reference_reset( h );
3096 h->frames.i_poc_last_open_gop = -1;
3098 else if( h->fenc->i_type == X264_TYPE_I )
3100 i_nal_type = NAL_SLICE;
3101 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
3102 h->sh.i_type = SLICE_TYPE_I;
3103 x264_reference_hierarchy_reset( h );
3104 if( h->param.b_open_gop )
3105 h->frames.i_poc_last_open_gop = h->fenc->b_keyframe ? h->fenc->i_poc : -1;
3107 else if( h->fenc->i_type == X264_TYPE_P )
3109 i_nal_type = NAL_SLICE;
3110 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
3111 h->sh.i_type = SLICE_TYPE_P;
3112 x264_reference_hierarchy_reset( h );
3113 h->frames.i_poc_last_open_gop = -1;
3115 else if( h->fenc->i_type == X264_TYPE_BREF )
3117 i_nal_type = NAL_SLICE;
3118 i_nal_ref_idc = h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT ? NAL_PRIORITY_LOW : NAL_PRIORITY_HIGH;
3119 h->sh.i_type = SLICE_TYPE_B;
3120 x264_reference_hierarchy_reset( h );
3124 i_nal_type = NAL_SLICE;
3125 i_nal_ref_idc = NAL_PRIORITY_DISPOSABLE;
3126 h->sh.i_type = SLICE_TYPE_B;
3129 h->fdec->i_type = h->fenc->i_type;
3130 h->fdec->i_frame = h->fenc->i_frame;
3131 h->fenc->b_kept_as_ref =
3132 h->fdec->b_kept_as_ref = i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE && h->param.i_keyint_max > 1;
3134 h->fdec->mb_info = h->fenc->mb_info;
3135 h->fdec->mb_info_free = h->fenc->mb_info_free;
3136 h->fenc->mb_info = NULL;
3137 h->fenc->mb_info_free = NULL;
3139 h->fdec->i_pts = h->fenc->i_pts;
3140 if( h->frames.i_bframe_delay )
3142 int64_t *prev_reordered_pts = thread_current->frames.i_prev_reordered_pts;
3143 h->fdec->i_dts = h->i_frame > h->frames.i_bframe_delay
3144 ? prev_reordered_pts[ (h->i_frame - h->frames.i_bframe_delay) % h->frames.i_bframe_delay ]
3145 : h->fenc->i_reordered_pts - h->frames.i_bframe_delay_time;
3146 prev_reordered_pts[ h->i_frame % h->frames.i_bframe_delay ] = h->fenc->i_reordered_pts;
3149 h->fdec->i_dts = h->fenc->i_reordered_pts;
3150 if( h->fenc->i_type == X264_TYPE_IDR )
3151 h->i_last_idr_pts = h->fdec->i_pts;
3153 /* ------------------- Init ----------------------------- */
3154 /* build ref list 0/1 */
3155 x264_reference_build_list( h, h->fdec->i_poc );
3157 /* ---------------------- Write the bitstream -------------------------- */
3158 /* Init bitstream context */
3159 if( h->param.b_sliced_threads )
3161 for( int i = 0; i < h->param.i_threads; i++ )
3163 bs_init( &h->thread[i]->out.bs, h->thread[i]->out.p_bitstream, h->thread[i]->out.i_bitstream );
3164 h->thread[i]->out.i_nal = 0;
3169 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
3173 if( h->param.b_aud )
3177 if( h->sh.i_type == SLICE_TYPE_I )
3179 else if( h->sh.i_type == SLICE_TYPE_P )
3181 else if( h->sh.i_type == SLICE_TYPE_B )
3186 x264_nal_start( h, NAL_AUD, NAL_PRIORITY_DISPOSABLE );
3187 bs_write( &h->out.bs, 3, pic_type );
3188 bs_rbsp_trailing( &h->out.bs );
3189 if( x264_nal_end( h ) )
3191 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3194 h->i_nal_type = i_nal_type;
3195 h->i_nal_ref_idc = i_nal_ref_idc;
3197 if( h->param.b_intra_refresh )
3199 if( IS_X264_TYPE_I( h->fenc->i_type ) )
3201 h->fdec->i_frames_since_pir = 0;
3202 h->b_queued_intra_refresh = 0;
3203 /* PIR is currently only supported with ref == 1, so any intra frame effectively refreshes
3204 * the whole frame and counts as an intra refresh. */
3205 h->fdec->f_pir_position = h->mb.i_mb_width;
3207 else if( h->fenc->i_type == X264_TYPE_P )
3209 int pocdiff = (h->fdec->i_poc - h->fref[0][0]->i_poc)/2;
3210 float increment = X264_MAX( ((float)h->mb.i_mb_width-1) / h->param.i_keyint_max, 1 );
3211 h->fdec->f_pir_position = h->fref[0][0]->f_pir_position;
3212 h->fdec->i_frames_since_pir = h->fref[0][0]->i_frames_since_pir + pocdiff;
3213 if( h->fdec->i_frames_since_pir >= h->param.i_keyint_max ||
3214 (h->b_queued_intra_refresh && h->fdec->f_pir_position + 0.5 >= h->mb.i_mb_width) )
3216 h->fdec->f_pir_position = 0;
3217 h->fdec->i_frames_since_pir = 0;
3218 h->b_queued_intra_refresh = 0;
3219 h->fenc->b_keyframe = 1;
3221 h->fdec->i_pir_start_col = h->fdec->f_pir_position+0.5;
3222 h->fdec->f_pir_position += increment * pocdiff;
3223 h->fdec->i_pir_end_col = h->fdec->f_pir_position+0.5;
3224 /* If our intra refresh has reached the right side of the frame, we're done. */
3225 if( h->fdec->i_pir_end_col >= h->mb.i_mb_width - 1 )
3227 h->fdec->f_pir_position = h->mb.i_mb_width;
3228 h->fdec->i_pir_end_col = h->mb.i_mb_width - 1;
3233 if( h->fenc->b_keyframe )
3235 /* Write SPS and PPS */
3236 if( h->param.b_repeat_headers )
3238 /* generate sequence parameters */
3239 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
3240 x264_sps_write( &h->out.bs, h->sps );
3241 if( x264_nal_end( h ) )
3243 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3245 /* generate picture parameters */
3246 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
3247 x264_pps_write( &h->out.bs, h->sps, h->pps );
3248 if( x264_nal_end( h ) )
3250 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3253 /* when frame threading is used, buffering period sei is written in x264_encoder_frame_end */
3254 if( h->i_thread_frames == 1 && h->sps->vui.b_nal_hrd_parameters_present )
3256 x264_hrd_fullness( h );
3257 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3258 x264_sei_buffering_period_write( h, &h->out.bs );
3259 if( x264_nal_end( h ) )
3261 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3265 /* write extra sei */
3266 for( int i = 0; i < h->fenc->extra_sei.num_payloads; i++ )
3268 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3269 x264_sei_write( &h->out.bs, h->fenc->extra_sei.payloads[i].payload, h->fenc->extra_sei.payloads[i].payload_size,
3270 h->fenc->extra_sei.payloads[i].payload_type );
3271 if( x264_nal_end( h ) )
3273 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3274 if( h->fenc->extra_sei.sei_free )
3276 h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads[i].payload );
3277 h->fenc->extra_sei.payloads[i].payload = NULL;
3281 if( h->fenc->extra_sei.sei_free )
3283 h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads );
3284 h->fenc->extra_sei.payloads = NULL;
3285 h->fenc->extra_sei.sei_free = NULL;
3288 if( h->fenc->b_keyframe )
3290 if( h->param.b_repeat_headers && h->fenc->i_frame == 0 )
3292 /* identify ourself */
3293 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3294 if( x264_sei_version_write( h, &h->out.bs ) )
3296 if( x264_nal_end( h ) )
3298 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3301 if( h->fenc->i_type != X264_TYPE_IDR )
3303 int time_to_recovery = h->param.b_open_gop ? 0 : X264_MIN( h->mb.i_mb_width - 1, h->param.i_keyint_max ) + h->param.i_bframe - 1;
3304 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3305 x264_sei_recovery_point_write( h, &h->out.bs, time_to_recovery );
3306 if( x264_nal_end( h ) )
3308 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3311 if ( h->param.i_frame_packing >= 0 )
3313 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3314 x264_sei_frame_packing_write( h, &h->out.bs );
3315 if( x264_nal_end( h ) )
3317 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3321 /* generate sei pic timing */
3322 if( h->sps->vui.b_pic_struct_present || h->sps->vui.b_nal_hrd_parameters_present )
3324 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3325 x264_sei_pic_timing_write( h, &h->out.bs );
3326 if( x264_nal_end( h ) )
3328 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3331 /* As required by Blu-ray. */
3332 if( !IS_X264_TYPE_B( h->fenc->i_type ) && h->b_sh_backup )
3335 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3336 x264_sei_dec_ref_pic_marking_write( h, &h->out.bs );
3337 if( x264_nal_end( h ) )
3339 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3342 if( h->fenc->b_keyframe && h->param.b_intra_refresh )
3343 h->i_cpb_delay_pir_offset_next = h->fenc->i_cpb_delay;
3345 /* Init the rate control */
3346 /* FIXME: Include slice header bit cost. */
3347 x264_ratecontrol_start( h, h->fenc->i_qpplus1, overhead*8 );
3348 i_global_qp = x264_ratecontrol_qp( h );
3350 pic_out->i_qpplus1 =
3351 h->fdec->i_qpplus1 = i_global_qp + 1;
3353 if( h->param.rc.b_stat_read && h->sh.i_type != SLICE_TYPE_I )
3355 x264_reference_build_list_optimal( h );
3356 x264_reference_check_reorder( h );
3360 h->fdec->i_poc_l0ref0 = h->fref[0][0]->i_poc;
3362 /* ------------------------ Create slice header ----------------------- */
3363 x264_slice_init( h, i_nal_type, i_global_qp );
3365 /*------------------------- Weights -------------------------------------*/
3366 if( h->sh.i_type == SLICE_TYPE_B )
3367 x264_macroblock_bipred_init( h );
3369 x264_weighted_pred_init( h );
3371 if( i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE )
3375 h->i_threadslice_start = 0;
3376 h->i_threadslice_end = h->mb.i_mb_height;
3377 if( h->i_thread_frames > 1 )
3379 x264_threadpool_run( h->threadpool, (void*)x264_slices_write, h );
3380 h->b_thread_active = 1;
3382 else if( h->param.b_sliced_threads )
3384 if( x264_threaded_slices_write( h ) )
3388 if( (intptr_t)x264_slices_write( h ) )
3391 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
3394 static int x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
3395 x264_nal_t **pp_nal, int *pi_nal,
3396 x264_picture_t *pic_out )
3398 char psz_message[80];
3400 if( !h->param.b_sliced_threads && h->b_thread_active )
3402 h->b_thread_active = 0;
3403 if( (intptr_t)x264_threadpool_wait( h->threadpool, h ) )
3408 pic_out->i_type = X264_TYPE_AUTO;
3414 /* generate buffering period sei and insert it into place */
3415 if( h->i_thread_frames > 1 && h->fenc->b_keyframe && h->sps->vui.b_nal_hrd_parameters_present )
3417 x264_hrd_fullness( h );
3418 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3419 x264_sei_buffering_period_write( h, &h->out.bs );
3420 if( x264_nal_end( h ) )
3422 /* buffering period sei must follow AUD, SPS and PPS and precede all other SEIs */
3424 while( h->out.nal[idx].i_type == NAL_AUD ||
3425 h->out.nal[idx].i_type == NAL_SPS ||
3426 h->out.nal[idx].i_type == NAL_PPS )
3428 x264_nal_t nal_tmp = h->out.nal[h->out.i_nal-1];
3429 memmove( &h->out.nal[idx+1], &h->out.nal[idx], (h->out.i_nal-idx-1)*sizeof(x264_nal_t) );
3430 h->out.nal[idx] = nal_tmp;
3433 int frame_size = x264_encoder_encapsulate_nals( h, 0 );
3434 if( frame_size < 0 )
3437 /* Set output picture properties */
3438 pic_out->i_type = h->fenc->i_type;
3440 pic_out->b_keyframe = h->fenc->b_keyframe;
3441 pic_out->i_pic_struct = h->fenc->i_pic_struct;
3443 pic_out->i_pts = h->fdec->i_pts;
3444 pic_out->i_dts = h->fdec->i_dts;
3446 if( pic_out->i_pts < pic_out->i_dts )
3447 x264_log( h, X264_LOG_WARNING, "invalid DTS: PTS is less than DTS\n" );
3449 pic_out->opaque = h->fenc->opaque;
3451 pic_out->img.i_csp = h->fdec->i_csp;
3453 pic_out->img.i_csp |= X264_CSP_HIGH_DEPTH;
3455 pic_out->img.i_plane = h->fdec->i_plane;
3456 for( int i = 0; i < pic_out->img.i_plane; i++ )
3458 pic_out->img.i_stride[i] = h->fdec->i_stride[i] * sizeof(pixel);
3459 pic_out->img.plane[i] = (uint8_t*)h->fdec->plane[i];
3462 x264_frame_push_unused( thread_current, h->fenc );
3464 /* ---------------------- Update encoder state ------------------------- */
3468 if( x264_ratecontrol_end( h, frame_size * 8, &filler ) < 0 )
3471 pic_out->hrd_timing = h->fenc->hrd_timing;
3472 pic_out->prop.f_crf_avg = h->fdec->f_crf_avg;
3477 overhead = (FILLER_OVERHEAD - h->param.b_annexb);
3478 if( h->param.i_slice_max_size && filler > h->param.i_slice_max_size )
3480 int next_size = filler - h->param.i_slice_max_size;
3481 int overflow = X264_MAX( overhead - next_size, 0 );
3482 f = h->param.i_slice_max_size - overhead - overflow;
3485 f = X264_MAX( 0, filler - overhead );
3487 if( x264_bitstream_check_buffer_filler( h, f ) )
3489 x264_nal_start( h, NAL_FILLER, NAL_PRIORITY_DISPOSABLE );
3490 x264_filler_write( h, &h->out.bs, f );
3491 if( x264_nal_end( h ) )
3493 int total_size = x264_encoder_encapsulate_nals( h, h->out.i_nal-1 );
3494 if( total_size < 0 )
3496 frame_size += total_size;
3497 filler -= total_size;
3500 /* End bitstream, set output */
3501 *pi_nal = h->out.i_nal;
3502 *pp_nal = h->out.nal;
3506 x264_noise_reduction_update( h );
3508 /* ---------------------- Compute/Print statistics --------------------- */
3509 x264_thread_sync_stat( h, h->thread[0] );
3512 h->stat.i_frame_count[h->sh.i_type]++;
3513 h->stat.i_frame_size[h->sh.i_type] += frame_size;
3514 h->stat.f_frame_qp[h->sh.i_type] += h->fdec->f_qp_avg_aq;
3516 for( int i = 0; i < X264_MBTYPE_MAX; i++ )
3517 h->stat.i_mb_count[h->sh.i_type][i] += h->stat.frame.i_mb_count[i];
3518 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
3519 h->stat.i_mb_partition[h->sh.i_type][i] += h->stat.frame.i_mb_partition[i];
3520 for( int i = 0; i < 2; i++ )
3521 h->stat.i_mb_count_8x8dct[i] += h->stat.frame.i_mb_count_8x8dct[i];
3522 for( int i = 0; i < 6; i++ )
3523 h->stat.i_mb_cbp[i] += h->stat.frame.i_mb_cbp[i];
3524 for( int i = 0; i < 4; i++ )
3525 for( int j = 0; j < 13; j++ )
3526 h->stat.i_mb_pred_mode[i][j] += h->stat.frame.i_mb_pred_mode[i][j];
3527 if( h->sh.i_type != SLICE_TYPE_I )
3528 for( int i_list = 0; i_list < 2; i_list++ )
3529 for( int i = 0; i < X264_REF_MAX*2; i++ )
3530 h->stat.i_mb_count_ref[h->sh.i_type][i_list][i] += h->stat.frame.i_mb_count_ref[i_list][i];
3531 for( int i = 0; i < 3; i++ )
3532 h->stat.i_mb_field[i] += h->stat.frame.i_mb_field[i];
3533 if( h->sh.i_type == SLICE_TYPE_P && h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE )
3535 h->stat.i_wpred[0] += !!h->sh.weight[0][0].weightfn;
3536 h->stat.i_wpred[1] += !!h->sh.weight[0][1].weightfn || !!h->sh.weight[0][2].weightfn;
3538 if( h->sh.i_type == SLICE_TYPE_B )
3540 h->stat.i_direct_frames[ h->sh.b_direct_spatial_mv_pred ] ++;
3541 if( h->mb.b_direct_auto_write )
3543 //FIXME somewhat arbitrary time constants
3544 if( h->stat.i_direct_score[0] + h->stat.i_direct_score[1] > h->mb.i_mb_count )
3545 for( int i = 0; i < 2; i++ )
3546 h->stat.i_direct_score[i] = h->stat.i_direct_score[i] * 9/10;
3547 for( int i = 0; i < 2; i++ )
3548 h->stat.i_direct_score[i] += h->stat.frame.i_direct_score[i];
3552 h->stat.i_consecutive_bframes[h->fenc->i_bframes]++;
3554 psz_message[0] = '\0';
3555 double dur = h->fenc->f_duration;
3556 h->stat.f_frame_duration[h->sh.i_type] += dur;
3557 if( h->param.analyse.b_psnr )
3561 h->stat.frame.i_ssd[0],
3562 h->stat.frame.i_ssd[1],
3563 h->stat.frame.i_ssd[2],
3565 int luma_size = h->param.i_width * h->param.i_height;
3566 int chroma_size = CHROMA_SIZE( luma_size );
3567 pic_out->prop.f_psnr[0] = x264_psnr( ssd[0], luma_size );
3568 pic_out->prop.f_psnr[1] = x264_psnr( ssd[1], chroma_size );
3569 pic_out->prop.f_psnr[2] = x264_psnr( ssd[2], chroma_size );
3570 pic_out->prop.f_psnr_avg = x264_psnr( ssd[0] + ssd[1] + ssd[2], luma_size + chroma_size*2 );
3572 h->stat.f_ssd_global[h->sh.i_type] += dur * (ssd[0] + ssd[1] + ssd[2]);
3573 h->stat.f_psnr_average[h->sh.i_type] += dur * pic_out->prop.f_psnr_avg;
3574 h->stat.f_psnr_mean_y[h->sh.i_type] += dur * pic_out->prop.f_psnr[0];
3575 h->stat.f_psnr_mean_u[h->sh.i_type] += dur * pic_out->prop.f_psnr[1];
3576 h->stat.f_psnr_mean_v[h->sh.i_type] += dur * pic_out->prop.f_psnr[2];
3578 snprintf( psz_message, 80, " PSNR Y:%5.2f U:%5.2f V:%5.2f", pic_out->prop.f_psnr[0],
3579 pic_out->prop.f_psnr[1],
3580 pic_out->prop.f_psnr[2] );
3583 if( h->param.analyse.b_ssim )
3585 pic_out->prop.f_ssim = h->stat.frame.f_ssim / h->stat.frame.i_ssim_cnt;
3586 h->stat.f_ssim_mean_y[h->sh.i_type] += pic_out->prop.f_ssim * dur;
3587 snprintf( psz_message + strlen(psz_message), 80 - strlen(psz_message),
3588 " SSIM Y:%.5f", pic_out->prop.f_ssim );
3590 psz_message[79] = '\0';
3592 x264_log( h, X264_LOG_DEBUG,
3593 "frame=%4d QP=%.2f NAL=%d Slice:%c Poc:%-3d I:%-4d P:%-4d SKIP:%-4d size=%d bytes%s\n",
3595 h->fdec->f_qp_avg_aq,
3597 h->sh.i_type == SLICE_TYPE_I ? 'I' : (h->sh.i_type == SLICE_TYPE_P ? 'P' : 'B' ),
3599 h->stat.frame.i_mb_count_i,
3600 h->stat.frame.i_mb_count_p,
3601 h->stat.frame.i_mb_count_skip,
3605 // keep stats all in one place
3606 x264_thread_sync_stat( h->thread[0], h );
3607 // for the use of the next frame
3608 x264_thread_sync_stat( thread_current, h );
3610 #ifdef DEBUG_MB_TYPE
3612 static const char mb_chars[] = { 'i', 'i', 'I', 'C', 'P', '8', 'S',
3613 'D', '<', 'X', 'B', 'X', '>', 'B', 'B', 'B', 'B', '8', 'S' };
3614 for( int mb_xy = 0; mb_xy < h->mb.i_mb_width * h->mb.i_mb_height; mb_xy++ )
3616 if( h->mb.type[mb_xy] < X264_MBTYPE_MAX && h->mb.type[mb_xy] >= 0 )
3617 fprintf( stderr, "%c ", mb_chars[ h->mb.type[mb_xy] ] );
3619 fprintf( stderr, "? " );
3621 if( (mb_xy+1) % h->mb.i_mb_width == 0 )
3622 fprintf( stderr, "\n" );
3627 /* Remove duplicates, must be done near the end as breaks h->fref0 array
3628 * by freeing some of its pointers. */
3629 for( int i = 0; i < h->i_ref[0]; i++ )
3630 if( h->fref[0][i] && h->fref[0][i]->b_duplicate )
3632 x264_frame_push_blank_unused( h, h->fref[0][i] );
3636 if( h->param.psz_dump_yuv )
3637 x264_frame_dump( h );
3643 static void x264_print_intra( int64_t *i_mb_count, double i_count, int b_print_pcm, char *intra )
3645 intra += sprintf( intra, "I16..4%s: %4.1f%% %4.1f%% %4.1f%%",
3646 b_print_pcm ? "..PCM" : "",
3647 i_mb_count[I_16x16]/ i_count,
3648 i_mb_count[I_8x8] / i_count,
3649 i_mb_count[I_4x4] / i_count );
3651 sprintf( intra, " %4.1f%%", i_mb_count[I_PCM] / i_count );
3654 /****************************************************************************
3655 * x264_encoder_close:
3656 ****************************************************************************/
3657 void x264_encoder_close ( x264_t *h )
3659 int64_t i_yuv_size = FRAME_SIZE( h->param.i_width * h->param.i_height );
3660 int64_t i_mb_count_size[2][7] = {{0}};
3662 int b_print_pcm = h->stat.i_mb_count[SLICE_TYPE_I][I_PCM]
3663 || h->stat.i_mb_count[SLICE_TYPE_P][I_PCM]
3664 || h->stat.i_mb_count[SLICE_TYPE_B][I_PCM];
3666 x264_lookahead_delete( h );
3669 x264_opencl_lookahead_delete( h );
3670 x264_opencl_function_t *ocl = h->opencl.ocl;
3673 if( h->param.b_sliced_threads )
3674 x264_threadpool_wait_all( h );
3675 if( h->param.i_threads > 1 )
3676 x264_threadpool_delete( h->threadpool );
3677 if( h->param.i_lookahead_threads > 1 )
3678 x264_threadpool_delete( h->lookaheadpool );
3679 if( h->i_thread_frames > 1 )
3681 for( int i = 0; i < h->i_thread_frames; i++ )
3682 if( h->thread[i]->b_thread_active )
3684 assert( h->thread[i]->fenc->i_reference_count == 1 );
3685 x264_frame_delete( h->thread[i]->fenc );
3688 x264_t *thread_prev = h->thread[h->i_thread_phase];
3689 x264_thread_sync_ratecontrol( h, thread_prev, h );
3690 x264_thread_sync_ratecontrol( thread_prev, thread_prev, h );
3691 h->i_frame = thread_prev->i_frame + 1 - h->i_thread_frames;
3695 /* Slices used and PSNR */
3696 for( int i = 0; i < 3; i++ )
3698 static const uint8_t slice_order[] = { SLICE_TYPE_I, SLICE_TYPE_P, SLICE_TYPE_B };
3699 int i_slice = slice_order[i];
3701 if( h->stat.i_frame_count[i_slice] > 0 )
3703 int i_count = h->stat.i_frame_count[i_slice];
3704 double dur = h->stat.f_frame_duration[i_slice];
3705 if( h->param.analyse.b_psnr )
3707 x264_log( h, X264_LOG_INFO,
3708 "frame %c:%-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",
3709 slice_type_to_char[i_slice],
3711 h->stat.f_frame_qp[i_slice] / i_count,
3712 (double)h->stat.i_frame_size[i_slice] / i_count,
3713 h->stat.f_psnr_mean_y[i_slice] / dur, h->stat.f_psnr_mean_u[i_slice] / dur, h->stat.f_psnr_mean_v[i_slice] / dur,
3714 h->stat.f_psnr_average[i_slice] / dur,
3715 x264_psnr( h->stat.f_ssd_global[i_slice], dur * i_yuv_size ) );
3719 x264_log( h, X264_LOG_INFO,
3720 "frame %c:%-5d Avg QP:%5.2f size:%6.0f\n",
3721 slice_type_to_char[i_slice],
3723 h->stat.f_frame_qp[i_slice] / i_count,
3724 (double)h->stat.i_frame_size[i_slice] / i_count );
3728 if( h->param.i_bframe && h->stat.i_frame_count[SLICE_TYPE_B] )
3732 // weight by number of frames (including the I/P-frames) that are in a sequence of N B-frames
3733 for( int i = 0; i <= h->param.i_bframe; i++ )
3734 den += (i+1) * h->stat.i_consecutive_bframes[i];
3735 for( int i = 0; i <= h->param.i_bframe; i++ )
3736 p += sprintf( p, " %4.1f%%", 100. * (i+1) * h->stat.i_consecutive_bframes[i] / den );
3737 x264_log( h, X264_LOG_INFO, "consecutive B-frames:%s\n", buf );
3740 for( int i_type = 0; i_type < 2; i_type++ )
3741 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
3743 if( i == D_DIRECT_8x8 ) continue; /* direct is counted as its own type */
3744 i_mb_count_size[i_type][x264_mb_partition_pixel_table[i]] += h->stat.i_mb_partition[i_type][i];
3748 if( h->stat.i_frame_count[SLICE_TYPE_I] > 0 )
3750 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_I];
3751 double i_count = h->stat.i_frame_count[SLICE_TYPE_I] * h->mb.i_mb_count / 100.0;
3752 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
3753 x264_log( h, X264_LOG_INFO, "mb I %s\n", buf );
3755 if( h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
3757 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_P];
3758 double i_count = h->stat.i_frame_count[SLICE_TYPE_P] * h->mb.i_mb_count / 100.0;
3759 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_P];
3760 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
3761 x264_log( h, X264_LOG_INFO,
3762 "mb P %s P16..4: %4.1f%% %4.1f%% %4.1f%% %4.1f%% %4.1f%% skip:%4.1f%%\n",
3764 i_mb_size[PIXEL_16x16] / (i_count*4),
3765 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
3766 i_mb_size[PIXEL_8x8] / (i_count*4),
3767 (i_mb_size[PIXEL_8x4] + i_mb_size[PIXEL_4x8]) / (i_count*4),
3768 i_mb_size[PIXEL_4x4] / (i_count*4),
3769 i_mb_count[P_SKIP] / i_count );
3771 if( h->stat.i_frame_count[SLICE_TYPE_B] > 0 )
3773 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_B];
3774 double i_count = h->stat.i_frame_count[SLICE_TYPE_B] * h->mb.i_mb_count / 100.0;
3775 double i_mb_list_count;
3776 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_B];
3777 int64_t list_count[3] = {0}; /* 0 == L0, 1 == L1, 2 == BI */
3778 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
3779 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
3780 for( int j = 0; j < 2; j++ )
3782 int l0 = x264_mb_type_list_table[i][0][j];
3783 int l1 = x264_mb_type_list_table[i][1][j];
3785 list_count[l1+l0*l1] += h->stat.i_mb_count[SLICE_TYPE_B][i] * 2;
3787 list_count[0] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L0_8x8];
3788 list_count[1] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L1_8x8];
3789 list_count[2] += h->stat.i_mb_partition[SLICE_TYPE_B][D_BI_8x8];
3790 i_mb_count[B_DIRECT] += (h->stat.i_mb_partition[SLICE_TYPE_B][D_DIRECT_8x8]+2)/4;
3791 i_mb_list_count = (list_count[0] + list_count[1] + list_count[2]) / 100.0;
3792 sprintf( buf + strlen(buf), " B16..8: %4.1f%% %4.1f%% %4.1f%% direct:%4.1f%% skip:%4.1f%%",
3793 i_mb_size[PIXEL_16x16] / (i_count*4),
3794 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
3795 i_mb_size[PIXEL_8x8] / (i_count*4),
3796 i_mb_count[B_DIRECT] / i_count,
3797 i_mb_count[B_SKIP] / i_count );
3798 if( i_mb_list_count != 0 )
3799 sprintf( buf + strlen(buf), " L0:%4.1f%% L1:%4.1f%% BI:%4.1f%%",
3800 list_count[0] / i_mb_list_count,
3801 list_count[1] / i_mb_list_count,
3802 list_count[2] / i_mb_list_count );
3803 x264_log( h, X264_LOG_INFO, "mb B %s\n", buf );
3806 x264_ratecontrol_summary( h );
3808 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 )
3810 #define SUM3(p) (p[SLICE_TYPE_I] + p[SLICE_TYPE_P] + p[SLICE_TYPE_B])
3811 #define SUM3b(p,o) (p[SLICE_TYPE_I][o] + p[SLICE_TYPE_P][o] + p[SLICE_TYPE_B][o])
3812 int64_t i_i8x8 = SUM3b( h->stat.i_mb_count, I_8x8 );
3813 int64_t i_intra = i_i8x8 + SUM3b( h->stat.i_mb_count, I_4x4 )
3814 + SUM3b( h->stat.i_mb_count, I_16x16 );
3815 int64_t i_all_intra = i_intra + SUM3b( h->stat.i_mb_count, I_PCM);
3816 int64_t i_skip = SUM3b( h->stat.i_mb_count, P_SKIP )
3817 + SUM3b( h->stat.i_mb_count, B_SKIP );
3818 const int i_count = h->stat.i_frame_count[SLICE_TYPE_I] +
3819 h->stat.i_frame_count[SLICE_TYPE_P] +
3820 h->stat.i_frame_count[SLICE_TYPE_B];
3821 int64_t i_mb_count = (int64_t)i_count * h->mb.i_mb_count;
3822 int64_t i_inter = i_mb_count - i_skip - i_intra;
3823 const double duration = h->stat.f_frame_duration[SLICE_TYPE_I] +
3824 h->stat.f_frame_duration[SLICE_TYPE_P] +
3825 h->stat.f_frame_duration[SLICE_TYPE_B];
3826 float f_bitrate = SUM3(h->stat.i_frame_size) / duration / 125;
3828 if( PARAM_INTERLACED )
3830 char *fieldstats = buf;
3833 fieldstats += sprintf( fieldstats, " inter:%.1f%%", h->stat.i_mb_field[1] * 100.0 / i_inter );
3835 fieldstats += sprintf( fieldstats, " skip:%.1f%%", h->stat.i_mb_field[2] * 100.0 / i_skip );
3836 x264_log( h, X264_LOG_INFO, "field mbs: intra: %.1f%%%s\n",
3837 h->stat.i_mb_field[0] * 100.0 / i_intra, buf );
3840 if( h->pps->b_transform_8x8_mode )
3843 if( h->stat.i_mb_count_8x8dct[0] )
3844 sprintf( buf, " inter:%.1f%%", 100. * h->stat.i_mb_count_8x8dct[1] / h->stat.i_mb_count_8x8dct[0] );
3845 x264_log( h, X264_LOG_INFO, "8x8 transform intra:%.1f%%%s\n", 100. * i_i8x8 / i_intra, buf );
3848 if( (h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO ||
3849 (h->stat.i_direct_frames[0] && h->stat.i_direct_frames[1]))
3850 && h->stat.i_frame_count[SLICE_TYPE_B] )
3852 x264_log( h, X264_LOG_INFO, "direct mvs spatial:%.1f%% temporal:%.1f%%\n",
3853 h->stat.i_direct_frames[1] * 100. / h->stat.i_frame_count[SLICE_TYPE_B],
3854 h->stat.i_direct_frames[0] * 100. / h->stat.i_frame_count[SLICE_TYPE_B] );
3858 int csize = CHROMA444 ? 4 : 1;
3859 if( i_mb_count != i_all_intra )
3860 sprintf( buf, " inter: %.1f%% %.1f%% %.1f%%",
3861 h->stat.i_mb_cbp[1] * 100.0 / ((i_mb_count - i_all_intra)*4),
3862 h->stat.i_mb_cbp[3] * 100.0 / ((i_mb_count - i_all_intra)*csize),
3863 h->stat.i_mb_cbp[5] * 100.0 / ((i_mb_count - i_all_intra)*csize) );
3864 x264_log( h, X264_LOG_INFO, "coded y,%s,%s intra: %.1f%% %.1f%% %.1f%%%s\n",
3865 CHROMA444?"u":"uvDC", CHROMA444?"v":"uvAC",
3866 h->stat.i_mb_cbp[0] * 100.0 / (i_all_intra*4),
3867 h->stat.i_mb_cbp[2] * 100.0 / (i_all_intra*csize),
3868 h->stat.i_mb_cbp[4] * 100.0 / (i_all_intra*csize), buf );
3870 int64_t fixed_pred_modes[4][9] = {{0}};
3871 int64_t sum_pred_modes[4] = {0};
3872 for( int i = 0; i <= I_PRED_16x16_DC_128; i++ )
3874 fixed_pred_modes[0][x264_mb_pred_mode16x16_fix[i]] += h->stat.i_mb_pred_mode[0][i];
3875 sum_pred_modes[0] += h->stat.i_mb_pred_mode[0][i];
3877 if( sum_pred_modes[0] )
3878 x264_log( h, X264_LOG_INFO, "i16 v,h,dc,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
3879 fixed_pred_modes[0][0] * 100.0 / sum_pred_modes[0],
3880 fixed_pred_modes[0][1] * 100.0 / sum_pred_modes[0],
3881 fixed_pred_modes[0][2] * 100.0 / sum_pred_modes[0],
3882 fixed_pred_modes[0][3] * 100.0 / sum_pred_modes[0] );
3883 for( int i = 1; i <= 2; i++ )
3885 for( int j = 0; j <= I_PRED_8x8_DC_128; j++ )
3887 fixed_pred_modes[i][x264_mb_pred_mode4x4_fix(j)] += h->stat.i_mb_pred_mode[i][j];
3888 sum_pred_modes[i] += h->stat.i_mb_pred_mode[i][j];
3890 if( sum_pred_modes[i] )
3891 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,
3892 fixed_pred_modes[i][0] * 100.0 / sum_pred_modes[i],
3893 fixed_pred_modes[i][1] * 100.0 / sum_pred_modes[i],
3894 fixed_pred_modes[i][2] * 100.0 / sum_pred_modes[i],
3895 fixed_pred_modes[i][3] * 100.0 / sum_pred_modes[i],
3896 fixed_pred_modes[i][4] * 100.0 / sum_pred_modes[i],
3897 fixed_pred_modes[i][5] * 100.0 / sum_pred_modes[i],
3898 fixed_pred_modes[i][6] * 100.0 / sum_pred_modes[i],
3899 fixed_pred_modes[i][7] * 100.0 / sum_pred_modes[i],
3900 fixed_pred_modes[i][8] * 100.0 / sum_pred_modes[i] );
3902 for( int i = 0; i <= I_PRED_CHROMA_DC_128; i++ )
3904 fixed_pred_modes[3][x264_mb_chroma_pred_mode_fix[i]] += h->stat.i_mb_pred_mode[3][i];
3905 sum_pred_modes[3] += h->stat.i_mb_pred_mode[3][i];
3907 if( sum_pred_modes[3] && !CHROMA444 )
3908 x264_log( h, X264_LOG_INFO, "i8c dc,h,v,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
3909 fixed_pred_modes[3][0] * 100.0 / sum_pred_modes[3],
3910 fixed_pred_modes[3][1] * 100.0 / sum_pred_modes[3],
3911 fixed_pred_modes[3][2] * 100.0 / sum_pred_modes[3],
3912 fixed_pred_modes[3][3] * 100.0 / sum_pred_modes[3] );
3914 if( h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE && h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
3915 x264_log( h, X264_LOG_INFO, "Weighted P-Frames: Y:%.1f%% UV:%.1f%%\n",
3916 h->stat.i_wpred[0] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P],
3917 h->stat.i_wpred[1] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P] );
3919 for( int i_list = 0; i_list < 2; i_list++ )
3920 for( int i_slice = 0; i_slice < 2; i_slice++ )
3925 for( int i = 0; i < X264_REF_MAX*2; i++ )
3926 if( h->stat.i_mb_count_ref[i_slice][i_list][i] )
3928 i_den += h->stat.i_mb_count_ref[i_slice][i_list][i];
3933 for( int i = 0; i <= i_max; i++ )
3934 p += sprintf( p, " %4.1f%%", 100. * h->stat.i_mb_count_ref[i_slice][i_list][i] / i_den );
3935 x264_log( h, X264_LOG_INFO, "ref %c L%d:%s\n", "PB"[i_slice], i_list, buf );
3938 if( h->param.analyse.b_ssim )
3940 float ssim = SUM3( h->stat.f_ssim_mean_y ) / duration;
3941 x264_log( h, X264_LOG_INFO, "SSIM Mean Y:%.7f (%6.3fdb)\n", ssim, x264_ssim( ssim ) );
3943 if( h->param.analyse.b_psnr )
3945 x264_log( h, X264_LOG_INFO,
3946 "PSNR Mean Y:%6.3f U:%6.3f V:%6.3f Avg:%6.3f Global:%6.3f kb/s:%.2f\n",
3947 SUM3( h->stat.f_psnr_mean_y ) / duration,
3948 SUM3( h->stat.f_psnr_mean_u ) / duration,
3949 SUM3( h->stat.f_psnr_mean_v ) / duration,
3950 SUM3( h->stat.f_psnr_average ) / duration,
3951 x264_psnr( SUM3( h->stat.f_ssd_global ), duration * i_yuv_size ),
3955 x264_log( h, X264_LOG_INFO, "kb/s:%.2f\n", f_bitrate );
3959 x264_ratecontrol_delete( h );
3962 if( h->param.rc.psz_stat_out )
3963 free( h->param.rc.psz_stat_out );
3964 if( h->param.rc.psz_stat_in )
3965 free( h->param.rc.psz_stat_in );
3967 x264_cqm_delete( h );
3968 x264_free( h->nal_buffer );
3969 x264_analyse_free_costs( h );
3971 if( h->i_thread_frames > 1 )
3972 h = h->thread[h->i_thread_phase];
3975 x264_frame_delete_list( h->frames.unused[0] );
3976 x264_frame_delete_list( h->frames.unused[1] );
3977 x264_frame_delete_list( h->frames.current );
3978 x264_frame_delete_list( h->frames.blank_unused );
3982 for( int i = 0; i < h->i_thread_frames; i++ )
3983 if( h->thread[i]->b_thread_active )
3984 for( int j = 0; j < h->thread[i]->i_ref[0]; j++ )
3985 if( h->thread[i]->fref[0][j] && h->thread[i]->fref[0][j]->b_duplicate )
3986 x264_frame_delete( h->thread[i]->fref[0][j] );
3988 if( h->param.i_lookahead_threads > 1 )
3989 for( int i = 0; i < h->param.i_lookahead_threads; i++ )
3990 x264_free( h->lookahead_thread[i] );
3992 for( int i = h->param.i_threads - 1; i >= 0; i-- )
3994 x264_frame_t **frame;
3996 if( !h->param.b_sliced_threads || i == 0 )
3998 for( frame = h->thread[i]->frames.reference; *frame; frame++ )
4000 assert( (*frame)->i_reference_count > 0 );
4001 (*frame)->i_reference_count--;
4002 if( (*frame)->i_reference_count == 0 )
4003 x264_frame_delete( *frame );
4005 frame = &h->thread[i]->fdec;
4008 assert( (*frame)->i_reference_count > 0 );
4009 (*frame)->i_reference_count--;
4010 if( (*frame)->i_reference_count == 0 )
4011 x264_frame_delete( *frame );
4013 x264_macroblock_cache_free( h->thread[i] );
4015 x264_macroblock_thread_free( h->thread[i], 0 );
4016 x264_free( h->thread[i]->out.p_bitstream );
4017 x264_free( h->thread[i]->out.nal );
4018 x264_pthread_mutex_destroy( &h->thread[i]->mutex );
4019 x264_pthread_cond_destroy( &h->thread[i]->cv );
4020 x264_free( h->thread[i] );
4023 x264_opencl_close_library( ocl );
4027 int x264_encoder_delayed_frames( x264_t *h )
4029 int delayed_frames = 0;
4030 if( h->i_thread_frames > 1 )
4032 for( int i = 0; i < h->i_thread_frames; i++ )
4033 delayed_frames += h->thread[i]->b_thread_active;
4034 h = h->thread[h->i_thread_phase];
4036 for( int i = 0; h->frames.current[i]; i++ )
4038 x264_pthread_mutex_lock( &h->lookahead->ofbuf.mutex );
4039 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
4040 x264_pthread_mutex_lock( &h->lookahead->next.mutex );
4041 delayed_frames += h->lookahead->ifbuf.i_size + h->lookahead->next.i_size + h->lookahead->ofbuf.i_size;
4042 x264_pthread_mutex_unlock( &h->lookahead->next.mutex );
4043 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
4044 x264_pthread_mutex_unlock( &h->lookahead->ofbuf.mutex );
4045 return delayed_frames;
4048 int x264_encoder_maximum_delayed_frames( x264_t *h )
4050 return h->frames.i_delay;