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 );
408 /* Misalign mask has to be set separately for each thread. */
409 if( h->param.cpu&X264_CPU_SSE_MISALIGN )
410 x264_cpu_mask_misalign_sse();
414 static void x264_lookahead_thread_init( x264_t *h )
417 /* Misalign mask has to be set separately for each thread. */
418 if( h->param.cpu&X264_CPU_SSE_MISALIGN )
419 x264_cpu_mask_misalign_sse();
424 /****************************************************************************
426 ****************************************************************************
427 ****************************** External API*********************************
428 ****************************************************************************
430 ****************************************************************************/
432 static int x264_validate_parameters( x264_t *h, int b_open )
437 int cpuflags = x264_cpu_detect();
440 if( !(cpuflags & X264_CPU_SSE) )
442 x264_log( h, X264_LOG_ERROR, "your cpu does not support SSE1, but x264 was compiled with asm\n");
446 if( !(cpuflags & X264_CPU_MMX2) )
448 x264_log( h, X264_LOG_ERROR, "your cpu does not support MMXEXT, but x264 was compiled with asm\n");
452 if( !fail && !(cpuflags & X264_CPU_CMOV) )
454 x264_log( h, X264_LOG_ERROR, "your cpu does not support CMOV, but x264 was compiled with asm\n");
459 x264_log( h, X264_LOG_ERROR, "to run x264, recompile without asm (configure --disable-asm)\n");
466 h->param.b_interlaced = !!PARAM_INTERLACED;
468 if( h->param.b_interlaced )
470 x264_log( h, X264_LOG_ERROR, "not compiled with interlaced support\n" );
475 if( h->param.i_width <= 0 || h->param.i_height <= 0 )
477 x264_log( h, X264_LOG_ERROR, "invalid width x height (%dx%d)\n",
478 h->param.i_width, h->param.i_height );
482 int i_csp = h->param.i_csp & X264_CSP_MASK;
483 #if X264_CHROMA_FORMAT
484 if( CHROMA_FORMAT != CHROMA_420 && i_csp >= X264_CSP_I420 && i_csp <= X264_CSP_NV12 )
486 x264_log( h, X264_LOG_ERROR, "not compiled with 4:2:0 support\n" );
489 else if( CHROMA_FORMAT != CHROMA_422 && i_csp >= X264_CSP_I422 && i_csp <= X264_CSP_NV16 )
491 x264_log( h, X264_LOG_ERROR, "not compiled with 4:2:2 support\n" );
494 else if( CHROMA_FORMAT != CHROMA_444 && i_csp >= X264_CSP_I444 && i_csp <= X264_CSP_RGB )
496 x264_log( h, X264_LOG_ERROR, "not compiled with 4:4:4 support\n" );
500 if( i_csp <= X264_CSP_NONE || i_csp >= X264_CSP_MAX )
502 x264_log( h, X264_LOG_ERROR, "invalid CSP (only I420/YV12/NV12/I422/YV16/NV16/I444/YV24/BGR/BGRA/RGB supported)\n" );
506 if( i_csp < X264_CSP_I444 && h->param.i_width % 2 )
508 x264_log( h, X264_LOG_ERROR, "width not divisible by 2 (%dx%d)\n",
509 h->param.i_width, h->param.i_height );
513 if( i_csp < X264_CSP_I422 && PARAM_INTERLACED && h->param.i_height % 4 )
515 x264_log( h, X264_LOG_ERROR, "height not divisible by 4 (%dx%d)\n",
516 h->param.i_width, h->param.i_height );
520 if( (i_csp < X264_CSP_I422 || PARAM_INTERLACED) && h->param.i_height % 2 )
522 x264_log( h, X264_LOG_ERROR, "height not divisible by 2 (%dx%d)\n",
523 h->param.i_width, h->param.i_height );
527 if( (h->param.crop_rect.i_left + h->param.crop_rect.i_right ) >= h->param.i_width ||
528 (h->param.crop_rect.i_top + h->param.crop_rect.i_bottom) >= h->param.i_height )
530 x264_log( h, X264_LOG_ERROR, "invalid crop-rect %u,%u,%u,%u\n", h->param.crop_rect.i_left,
531 h->param.crop_rect.i_top, h->param.crop_rect.i_right, h->param.crop_rect.i_bottom );
535 if( h->param.i_threads == X264_THREADS_AUTO )
536 h->param.i_threads = x264_cpu_num_processors() * (h->param.b_sliced_threads?2:3)/2;
537 int max_sliced_threads = X264_MAX( 1, (h->param.i_height+15)/16 / 4 );
538 if( h->param.i_threads > 1 )
541 x264_log( h, X264_LOG_WARNING, "not compiled with thread support!\n");
542 h->param.i_threads = 1;
544 /* Avoid absurdly small thread slices as they can reduce performance
545 * and VBV compliance. Capped at an arbitrary 4 rows per thread. */
546 if( h->param.b_sliced_threads )
547 h->param.i_threads = X264_MIN( h->param.i_threads, max_sliced_threads );
549 h->param.i_threads = x264_clip3( h->param.i_threads, 1, X264_THREAD_MAX );
550 if( h->param.i_threads == 1 )
552 h->param.b_sliced_threads = 0;
553 h->param.i_lookahead_threads = 1;
555 h->i_thread_frames = h->param.b_sliced_threads ? 1 : h->param.i_threads;
556 if( h->i_thread_frames > 1 )
557 h->param.nalu_process = NULL;
559 if( h->param.b_opencl )
562 x264_log( h, X264_LOG_WARNING, "OpenCL: not compiled with OpenCL support, disabling\n" );
563 h->param.b_opencl = 0;
565 x264_log( h, X264_LOG_WARNING, "OpenCL lookahead does not support high bit depth, disabling opencl\n" );
566 h->param.b_opencl = 0;
568 if( h->param.i_width < 32 || h->param.i_height < 32 )
570 x264_log( h, X264_LOG_WARNING, "OpenCL: frame size is too small, disabling opencl\n" );
571 h->param.b_opencl = 0;
574 if( h->param.opencl_device_id && h->param.i_opencl_device )
576 x264_log( h, X264_LOG_WARNING, "OpenCL: device id and device skip count configured; dropping skip\n" );
577 h->param.i_opencl_device = 0;
581 h->param.i_keyint_max = x264_clip3( h->param.i_keyint_max, 1, X264_KEYINT_MAX_INFINITE );
582 if( h->param.i_keyint_max == 1 )
584 h->param.b_intra_refresh = 0;
585 h->param.analyse.i_weighted_pred = 0;
588 h->param.i_frame_packing = x264_clip3( h->param.i_frame_packing, -1, 5 );
590 /* Detect default ffmpeg settings and terminate with an error. */
594 score += h->param.analyse.i_me_range == 0;
595 score += h->param.rc.i_qp_step == 3;
596 score += h->param.i_keyint_max == 12;
597 score += h->param.rc.i_qp_min == 2;
598 score += h->param.rc.i_qp_max == 31;
599 score += h->param.rc.f_qcompress == 0.5;
600 score += fabs(h->param.rc.f_ip_factor - 1.25) < 0.01;
601 score += fabs(h->param.rc.f_pb_factor - 1.25) < 0.01;
602 score += h->param.analyse.inter == 0 && h->param.analyse.i_subpel_refine == 8;
605 x264_log( h, X264_LOG_ERROR, "broken ffmpeg default settings detected\n" );
606 x264_log( h, X264_LOG_ERROR, "use an encoding preset (e.g. -vpre medium)\n" );
607 x264_log( h, X264_LOG_ERROR, "preset usage: -vpre <speed> -vpre <profile>\n" );
608 x264_log( h, X264_LOG_ERROR, "speed presets are listed in x264 --help\n" );
609 x264_log( h, X264_LOG_ERROR, "profile is optional; x264 defaults to high\n" );
614 if( h->param.rc.i_rc_method < 0 || h->param.rc.i_rc_method > 2 )
616 x264_log( h, X264_LOG_ERROR, "no ratecontrol method specified\n" );
619 h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, -QP_BD_OFFSET, 51 );
620 h->param.rc.f_rf_constant_max = x264_clip3f( h->param.rc.f_rf_constant_max, -QP_BD_OFFSET, 51 );
621 h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, 0, QP_MAX );
622 h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 11 );
623 h->param.rc.f_ip_factor = X264_MAX( h->param.rc.f_ip_factor, 0.01f );
624 h->param.rc.f_pb_factor = X264_MAX( h->param.rc.f_pb_factor, 0.01f );
625 if( h->param.rc.i_rc_method == X264_RC_CRF )
627 h->param.rc.i_qp_constant = h->param.rc.f_rf_constant + QP_BD_OFFSET;
628 h->param.rc.i_bitrate = 0;
630 if( b_open && (h->param.rc.i_rc_method == X264_RC_CQP || h->param.rc.i_rc_method == X264_RC_CRF)
631 && h->param.rc.i_qp_constant == 0 )
633 h->mb.b_lossless = 1;
634 h->param.i_cqm_preset = X264_CQM_FLAT;
635 h->param.psz_cqm_file = NULL;
636 h->param.rc.i_rc_method = X264_RC_CQP;
637 h->param.rc.f_ip_factor = 1;
638 h->param.rc.f_pb_factor = 1;
639 h->param.analyse.b_psnr = 0;
640 h->param.analyse.b_ssim = 0;
641 h->param.analyse.i_chroma_qp_offset = 0;
642 h->param.analyse.i_trellis = 0;
643 h->param.analyse.b_fast_pskip = 0;
644 h->param.analyse.i_noise_reduction = 0;
645 h->param.analyse.b_psy = 0;
646 h->param.i_bframe = 0;
647 /* 8x8dct is not useful without RD in CAVLC lossless */
648 if( !h->param.b_cabac && h->param.analyse.i_subpel_refine < 6 )
649 h->param.analyse.b_transform_8x8 = 0;
651 if( h->param.rc.i_rc_method == X264_RC_CQP )
653 float qp_p = h->param.rc.i_qp_constant;
654 float qp_i = qp_p - 6*log2f( h->param.rc.f_ip_factor );
655 float qp_b = qp_p + 6*log2f( h->param.rc.f_pb_factor );
656 h->param.rc.i_qp_min = x264_clip3( (int)(X264_MIN3( qp_p, qp_i, qp_b )), 0, QP_MAX );
657 h->param.rc.i_qp_max = x264_clip3( (int)(X264_MAX3( qp_p, qp_i, qp_b ) + .999), 0, QP_MAX );
658 h->param.rc.i_aq_mode = 0;
659 h->param.rc.b_mb_tree = 0;
660 h->param.rc.i_bitrate = 0;
662 h->param.rc.i_qp_max = x264_clip3( h->param.rc.i_qp_max, 0, QP_MAX );
663 h->param.rc.i_qp_min = x264_clip3( h->param.rc.i_qp_min, 0, h->param.rc.i_qp_max );
664 h->param.rc.i_qp_step = x264_clip3( h->param.rc.i_qp_step, 2, QP_MAX );
665 h->param.rc.i_bitrate = x264_clip3( h->param.rc.i_bitrate, 0, 2000000 );
666 if( h->param.rc.i_rc_method == X264_RC_ABR && !h->param.rc.i_bitrate )
668 x264_log( h, X264_LOG_ERROR, "bitrate not specified\n" );
671 h->param.rc.i_vbv_buffer_size = x264_clip3( h->param.rc.i_vbv_buffer_size, 0, 2000000 );
672 h->param.rc.i_vbv_max_bitrate = x264_clip3( h->param.rc.i_vbv_max_bitrate, 0, 2000000 );
673 h->param.rc.f_vbv_buffer_init = x264_clip3f( h->param.rc.f_vbv_buffer_init, 0, 2000000 );
674 if( h->param.rc.i_vbv_buffer_size )
676 if( h->param.rc.i_rc_method == X264_RC_CQP )
678 x264_log( h, X264_LOG_WARNING, "VBV is incompatible with constant QP, ignored.\n" );
679 h->param.rc.i_vbv_max_bitrate = 0;
680 h->param.rc.i_vbv_buffer_size = 0;
682 else if( h->param.rc.i_vbv_max_bitrate == 0 )
684 if( h->param.rc.i_rc_method == X264_RC_ABR )
686 x264_log( h, X264_LOG_WARNING, "VBV maxrate unspecified, assuming CBR\n" );
687 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
691 x264_log( h, X264_LOG_WARNING, "VBV bufsize set but maxrate unspecified, ignored\n" );
692 h->param.rc.i_vbv_buffer_size = 0;
695 else if( h->param.rc.i_vbv_max_bitrate < h->param.rc.i_bitrate &&
696 h->param.rc.i_rc_method == X264_RC_ABR )
698 x264_log( h, X264_LOG_WARNING, "max bitrate less than average bitrate, assuming CBR\n" );
699 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
702 else if( h->param.rc.i_vbv_max_bitrate )
704 x264_log( h, X264_LOG_WARNING, "VBV maxrate specified, but no bufsize, ignored\n" );
705 h->param.rc.i_vbv_max_bitrate = 0;
708 h->param.i_slice_max_size = X264_MAX( h->param.i_slice_max_size, 0 );
709 h->param.i_slice_max_mbs = X264_MAX( h->param.i_slice_max_mbs, 0 );
710 h->param.i_slice_min_mbs = X264_MAX( h->param.i_slice_min_mbs, 0 );
711 if( h->param.i_slice_max_mbs )
712 h->param.i_slice_min_mbs = X264_MIN( h->param.i_slice_min_mbs, h->param.i_slice_max_mbs/2 );
713 else if( !h->param.i_slice_max_size )
714 h->param.i_slice_min_mbs = 0;
715 if( PARAM_INTERLACED && h->param.i_slice_min_mbs )
717 x264_log( h, X264_LOG_WARNING, "interlace + slice-min-mbs is not implemented\n" );
718 h->param.i_slice_min_mbs = 0;
720 int mb_width = (h->param.i_width+15)/16;
721 if( h->param.i_slice_min_mbs > mb_width )
723 x264_log( h, X264_LOG_WARNING, "slice-min-mbs > row mb size (%d) not implemented\n", mb_width );
724 h->param.i_slice_min_mbs = mb_width;
727 int max_slices = (h->param.i_height+((16<<PARAM_INTERLACED)-1))/(16<<PARAM_INTERLACED);
728 if( h->param.b_sliced_threads )
729 h->param.i_slice_count = x264_clip3( h->param.i_threads, 0, max_slices );
732 h->param.i_slice_count = x264_clip3( h->param.i_slice_count, 0, max_slices );
733 if( h->param.i_slice_max_mbs || h->param.i_slice_max_size )
734 h->param.i_slice_count = 0;
736 if( h->param.i_slice_count_max > 0 )
737 h->param.i_slice_count_max = X264_MAX( h->param.i_slice_count, h->param.i_slice_count_max );
739 if( h->param.b_bluray_compat )
741 h->param.i_bframe_pyramid = X264_MIN( X264_B_PYRAMID_STRICT, h->param.i_bframe_pyramid );
742 h->param.i_bframe = X264_MIN( h->param.i_bframe, 3 );
744 h->param.i_nal_hrd = X264_MAX( h->param.i_nal_hrd, X264_NAL_HRD_VBR );
745 h->param.i_slice_max_size = 0;
746 h->param.i_slice_max_mbs = 0;
747 h->param.b_intra_refresh = 0;
748 h->param.i_frame_reference = X264_MIN( h->param.i_frame_reference, 6 );
749 h->param.i_dpb_size = X264_MIN( h->param.i_dpb_size, 6 );
750 /* Don't use I-frames, because Blu-ray treats them the same as IDR. */
751 h->param.i_keyint_min = 1;
752 /* Due to the proliferation of broken players that don't handle dupes properly. */
753 h->param.analyse.i_weighted_pred = X264_MIN( h->param.analyse.i_weighted_pred, X264_WEIGHTP_SIMPLE );
754 if( h->param.b_fake_interlaced )
755 h->param.b_pic_struct = 1;
758 h->param.i_frame_reference = x264_clip3( h->param.i_frame_reference, 1, X264_REF_MAX );
759 h->param.i_dpb_size = x264_clip3( h->param.i_dpb_size, 1, X264_REF_MAX );
760 if( h->param.i_scenecut_threshold < 0 )
761 h->param.i_scenecut_threshold = 0;
762 h->param.analyse.i_direct_mv_pred = x264_clip3( h->param.analyse.i_direct_mv_pred, X264_DIRECT_PRED_NONE, X264_DIRECT_PRED_AUTO );
763 if( !h->param.analyse.i_subpel_refine && h->param.analyse.i_direct_mv_pred > X264_DIRECT_PRED_SPATIAL )
765 x264_log( h, X264_LOG_WARNING, "subme=0 + direct=temporal is not supported\n" );
766 h->param.analyse.i_direct_mv_pred = X264_DIRECT_PRED_SPATIAL;
768 h->param.i_bframe = x264_clip3( h->param.i_bframe, 0, X264_MIN( X264_BFRAME_MAX, h->param.i_keyint_max-1 ) );
769 h->param.i_bframe_bias = x264_clip3( h->param.i_bframe_bias, -90, 100 );
770 if( h->param.i_bframe <= 1 )
771 h->param.i_bframe_pyramid = X264_B_PYRAMID_NONE;
772 h->param.i_bframe_pyramid = x264_clip3( h->param.i_bframe_pyramid, X264_B_PYRAMID_NONE, X264_B_PYRAMID_NORMAL );
773 h->param.i_bframe_adaptive = x264_clip3( h->param.i_bframe_adaptive, X264_B_ADAPT_NONE, X264_B_ADAPT_TRELLIS );
774 if( !h->param.i_bframe )
776 h->param.i_bframe_adaptive = X264_B_ADAPT_NONE;
777 h->param.analyse.i_direct_mv_pred = 0;
778 h->param.analyse.b_weighted_bipred = 0;
779 h->param.b_open_gop = 0;
781 if( h->param.b_intra_refresh && h->param.i_bframe_pyramid == X264_B_PYRAMID_NORMAL )
783 x264_log( h, X264_LOG_WARNING, "b-pyramid normal + intra-refresh is not supported\n" );
784 h->param.i_bframe_pyramid = X264_B_PYRAMID_STRICT;
786 if( h->param.b_intra_refresh && (h->param.i_frame_reference > 1 || h->param.i_dpb_size > 1) )
788 x264_log( h, X264_LOG_WARNING, "ref > 1 + intra-refresh is not supported\n" );
789 h->param.i_frame_reference = 1;
790 h->param.i_dpb_size = 1;
792 if( h->param.b_intra_refresh && h->param.b_open_gop )
794 x264_log( h, X264_LOG_WARNING, "intra-refresh is not compatible with open-gop\n" );
795 h->param.b_open_gop = 0;
797 if( !h->param.i_fps_num || !h->param.i_fps_den )
799 h->param.i_fps_num = 25;
800 h->param.i_fps_den = 1;
802 float fps = (float) h->param.i_fps_num / h->param.i_fps_den;
803 if( h->param.i_keyint_min == X264_KEYINT_MIN_AUTO )
804 h->param.i_keyint_min = X264_MIN( h->param.i_keyint_max / 10, fps );
805 h->param.i_keyint_min = x264_clip3( h->param.i_keyint_min, 1, h->param.i_keyint_max/2+1 );
806 h->param.rc.i_lookahead = x264_clip3( h->param.rc.i_lookahead, 0, X264_LOOKAHEAD_MAX );
808 int maxrate = X264_MAX( h->param.rc.i_vbv_max_bitrate, h->param.rc.i_bitrate );
809 float bufsize = maxrate ? (float)h->param.rc.i_vbv_buffer_size / maxrate : 0;
810 h->param.rc.i_lookahead = X264_MIN( h->param.rc.i_lookahead, X264_MAX( h->param.i_keyint_max, bufsize*fps ) );
813 if( !h->param.i_timebase_num || !h->param.i_timebase_den || !(h->param.b_vfr_input || h->param.b_pulldown) )
815 h->param.i_timebase_num = h->param.i_fps_den;
816 h->param.i_timebase_den = h->param.i_fps_num;
819 h->param.rc.f_qcompress = x264_clip3f( h->param.rc.f_qcompress, 0.0, 1.0 );
820 if( h->param.i_keyint_max == 1 || h->param.rc.f_qcompress == 1 )
821 h->param.rc.b_mb_tree = 0;
822 if( (!h->param.b_intra_refresh && h->param.i_keyint_max != X264_KEYINT_MAX_INFINITE) &&
823 !h->param.rc.i_lookahead && h->param.rc.b_mb_tree )
825 x264_log( h, X264_LOG_WARNING, "lookaheadless mb-tree requires intra refresh or infinite keyint\n" );
826 h->param.rc.b_mb_tree = 0;
828 if( b_open && h->param.rc.b_stat_read )
829 h->param.rc.i_lookahead = 0;
831 if( h->param.i_sync_lookahead < 0 )
832 h->param.i_sync_lookahead = h->param.i_bframe + 1;
833 h->param.i_sync_lookahead = X264_MIN( h->param.i_sync_lookahead, X264_LOOKAHEAD_MAX );
834 if( h->param.rc.b_stat_read || h->i_thread_frames == 1 )
835 h->param.i_sync_lookahead = 0;
837 h->param.i_sync_lookahead = 0;
840 h->param.i_deblocking_filter_alphac0 = x264_clip3( h->param.i_deblocking_filter_alphac0, -6, 6 );
841 h->param.i_deblocking_filter_beta = x264_clip3( h->param.i_deblocking_filter_beta, -6, 6 );
842 h->param.analyse.i_luma_deadzone[0] = x264_clip3( h->param.analyse.i_luma_deadzone[0], 0, 32 );
843 h->param.analyse.i_luma_deadzone[1] = x264_clip3( h->param.analyse.i_luma_deadzone[1], 0, 32 );
845 h->param.i_cabac_init_idc = x264_clip3( h->param.i_cabac_init_idc, 0, 2 );
847 if( h->param.i_cqm_preset < X264_CQM_FLAT || h->param.i_cqm_preset > X264_CQM_CUSTOM )
848 h->param.i_cqm_preset = X264_CQM_FLAT;
850 if( h->param.analyse.i_me_method < X264_ME_DIA ||
851 h->param.analyse.i_me_method > X264_ME_TESA )
852 h->param.analyse.i_me_method = X264_ME_HEX;
853 h->param.analyse.i_me_range = x264_clip3( h->param.analyse.i_me_range, 4, 1024 );
854 if( h->param.analyse.i_me_range > 16 && h->param.analyse.i_me_method <= X264_ME_HEX )
855 h->param.analyse.i_me_range = 16;
856 if( h->param.analyse.i_me_method == X264_ME_TESA &&
857 (h->mb.b_lossless || h->param.analyse.i_subpel_refine <= 1) )
858 h->param.analyse.i_me_method = X264_ME_ESA;
859 h->param.analyse.b_mixed_references = h->param.analyse.b_mixed_references && h->param.i_frame_reference > 1;
860 h->param.analyse.inter &= X264_ANALYSE_PSUB16x16|X264_ANALYSE_PSUB8x8|X264_ANALYSE_BSUB16x16|
861 X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
862 h->param.analyse.intra &= X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
863 if( !(h->param.analyse.inter & X264_ANALYSE_PSUB16x16) )
864 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
865 if( !h->param.analyse.b_transform_8x8 )
867 h->param.analyse.inter &= ~X264_ANALYSE_I8x8;
868 h->param.analyse.intra &= ~X264_ANALYSE_I8x8;
870 h->param.analyse.i_trellis = x264_clip3( h->param.analyse.i_trellis, 0, 2 );
871 h->param.rc.i_aq_mode = x264_clip3( h->param.rc.i_aq_mode, 0, 2 );
872 h->param.rc.f_aq_strength = x264_clip3f( h->param.rc.f_aq_strength, 0, 3 );
873 if( h->param.rc.f_aq_strength == 0 )
874 h->param.rc.i_aq_mode = 0;
876 if( h->param.i_log_level < X264_LOG_INFO )
878 h->param.analyse.b_psnr = 0;
879 h->param.analyse.b_ssim = 0;
881 /* Warn users trying to measure PSNR/SSIM with psy opts on. */
882 if( b_open && (h->param.analyse.b_psnr || h->param.analyse.b_ssim) )
886 if( h->param.analyse.b_psy )
888 s = h->param.analyse.b_psnr ? "psnr" : "ssim";
889 x264_log( h, X264_LOG_WARNING, "--%s used with psy on: results will be invalid!\n", s );
891 else if( !h->param.rc.i_aq_mode && h->param.analyse.b_ssim )
893 x264_log( h, X264_LOG_WARNING, "--ssim used with AQ off: results will be invalid!\n" );
896 else if( h->param.rc.i_aq_mode && h->param.analyse.b_psnr )
898 x264_log( h, X264_LOG_WARNING, "--psnr used with AQ on: results will be invalid!\n" );
902 x264_log( h, X264_LOG_WARNING, "--tune %s should be used if attempting to benchmark %s!\n", s, s );
905 if( !h->param.analyse.b_psy )
907 h->param.analyse.f_psy_rd = 0;
908 h->param.analyse.f_psy_trellis = 0;
910 h->param.analyse.f_psy_rd = x264_clip3f( h->param.analyse.f_psy_rd, 0, 10 );
911 h->param.analyse.f_psy_trellis = x264_clip3f( h->param.analyse.f_psy_trellis, 0, 10 );
912 h->mb.i_psy_rd = h->param.analyse.i_subpel_refine >= 6 ? FIX8( h->param.analyse.f_psy_rd ) : 0;
913 h->mb.i_psy_trellis = h->param.analyse.i_trellis ? FIX8( h->param.analyse.f_psy_trellis / 4 ) : 0;
914 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -32, 32);
915 /* In 4:4:4 mode, chroma gets twice as much resolution, so we can halve its quality. */
916 if( b_open && i_csp >= X264_CSP_I444 && i_csp < X264_CSP_BGR && h->param.analyse.b_psy )
917 h->param.analyse.i_chroma_qp_offset += 6;
918 /* Psy RDO increases overall quantizers to improve the quality of luma--this indirectly hurts chroma quality */
919 /* so we lower the chroma QP offset to compensate */
920 if( b_open && h->mb.i_psy_rd )
921 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_rd < 0.25 ? 1 : 2;
922 /* Psy trellis has a similar effect. */
923 if( b_open && h->mb.i_psy_trellis )
924 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_trellis < 0.25 ? 1 : 2;
925 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
926 /* MB-tree requires AQ to be on, even if the strength is zero. */
927 if( !h->param.rc.i_aq_mode && h->param.rc.b_mb_tree )
929 h->param.rc.i_aq_mode = 1;
930 h->param.rc.f_aq_strength = 0;
932 h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
933 if( h->param.analyse.i_subpel_refine >= 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
934 h->param.analyse.i_subpel_refine = 9;
937 const x264_level_t *l = x264_levels;
938 if( h->param.i_level_idc < 0 )
940 int maxrate_bak = h->param.rc.i_vbv_max_bitrate;
941 if( h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.i_vbv_buffer_size <= 0 )
942 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate * 2;
943 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
944 do h->param.i_level_idc = l->level_idc;
945 while( l[1].level_idc && x264_validate_levels( h, 0 ) && l++ );
946 h->param.rc.i_vbv_max_bitrate = maxrate_bak;
950 while( l->level_idc && l->level_idc != h->param.i_level_idc )
952 if( l->level_idc == 0 )
954 x264_log( h, X264_LOG_ERROR, "invalid level_idc: %d\n", h->param.i_level_idc );
958 if( h->param.analyse.i_mv_range <= 0 )
959 h->param.analyse.i_mv_range = l->mv_range >> PARAM_INTERLACED;
961 h->param.analyse.i_mv_range = x264_clip3(h->param.analyse.i_mv_range, 32, 512 >> PARAM_INTERLACED);
964 h->param.analyse.i_weighted_pred = x264_clip3( h->param.analyse.i_weighted_pred, X264_WEIGHTP_NONE, X264_WEIGHTP_SMART );
966 if( h->param.i_lookahead_threads == X264_THREADS_AUTO )
968 if( h->param.b_sliced_threads )
969 h->param.i_lookahead_threads = h->param.i_threads;
972 /* If we're using much slower lookahead settings than encoding settings, it helps a lot to use
973 * more lookahead threads. This typically happens in the first pass of a two-pass encode, so
974 * try to guess at this sort of case.
976 * Tuned by a little bit of real encoding with the various presets. */
977 int badapt = h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS;
978 int subme = X264_MIN( h->param.analyse.i_subpel_refine / 3, 3 ) + (h->param.analyse.i_subpel_refine > 1);
979 int bframes = X264_MIN( (h->param.i_bframe - 1) / 3, 3 );
981 /* [b-adapt 0/1 vs 2][quantized subme][quantized bframes] */
982 static const uint8_t lookahead_thread_div[2][5][4] =
983 {{{6,6,6,6}, {3,3,3,3}, {4,4,4,4}, {6,6,6,6}, {12,12,12,12}},
984 {{3,2,1,1}, {2,1,1,1}, {4,3,2,1}, {6,4,3,2}, {12, 9, 6, 4}}};
986 h->param.i_lookahead_threads = h->param.i_threads / lookahead_thread_div[badapt][subme][bframes];
987 /* Since too many lookahead threads significantly degrades lookahead accuracy, limit auto
988 * lookahead threads to about 8 macroblock rows high each at worst. This number is chosen
989 * pretty much arbitrarily. */
990 h->param.i_lookahead_threads = X264_MIN( h->param.i_lookahead_threads, h->param.i_height / 128 );
993 h->param.i_lookahead_threads = x264_clip3( h->param.i_lookahead_threads, 1, X264_MIN( max_sliced_threads, X264_LOOKAHEAD_THREAD_MAX ) );
995 if( PARAM_INTERLACED )
997 if( h->param.analyse.i_me_method >= X264_ME_ESA )
999 x264_log( h, X264_LOG_WARNING, "interlace + me=esa is not implemented\n" );
1000 h->param.analyse.i_me_method = X264_ME_UMH;
1002 if( h->param.analyse.i_weighted_pred > 0 )
1004 x264_log( h, X264_LOG_WARNING, "interlace + weightp is not implemented\n" );
1005 h->param.analyse.i_weighted_pred = X264_WEIGHTP_NONE;
1009 if( !h->param.analyse.i_weighted_pred && h->param.rc.b_mb_tree && h->param.analyse.b_psy )
1010 h->param.analyse.i_weighted_pred = X264_WEIGHTP_FAKE;
1012 if( h->i_thread_frames > 1 )
1014 int r = h->param.analyse.i_mv_range_thread;
1018 // half of the available space is reserved and divided evenly among the threads,
1019 // the rest is allocated to whichever thread is far enough ahead to use it.
1020 // reserving more space increases quality for some videos, but costs more time
1021 // in thread synchronization.
1022 int max_range = (h->param.i_height + X264_THREAD_HEIGHT) / h->i_thread_frames - X264_THREAD_HEIGHT;
1025 r = X264_MAX( r, h->param.analyse.i_me_range );
1026 r = X264_MIN( r, h->param.analyse.i_mv_range );
1027 // round up to use the whole mb row
1028 r2 = (r & ~15) + ((-X264_THREAD_HEIGHT) & 15);
1031 x264_log( h, X264_LOG_DEBUG, "using mv_range_thread = %d\n", r2 );
1032 h->param.analyse.i_mv_range_thread = r2;
1035 if( h->param.rc.f_rate_tolerance < 0 )
1036 h->param.rc.f_rate_tolerance = 0;
1037 if( h->param.rc.f_qblur < 0 )
1038 h->param.rc.f_qblur = 0;
1039 if( h->param.rc.f_complexity_blur < 0 )
1040 h->param.rc.f_complexity_blur = 0;
1042 h->param.i_sps_id &= 31;
1044 if( PARAM_INTERLACED )
1045 h->param.b_pic_struct = 1;
1047 h->param.i_nal_hrd = x264_clip3( h->param.i_nal_hrd, X264_NAL_HRD_NONE, X264_NAL_HRD_CBR );
1049 if( h->param.i_nal_hrd && !h->param.rc.i_vbv_buffer_size )
1051 x264_log( h, X264_LOG_WARNING, "NAL HRD parameters require VBV parameters\n" );
1052 h->param.i_nal_hrd = X264_NAL_HRD_NONE;
1055 if( h->param.i_nal_hrd == X264_NAL_HRD_CBR &&
1056 (h->param.rc.i_bitrate != h->param.rc.i_vbv_max_bitrate || !h->param.rc.i_vbv_max_bitrate) )
1058 x264_log( h, X264_LOG_WARNING, "CBR HRD requires constant bitrate\n" );
1059 h->param.i_nal_hrd = X264_NAL_HRD_VBR;
1062 /* ensure the booleans are 0 or 1 so they can be used in math */
1063 #define BOOLIFY(x) h->param.x = !!h->param.x
1065 BOOLIFY( b_constrained_intra );
1066 BOOLIFY( b_deblocking_filter );
1067 BOOLIFY( b_deterministic );
1068 BOOLIFY( b_sliced_threads );
1069 BOOLIFY( b_interlaced );
1070 BOOLIFY( b_intra_refresh );
1071 BOOLIFY( b_visualize );
1073 BOOLIFY( b_repeat_headers );
1074 BOOLIFY( b_annexb );
1075 BOOLIFY( b_vfr_input );
1076 BOOLIFY( b_pulldown );
1078 BOOLIFY( b_pic_struct );
1079 BOOLIFY( b_fake_interlaced );
1080 BOOLIFY( b_open_gop );
1081 BOOLIFY( b_bluray_compat );
1082 BOOLIFY( b_full_recon );
1083 BOOLIFY( b_opencl );
1084 BOOLIFY( analyse.b_transform_8x8 );
1085 BOOLIFY( analyse.b_weighted_bipred );
1086 BOOLIFY( analyse.b_chroma_me );
1087 BOOLIFY( analyse.b_mixed_references );
1088 BOOLIFY( analyse.b_fast_pskip );
1089 BOOLIFY( analyse.b_dct_decimate );
1090 BOOLIFY( analyse.b_psy );
1091 BOOLIFY( analyse.b_psnr );
1092 BOOLIFY( analyse.b_ssim );
1093 BOOLIFY( rc.b_stat_write );
1094 BOOLIFY( rc.b_stat_read );
1095 BOOLIFY( rc.b_mb_tree );
1101 static void mbcmp_init( x264_t *h )
1103 int satd = !h->mb.b_lossless && h->param.analyse.i_subpel_refine > 1;
1104 memcpy( h->pixf.mbcmp, satd ? h->pixf.satd : h->pixf.sad_aligned, sizeof(h->pixf.mbcmp) );
1105 memcpy( h->pixf.mbcmp_unaligned, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.mbcmp_unaligned) );
1106 h->pixf.intra_mbcmp_x3_16x16 = satd ? h->pixf.intra_satd_x3_16x16 : h->pixf.intra_sad_x3_16x16;
1107 h->pixf.intra_mbcmp_x3_8x16c = satd ? h->pixf.intra_satd_x3_8x16c : h->pixf.intra_sad_x3_8x16c;
1108 h->pixf.intra_mbcmp_x3_8x8c = satd ? h->pixf.intra_satd_x3_8x8c : h->pixf.intra_sad_x3_8x8c;
1109 h->pixf.intra_mbcmp_x3_8x8 = satd ? h->pixf.intra_sa8d_x3_8x8 : h->pixf.intra_sad_x3_8x8;
1110 h->pixf.intra_mbcmp_x3_4x4 = satd ? h->pixf.intra_satd_x3_4x4 : h->pixf.intra_sad_x3_4x4;
1111 h->pixf.intra_mbcmp_x9_4x4 = h->param.b_cpu_independent || h->mb.b_lossless ? NULL
1112 : satd ? h->pixf.intra_satd_x9_4x4 : h->pixf.intra_sad_x9_4x4;
1113 h->pixf.intra_mbcmp_x9_8x8 = h->param.b_cpu_independent || h->mb.b_lossless ? NULL
1114 : satd ? h->pixf.intra_sa8d_x9_8x8 : h->pixf.intra_sad_x9_8x8;
1115 satd &= h->param.analyse.i_me_method == X264_ME_TESA;
1116 memcpy( h->pixf.fpelcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.fpelcmp) );
1117 memcpy( h->pixf.fpelcmp_x3, satd ? h->pixf.satd_x3 : h->pixf.sad_x3, sizeof(h->pixf.fpelcmp_x3) );
1118 memcpy( h->pixf.fpelcmp_x4, satd ? h->pixf.satd_x4 : h->pixf.sad_x4, sizeof(h->pixf.fpelcmp_x4) );
1121 static void chroma_dsp_init( x264_t *h )
1123 memcpy( h->luma2chroma_pixel, x264_luma2chroma_pixel[CHROMA_FORMAT], sizeof(h->luma2chroma_pixel) );
1125 switch( CHROMA_FORMAT )
1128 memcpy( h->predict_chroma, h->predict_8x8c, sizeof(h->predict_chroma) );
1129 h->mc.prefetch_fenc = h->mc.prefetch_fenc_420;
1130 h->loopf.deblock_chroma[0] = h->loopf.deblock_h_chroma_420;
1131 h->loopf.deblock_chroma_intra[0] = h->loopf.deblock_h_chroma_420_intra;
1132 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_chroma_420_mbaff;
1133 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_chroma_420_intra_mbaff;
1134 h->pixf.intra_mbcmp_x3_chroma = h->pixf.intra_mbcmp_x3_8x8c;
1135 h->quantf.coeff_last[DCT_CHROMA_DC] = h->quantf.coeff_last4;
1136 h->quantf.coeff_level_run[DCT_CHROMA_DC] = h->quantf.coeff_level_run4;
1139 memcpy( h->predict_chroma, h->predict_8x16c, sizeof(h->predict_chroma) );
1140 h->mc.prefetch_fenc = h->mc.prefetch_fenc_422;
1141 h->loopf.deblock_chroma[0] = h->loopf.deblock_h_chroma_422;
1142 h->loopf.deblock_chroma_intra[0] = h->loopf.deblock_h_chroma_422_intra;
1143 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_chroma_422_mbaff;
1144 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_chroma_422_intra_mbaff;
1145 h->pixf.intra_mbcmp_x3_chroma = h->pixf.intra_mbcmp_x3_8x16c;
1146 h->quantf.coeff_last[DCT_CHROMA_DC] = h->quantf.coeff_last8;
1147 h->quantf.coeff_level_run[DCT_CHROMA_DC] = h->quantf.coeff_level_run8;
1150 h->mc.prefetch_fenc = h->mc.prefetch_fenc_422; /* FIXME: doesn't cover V plane */
1151 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_luma_mbaff;
1152 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_luma_intra_mbaff;
1157 static void x264_set_aspect_ratio( x264_t *h, x264_param_t *param, int initial )
1160 if( param->vui.i_sar_width > 0 && param->vui.i_sar_height > 0 )
1162 uint32_t i_w = param->vui.i_sar_width;
1163 uint32_t i_h = param->vui.i_sar_height;
1164 uint32_t old_w = h->param.vui.i_sar_width;
1165 uint32_t old_h = h->param.vui.i_sar_height;
1167 x264_reduce_fraction( &i_w, &i_h );
1169 while( i_w > 65535 || i_h > 65535 )
1175 x264_reduce_fraction( &i_w, &i_h );
1177 if( i_w != old_w || i_h != old_h || initial )
1179 h->param.vui.i_sar_width = 0;
1180 h->param.vui.i_sar_height = 0;
1181 if( i_w == 0 || i_h == 0 )
1182 x264_log( h, X264_LOG_WARNING, "cannot create valid sample aspect ratio\n" );
1185 x264_log( h, initial?X264_LOG_INFO:X264_LOG_DEBUG, "using SAR=%d/%d\n", i_w, i_h );
1186 h->param.vui.i_sar_width = i_w;
1187 h->param.vui.i_sar_height = i_h;
1189 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1194 /****************************************************************************
1195 * x264_encoder_open:
1196 ****************************************************************************/
1197 x264_t *x264_encoder_open( x264_param_t *param )
1201 int qp, i_slicetype_length;
1203 CHECKED_MALLOCZERO( h, sizeof(x264_t) );
1205 /* Create a copy of param */
1206 memcpy( &h->param, param, sizeof(x264_param_t) );
1208 if( param->param_free )
1209 param->param_free( param );
1211 if( x264_threading_init() )
1213 x264_log( h, X264_LOG_ERROR, "unable to initialize threading\n" );
1217 if( x264_validate_parameters( h, 1 ) < 0 )
1220 if( h->param.psz_cqm_file )
1221 if( x264_cqm_parse_file( h, h->param.psz_cqm_file ) < 0 )
1224 if( h->param.rc.psz_stat_out )
1225 h->param.rc.psz_stat_out = strdup( h->param.rc.psz_stat_out );
1226 if( h->param.rc.psz_stat_in )
1227 h->param.rc.psz_stat_in = strdup( h->param.rc.psz_stat_in );
1229 x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den );
1230 x264_reduce_fraction( &h->param.i_timebase_num, &h->param.i_timebase_den );
1235 h->i_idr_pic_id = 0;
1237 if( (uint64_t)h->param.i_timebase_den * 2 > UINT32_MAX )
1239 x264_log( h, X264_LOG_ERROR, "Effective timebase denominator %u exceeds H.264 maximum\n", h->param.i_timebase_den );
1243 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1244 x264_pps_init( h->pps, h->param.i_sps_id, &h->param, h->sps );
1246 x264_set_aspect_ratio( h, &h->param, 1 );
1248 x264_validate_levels( h, 1 );
1250 h->chroma_qp_table = i_chroma_qp_table + 12 + h->pps->i_chroma_qp_index_offset;
1252 if( x264_cqm_init( h ) < 0 )
1255 h->mb.i_mb_width = h->sps->i_mb_width;
1256 h->mb.i_mb_height = h->sps->i_mb_height;
1257 h->mb.i_mb_count = h->mb.i_mb_width * h->mb.i_mb_height;
1259 h->mb.chroma_h_shift = CHROMA_FORMAT == CHROMA_420 || CHROMA_FORMAT == CHROMA_422;
1260 h->mb.chroma_v_shift = CHROMA_FORMAT == CHROMA_420;
1262 /* Adaptive MBAFF and subme 0 are not supported as we require halving motion
1263 * vectors during prediction, resulting in hpel mvs.
1264 * The chosen solution is to make MBAFF non-adaptive in this case. */
1265 h->mb.b_adaptive_mbaff = PARAM_INTERLACED && h->param.analyse.i_subpel_refine;
1268 if( h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS && !h->param.rc.b_stat_read )
1269 h->frames.i_delay = X264_MAX(h->param.i_bframe,3)*4;
1271 h->frames.i_delay = h->param.i_bframe;
1272 if( h->param.rc.b_mb_tree || h->param.rc.i_vbv_buffer_size )
1273 h->frames.i_delay = X264_MAX( h->frames.i_delay, h->param.rc.i_lookahead );
1274 i_slicetype_length = h->frames.i_delay;
1275 h->frames.i_delay += h->i_thread_frames - 1;
1276 h->frames.i_delay += h->param.i_sync_lookahead;
1277 h->frames.i_delay += h->param.b_vfr_input;
1278 h->frames.i_bframe_delay = h->param.i_bframe ? (h->param.i_bframe_pyramid ? 2 : 1) : 0;
1280 h->frames.i_max_ref0 = h->param.i_frame_reference;
1281 h->frames.i_max_ref1 = X264_MIN( h->sps->vui.i_num_reorder_frames, h->param.i_frame_reference );
1282 h->frames.i_max_dpb = h->sps->vui.i_max_dec_frame_buffering;
1283 h->frames.b_have_lowres = !h->param.rc.b_stat_read
1284 && ( h->param.rc.i_rc_method == X264_RC_ABR
1285 || h->param.rc.i_rc_method == X264_RC_CRF
1286 || h->param.i_bframe_adaptive
1287 || h->param.i_scenecut_threshold
1288 || h->param.rc.b_mb_tree
1289 || h->param.analyse.i_weighted_pred );
1290 h->frames.b_have_lowres |= h->param.rc.b_stat_read && h->param.rc.i_vbv_buffer_size > 0;
1291 h->frames.b_have_sub8x8_esa = !!(h->param.analyse.inter & X264_ANALYSE_PSUB8x8);
1293 h->frames.i_last_idr =
1294 h->frames.i_last_keyframe = - h->param.i_keyint_max;
1295 h->frames.i_input = 0;
1296 h->frames.i_largest_pts = h->frames.i_second_largest_pts = -1;
1297 h->frames.i_poc_last_open_gop = -1;
1299 CHECKED_MALLOCZERO( h->frames.unused[0], (h->frames.i_delay + 3) * sizeof(x264_frame_t *) );
1300 /* Allocate room for max refs plus a few extra just in case. */
1301 CHECKED_MALLOCZERO( h->frames.unused[1], (h->i_thread_frames + X264_REF_MAX + 4) * sizeof(x264_frame_t *) );
1302 CHECKED_MALLOCZERO( h->frames.current, (h->param.i_sync_lookahead + h->param.i_bframe
1303 + h->i_thread_frames + 3) * sizeof(x264_frame_t *) );
1304 if( h->param.analyse.i_weighted_pred > 0 )
1305 CHECKED_MALLOCZERO( h->frames.blank_unused, h->i_thread_frames * 4 * sizeof(x264_frame_t *) );
1306 h->i_ref[0] = h->i_ref[1] = 0;
1307 h->i_cpb_delay = h->i_coded_fields = h->i_disp_fields = 0;
1308 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);
1309 h->i_disp_fields_last_frame = -1;
1312 /* init CPU functions */
1313 x264_predict_16x16_init( h->param.cpu, h->predict_16x16 );
1314 x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c );
1315 x264_predict_8x16c_init( h->param.cpu, h->predict_8x16c );
1316 x264_predict_8x8_init( h->param.cpu, h->predict_8x8, &h->predict_8x8_filter );
1317 x264_predict_4x4_init( h->param.cpu, h->predict_4x4 );
1318 x264_pixel_init( h->param.cpu, &h->pixf );
1319 x264_dct_init( h->param.cpu, &h->dctf );
1320 x264_zigzag_init( h->param.cpu, &h->zigzagf_progressive, &h->zigzagf_interlaced );
1321 memcpy( &h->zigzagf, PARAM_INTERLACED ? &h->zigzagf_interlaced : &h->zigzagf_progressive, sizeof(h->zigzagf) );
1322 x264_mc_init( h->param.cpu, &h->mc, h->param.b_cpu_independent );
1323 x264_quant_init( h, h->param.cpu, &h->quantf );
1324 x264_deblock_init( h->param.cpu, &h->loopf, PARAM_INTERLACED );
1325 x264_bitstream_init( h->param.cpu, &h->bsf );
1326 if( h->param.b_cabac )
1327 x264_cabac_init( h );
1329 x264_stack_align( x264_cavlc_init, h );
1332 chroma_dsp_init( h );
1334 p = buf + sprintf( buf, "using cpu capabilities:" );
1335 for( int i = 0; x264_cpu_names[i].flags; i++ )
1337 if( !strcmp(x264_cpu_names[i].name, "SSE")
1338 && h->param.cpu & (X264_CPU_SSE2) )
1340 if( !strcmp(x264_cpu_names[i].name, "SSE2")
1341 && h->param.cpu & (X264_CPU_SSE2_IS_FAST|X264_CPU_SSE2_IS_SLOW) )
1343 if( !strcmp(x264_cpu_names[i].name, "SSE3")
1344 && (h->param.cpu & X264_CPU_SSSE3 || !(h->param.cpu & X264_CPU_CACHELINE_64)) )
1346 if( !strcmp(x264_cpu_names[i].name, "SSE4.1")
1347 && (h->param.cpu & X264_CPU_SSE42) )
1349 if( !strcmp(x264_cpu_names[i].name, "BMI1")
1350 && (h->param.cpu & X264_CPU_BMI2) )
1352 if( (h->param.cpu & x264_cpu_names[i].flags) == x264_cpu_names[i].flags
1353 && (!i || x264_cpu_names[i].flags != x264_cpu_names[i-1].flags) )
1354 p += sprintf( p, " %s", x264_cpu_names[i].name );
1357 p += sprintf( p, " none!" );
1358 x264_log( h, X264_LOG_INFO, "%s\n", buf );
1360 float *logs = x264_analyse_prepare_costs( h );
1363 for( qp = X264_MIN( h->param.rc.i_qp_min, QP_MAX_SPEC ); qp <= h->param.rc.i_qp_max; qp++ )
1364 if( x264_analyse_init_costs( h, logs, qp ) )
1366 if( x264_analyse_init_costs( h, logs, X264_LOOKAHEAD_QP ) )
1370 static const uint16_t cost_mv_correct[7] = { 24, 47, 95, 189, 379, 757, 1515 };
1371 /* Checks for known miscompilation issues. */
1372 if( h->cost_mv[X264_LOOKAHEAD_QP][2013] != cost_mv_correct[BIT_DEPTH-8] )
1374 x264_log( h, X264_LOG_ERROR, "MV cost test failed: x264 has been miscompiled!\n" );
1378 /* Must be volatile or else GCC will optimize it out. */
1379 volatile int temp = 392;
1380 if( x264_clz( temp ) != 23 )
1382 x264_log( h, X264_LOG_ERROR, "CLZ test failed: x264 has been miscompiled!\n" );
1383 #if ARCH_X86 || ARCH_X86_64
1384 x264_log( h, X264_LOG_ERROR, "Are you attempting to run an SSE4a/LZCNT-targeted build on a CPU that\n" );
1385 x264_log( h, X264_LOG_ERROR, "doesn't support it?\n" );
1391 h->out.i_bitstream = X264_MAX( 1000000, h->param.i_width * h->param.i_height * 4
1392 * ( h->param.rc.i_rc_method == X264_RC_ABR ? pow( 0.95, h->param.rc.i_qp_min )
1393 : pow( 0.95, h->param.rc.i_qp_constant ) * X264_MAX( 1, h->param.rc.f_ip_factor )));
1395 h->nal_buffer_size = h->out.i_bitstream * 3/2 + 4 + 64; /* +4 for startcode, +64 for nal_escape assembly padding */
1396 CHECKED_MALLOC( h->nal_buffer, h->nal_buffer_size );
1398 if( h->param.i_threads > 1 &&
1399 x264_threadpool_init( &h->threadpool, h->param.i_threads, (void*)x264_encoder_thread_init, h ) )
1401 if( h->param.i_lookahead_threads > 1 &&
1402 x264_threadpool_init( &h->lookaheadpool, h->param.i_lookahead_threads, (void*)x264_lookahead_thread_init, h ) )
1406 if( h->param.b_opencl )
1408 h->opencl.ocl = x264_opencl_load_library();
1409 if( !h->opencl.ocl )
1411 x264_log( h, X264_LOG_WARNING, "failed to load OpenCL\n" );
1412 h->param.b_opencl = 0;
1418 for( int i = 1; i < h->param.i_threads + !!h->param.i_sync_lookahead; i++ )
1419 CHECKED_MALLOC( h->thread[i], sizeof(x264_t) );
1420 if( h->param.i_lookahead_threads > 1 )
1421 for( int i = 0; i < h->param.i_lookahead_threads; i++ )
1423 CHECKED_MALLOC( h->lookahead_thread[i], sizeof(x264_t) );
1424 *h->lookahead_thread[i] = *h;
1427 for( int i = 0; i < h->param.i_threads; i++ )
1429 int init_nal_count = h->param.i_slice_count + 3;
1430 int allocate_threadlocal_data = !h->param.b_sliced_threads || !i;
1434 if( x264_pthread_mutex_init( &h->thread[i]->mutex, NULL ) )
1436 if( x264_pthread_cond_init( &h->thread[i]->cv, NULL ) )
1439 if( allocate_threadlocal_data )
1441 h->thread[i]->fdec = x264_frame_pop_unused( h, 1 );
1442 if( !h->thread[i]->fdec )
1446 h->thread[i]->fdec = h->thread[0]->fdec;
1448 CHECKED_MALLOC( h->thread[i]->out.p_bitstream, h->out.i_bitstream );
1449 /* Start each thread with room for init_nal_count NAL units; it'll realloc later if needed. */
1450 CHECKED_MALLOC( h->thread[i]->out.nal, init_nal_count*sizeof(x264_nal_t) );
1451 h->thread[i]->out.i_nals_allocated = init_nal_count;
1453 if( allocate_threadlocal_data && x264_macroblock_cache_allocate( h->thread[i] ) < 0 )
1458 if( h->param.b_opencl && x264_opencl_lookahead_init( h ) < 0 )
1459 h->param.b_opencl = 0;
1462 if( x264_lookahead_init( h, i_slicetype_length ) )
1465 for( int i = 0; i < h->param.i_threads; i++ )
1466 if( x264_macroblock_thread_allocate( h->thread[i], 0 ) < 0 )
1469 if( x264_ratecontrol_new( h ) < 0 )
1472 if( h->param.i_nal_hrd )
1474 x264_log( h, X264_LOG_DEBUG, "HRD bitrate: %i bits/sec\n", h->sps->vui.hrd.i_bit_rate_unscaled );
1475 x264_log( h, X264_LOG_DEBUG, "CPB size: %i bits\n", h->sps->vui.hrd.i_cpb_size_unscaled );
1478 if( h->param.psz_dump_yuv )
1480 /* create or truncate the reconstructed video file */
1481 FILE *f = fopen( h->param.psz_dump_yuv, "w" );
1484 x264_log( h, X264_LOG_ERROR, "dump_yuv: can't write to %s\n", h->param.psz_dump_yuv );
1487 else if( !x264_is_regular_file( f ) )
1489 x264_log( h, X264_LOG_ERROR, "dump_yuv: incompatible with non-regular file %s\n", h->param.psz_dump_yuv );
1495 const char *profile = h->sps->i_profile_idc == PROFILE_BASELINE ? "Constrained Baseline" :
1496 h->sps->i_profile_idc == PROFILE_MAIN ? "Main" :
1497 h->sps->i_profile_idc == PROFILE_HIGH ? "High" :
1498 h->sps->i_profile_idc == PROFILE_HIGH10 ? (h->sps->b_constraint_set3 == 1 ? "High 10 Intra" : "High 10") :
1499 h->sps->i_profile_idc == PROFILE_HIGH422 ? (h->sps->b_constraint_set3 == 1 ? "High 4:2:2 Intra" : "High 4:2:2") :
1500 h->sps->b_constraint_set3 == 1 ? "High 4:4:4 Intra" : "High 4:4:4 Predictive";
1502 snprintf( level, sizeof(level), "%d.%d", h->sps->i_level_idc/10, h->sps->i_level_idc%10 );
1503 if( h->sps->i_level_idc == 9 || ( h->sps->i_level_idc == 11 && h->sps->b_constraint_set3 &&
1504 (h->sps->i_profile_idc == PROFILE_BASELINE || h->sps->i_profile_idc == PROFILE_MAIN) ) )
1505 strcpy( level, "1b" );
1507 if( h->sps->i_profile_idc < PROFILE_HIGH10 )
1509 x264_log( h, X264_LOG_INFO, "profile %s, level %s\n",
1514 static const char * const subsampling[4] = { "4:0:0", "4:2:0", "4:2:2", "4:4:4" };
1515 x264_log( h, X264_LOG_INFO, "profile %s, level %s, %s %d-bit\n",
1516 profile, level, subsampling[CHROMA_FORMAT], BIT_DEPTH );
1525 /****************************************************************************
1526 * x264_encoder_reconfig:
1527 ****************************************************************************/
1528 int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
1530 /* If the previous frame isn't done encoding, reconfiguring is probably dangerous. */
1531 if( h->param.b_sliced_threads )
1532 if( x264_threadpool_wait_all( h ) < 0 )
1535 int rc_reconfig = 0;
1536 h = h->thread[h->thread[0]->i_thread_phase];
1537 x264_set_aspect_ratio( h, param, 0 );
1538 #define COPY(var) h->param.var = param->var
1539 COPY( i_frame_reference ); // but never uses more refs than initially specified
1540 COPY( i_bframe_bias );
1541 if( h->param.i_scenecut_threshold )
1542 COPY( i_scenecut_threshold ); // can't turn it on or off, only vary the threshold
1543 COPY( b_deblocking_filter );
1544 COPY( i_deblocking_filter_alphac0 );
1545 COPY( i_deblocking_filter_beta );
1546 COPY( i_frame_packing );
1547 COPY( analyse.inter );
1548 COPY( analyse.intra );
1549 COPY( analyse.i_direct_mv_pred );
1550 /* Scratch buffer prevents me_range from being increased for esa/tesa */
1551 if( h->param.analyse.i_me_method < X264_ME_ESA || param->analyse.i_me_range < h->param.analyse.i_me_range )
1552 COPY( analyse.i_me_range );
1553 COPY( analyse.i_noise_reduction );
1554 /* We can't switch out of subme=0 during encoding. */
1555 if( h->param.analyse.i_subpel_refine )
1556 COPY( analyse.i_subpel_refine );
1557 COPY( analyse.i_trellis );
1558 COPY( analyse.b_chroma_me );
1559 COPY( analyse.b_dct_decimate );
1560 COPY( analyse.b_fast_pskip );
1561 COPY( analyse.b_mixed_references );
1562 COPY( analyse.f_psy_rd );
1563 COPY( analyse.f_psy_trellis );
1565 // can only twiddle these if they were enabled to begin with:
1566 if( h->param.analyse.i_me_method >= X264_ME_ESA || param->analyse.i_me_method < X264_ME_ESA )
1567 COPY( analyse.i_me_method );
1568 if( h->param.analyse.i_me_method >= X264_ME_ESA && !h->frames.b_have_sub8x8_esa )
1569 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
1570 if( h->pps->b_transform_8x8_mode )
1571 COPY( analyse.b_transform_8x8 );
1572 if( h->frames.i_max_ref1 > 1 )
1573 COPY( i_bframe_pyramid );
1574 COPY( i_slice_max_size );
1575 COPY( i_slice_max_mbs );
1576 COPY( i_slice_min_mbs );
1577 COPY( i_slice_count );
1578 COPY( i_slice_count_max );
1581 /* VBV can't be turned on if it wasn't on to begin with */
1582 if( h->param.rc.i_vbv_max_bitrate > 0 && h->param.rc.i_vbv_buffer_size > 0 &&
1583 param->rc.i_vbv_max_bitrate > 0 && param->rc.i_vbv_buffer_size > 0 )
1585 rc_reconfig |= h->param.rc.i_vbv_max_bitrate != param->rc.i_vbv_max_bitrate;
1586 rc_reconfig |= h->param.rc.i_vbv_buffer_size != param->rc.i_vbv_buffer_size;
1587 rc_reconfig |= h->param.rc.i_bitrate != param->rc.i_bitrate;
1588 COPY( rc.i_vbv_max_bitrate );
1589 COPY( rc.i_vbv_buffer_size );
1590 COPY( rc.i_bitrate );
1592 rc_reconfig |= h->param.rc.f_rf_constant != param->rc.f_rf_constant;
1593 rc_reconfig |= h->param.rc.f_rf_constant_max != param->rc.f_rf_constant_max;
1594 COPY( rc.f_rf_constant );
1595 COPY( rc.f_rf_constant_max );
1600 int ret = x264_validate_parameters( h, 0 );
1602 /* Supported reconfiguration options (1-pass only):
1606 * bitrate (CBR only) */
1607 if( !ret && rc_reconfig )
1608 x264_ratecontrol_init_reconfigurable( h, 0 );
1613 /****************************************************************************
1614 * x264_encoder_parameters:
1615 ****************************************************************************/
1616 void x264_encoder_parameters( x264_t *h, x264_param_t *param )
1618 memcpy( param, &h->thread[h->i_thread_phase]->param, sizeof(x264_param_t) );
1621 /* internal usage */
1622 static void x264_nal_start( x264_t *h, int i_type, int i_ref_idc )
1624 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1626 nal->i_ref_idc = i_ref_idc;
1627 nal->i_type = i_type;
1628 nal->b_long_startcode = 1;
1631 nal->p_payload= &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1634 /* if number of allocated nals is not enough, re-allocate a larger one. */
1635 static int x264_nal_check_buffer( x264_t *h )
1637 if( h->out.i_nal >= h->out.i_nals_allocated )
1639 x264_nal_t *new_out = x264_malloc( sizeof(x264_nal_t) * (h->out.i_nals_allocated*2) );
1642 memcpy( new_out, h->out.nal, sizeof(x264_nal_t) * (h->out.i_nals_allocated) );
1643 x264_free( h->out.nal );
1644 h->out.nal = new_out;
1645 h->out.i_nals_allocated *= 2;
1650 static int x264_nal_end( x264_t *h )
1652 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1653 uint8_t *end = &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1654 nal->i_payload = end - nal->p_payload;
1655 /* Assembly implementation of nal_escape reads past the end of the input.
1656 * While undefined padding wouldn't actually affect the output, it makes valgrind unhappy. */
1657 memset( end, 0xff, 64 );
1658 if( h->param.nalu_process )
1659 h->param.nalu_process( h, nal, h->fenc->opaque );
1662 return x264_nal_check_buffer( h );
1665 static int x264_encoder_encapsulate_nals( x264_t *h, int start )
1667 x264_t *h0 = h->thread[0];
1668 int nal_size = 0, previous_nal_size = 0;
1670 if( h->param.nalu_process )
1672 for( int i = start; i < h->out.i_nal; i++ )
1673 nal_size += h->out.nal[i].i_payload;
1677 for( int i = 0; i < start; i++ )
1678 previous_nal_size += h->out.nal[i].i_payload;
1680 for( int i = start; i < h->out.i_nal; i++ )
1681 nal_size += h->out.nal[i].i_payload;
1683 /* Worst-case NAL unit escaping: reallocate the buffer if it's too small. */
1684 int necessary_size = previous_nal_size + nal_size * 3/2 + h->out.i_nal * 4 + 4 + 64;
1685 if( h0->nal_buffer_size < necessary_size )
1687 necessary_size *= 2;
1688 uint8_t *buf = x264_malloc( necessary_size );
1691 if( previous_nal_size )
1692 memcpy( buf, h0->nal_buffer, previous_nal_size );
1694 intptr_t delta = buf - h0->nal_buffer;
1695 for( int i = 0; i < start; i++ )
1696 h->out.nal[i].p_payload += delta;
1698 x264_free( h0->nal_buffer );
1699 h0->nal_buffer = buf;
1700 h0->nal_buffer_size = necessary_size;
1703 uint8_t *nal_buffer = h0->nal_buffer + previous_nal_size;
1705 for( int i = start; i < h->out.i_nal; i++ )
1707 h->out.nal[i].b_long_startcode = !i || h->out.nal[i].i_type == NAL_SPS || h->out.nal[i].i_type == NAL_PPS;
1708 x264_nal_encode( h, nal_buffer, &h->out.nal[i] );
1709 nal_buffer += h->out.nal[i].i_payload;
1714 return nal_buffer - (h0->nal_buffer + previous_nal_size);
1717 /****************************************************************************
1718 * x264_encoder_headers:
1719 ****************************************************************************/
1720 int x264_encoder_headers( x264_t *h, x264_nal_t **pp_nal, int *pi_nal )
1723 /* init bitstream context */
1725 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
1727 /* Write SEI, SPS and PPS. */
1729 /* generate sequence parameters */
1730 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
1731 x264_sps_write( &h->out.bs, h->sps );
1732 if( x264_nal_end( h ) )
1735 /* generate picture parameters */
1736 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
1737 x264_pps_write( &h->out.bs, h->sps, h->pps );
1738 if( x264_nal_end( h ) )
1741 /* identify ourselves */
1742 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
1743 if( x264_sei_version_write( h, &h->out.bs ) )
1745 if( x264_nal_end( h ) )
1748 frame_size = x264_encoder_encapsulate_nals( h, 0 );
1749 if( frame_size < 0 )
1753 *pi_nal = h->out.i_nal;
1754 *pp_nal = &h->out.nal[0];
1760 /* Check to see whether we have chosen a reference list ordering different
1761 * from the standard's default. */
1762 static inline void x264_reference_check_reorder( x264_t *h )
1764 /* The reorder check doesn't check for missing frames, so just
1765 * force a reorder if one of the reference list is corrupt. */
1766 for( int i = 0; h->frames.reference[i]; i++ )
1767 if( h->frames.reference[i]->b_corrupt )
1769 h->b_ref_reorder[0] = 1;
1772 for( int list = 0; list <= (h->sh.i_type == SLICE_TYPE_B); list++ )
1773 for( int i = 0; i < h->i_ref[list] - 1; i++ )
1775 int framenum_diff = h->fref[list][i+1]->i_frame_num - h->fref[list][i]->i_frame_num;
1776 int poc_diff = h->fref[list][i+1]->i_poc - h->fref[list][i]->i_poc;
1777 /* P and B-frames use different default orders. */
1778 if( h->sh.i_type == SLICE_TYPE_P ? framenum_diff > 0 : list == 1 ? poc_diff < 0 : poc_diff > 0 )
1780 h->b_ref_reorder[list] = 1;
1786 /* return -1 on failure, else return the index of the new reference frame */
1787 int x264_weighted_reference_duplicate( x264_t *h, int i_ref, const x264_weight_t *w )
1789 int i = h->i_ref[0];
1791 x264_frame_t *newframe;
1792 if( i <= 1 ) /* empty list, definitely can't duplicate frame */
1795 //Duplication is only used in X264_WEIGHTP_SMART
1796 if( h->param.analyse.i_weighted_pred != X264_WEIGHTP_SMART )
1799 /* Duplication is a hack to compensate for crappy rounding in motion compensation.
1800 * With high bit depth, it's not worth doing, so turn it off except in the case of
1801 * unweighted dupes. */
1802 if( BIT_DEPTH > 8 && w != x264_weight_none )
1805 newframe = x264_frame_pop_blank_unused( h );
1809 //FIXME: probably don't need to copy everything
1810 *newframe = *h->fref[0][i_ref];
1811 newframe->i_reference_count = 1;
1812 newframe->orig = h->fref[0][i_ref];
1813 newframe->b_duplicate = 1;
1814 memcpy( h->fenc->weight[j], w, sizeof(h->fenc->weight[i]) );
1816 /* shift the frames to make space for the dupe. */
1817 h->b_ref_reorder[0] = 1;
1818 if( h->i_ref[0] < X264_REF_MAX )
1820 h->fref[0][X264_REF_MAX-1] = NULL;
1821 x264_frame_unshift( &h->fref[0][j], newframe );
1826 static void x264_weighted_pred_init( x264_t *h )
1828 /* for now no analysis and set all weights to nothing */
1829 for( int i_ref = 0; i_ref < h->i_ref[0]; i_ref++ )
1830 h->fenc->weighted[i_ref] = h->fref[0][i_ref]->filtered[0][0];
1832 // FIXME: This only supports weighting of one reference frame
1833 // and duplicates of that frame.
1834 h->fenc->i_lines_weighted = 0;
1836 for( int i_ref = 0; i_ref < (h->i_ref[0] << SLICE_MBAFF); i_ref++ )
1837 for( int i = 0; i < 3; i++ )
1838 h->sh.weight[i_ref][i].weightfn = NULL;
1841 if( h->sh.i_type != SLICE_TYPE_P || h->param.analyse.i_weighted_pred <= 0 )
1844 int i_padv = PADV << PARAM_INTERLACED;
1846 int weightplane[2] = { 0, 0 };
1847 int buffer_next = 0;
1848 for( int i = 0; i < 3; i++ )
1850 for( int j = 0; j < h->i_ref[0]; j++ )
1852 if( h->fenc->weight[j][i].weightfn )
1854 h->sh.weight[j][i] = h->fenc->weight[j][i];
1855 // if weight is useless, don't write it to stream
1856 if( h->sh.weight[j][i].i_scale == 1<<h->sh.weight[j][i].i_denom && h->sh.weight[j][i].i_offset == 0 )
1857 h->sh.weight[j][i].weightfn = NULL;
1860 if( !weightplane[!!i] )
1862 weightplane[!!i] = 1;
1863 h->sh.weight[0][!!i].i_denom = denom = h->sh.weight[j][i].i_denom;
1864 assert( x264_clip3( denom, 0, 7 ) == denom );
1867 assert( h->sh.weight[j][i].i_denom == denom );
1870 h->fenc->weighted[j] = h->mb.p_weight_buf[buffer_next++] + h->fenc->i_stride[0] * i_padv + PADH;
1871 //scale full resolution frame
1872 if( h->param.i_threads == 1 )
1874 pixel *src = h->fref[0][j]->filtered[0][0] - h->fref[0][j]->i_stride[0]*i_padv - PADH;
1875 pixel *dst = h->fenc->weighted[j] - h->fenc->i_stride[0]*i_padv - PADH;
1876 int stride = h->fenc->i_stride[0];
1877 int width = h->fenc->i_width[0] + PADH*2;
1878 int height = h->fenc->i_lines[0] + i_padv*2;
1879 x264_weight_scale_plane( h, dst, stride, src, stride, width, height, &h->sh.weight[j][0] );
1880 h->fenc->i_lines_weighted = height;
1888 if( weightplane[1] )
1889 for( int i = 0; i < h->i_ref[0]; i++ )
1891 if( h->sh.weight[i][1].weightfn && !h->sh.weight[i][2].weightfn )
1893 h->sh.weight[i][2].i_scale = 1 << h->sh.weight[0][1].i_denom;
1894 h->sh.weight[i][2].i_offset = 0;
1896 else if( h->sh.weight[i][2].weightfn && !h->sh.weight[i][1].weightfn )
1898 h->sh.weight[i][1].i_scale = 1 << h->sh.weight[0][1].i_denom;
1899 h->sh.weight[i][1].i_offset = 0;
1903 if( !weightplane[0] )
1904 h->sh.weight[0][0].i_denom = 0;
1905 if( !weightplane[1] )
1906 h->sh.weight[0][1].i_denom = 0;
1907 h->sh.weight[0][2].i_denom = h->sh.weight[0][1].i_denom;
1910 static inline int x264_reference_distance( x264_t *h, x264_frame_t *frame )
1912 if( h->param.i_frame_packing == 5 )
1913 return abs((h->fenc->i_frame&~1) - (frame->i_frame&~1)) +
1914 ((h->fenc->i_frame&1) != (frame->i_frame&1));
1916 return abs(h->fenc->i_frame - frame->i_frame);
1919 static inline void x264_reference_build_list( x264_t *h, int i_poc )
1923 /* build ref list 0/1 */
1924 h->mb.pic.i_fref[0] = h->i_ref[0] = 0;
1925 h->mb.pic.i_fref[1] = h->i_ref[1] = 0;
1926 if( h->sh.i_type == SLICE_TYPE_I )
1929 for( int i = 0; h->frames.reference[i]; i++ )
1931 if( h->frames.reference[i]->b_corrupt )
1933 if( h->frames.reference[i]->i_poc < i_poc )
1934 h->fref[0][h->i_ref[0]++] = h->frames.reference[i];
1935 else if( h->frames.reference[i]->i_poc > i_poc )
1936 h->fref[1][h->i_ref[1]++] = h->frames.reference[i];
1939 /* Order reference lists by distance from the current frame. */
1940 for( int list = 0; list < 2; list++ )
1942 h->fref_nearest[list] = h->fref[list][0];
1946 for( int i = 0; i < h->i_ref[list] - 1; i++ )
1948 if( list ? h->fref[list][i+1]->i_poc < h->fref_nearest[list]->i_poc
1949 : h->fref[list][i+1]->i_poc > h->fref_nearest[list]->i_poc )
1950 h->fref_nearest[list] = h->fref[list][i+1];
1951 if( x264_reference_distance( h, h->fref[list][i] ) > x264_reference_distance( h, h->fref[list][i+1] ) )
1953 XCHG( x264_frame_t*, h->fref[list][i], h->fref[list][i+1] );
1961 if( h->sh.i_mmco_remove_from_end )
1962 for( int i = h->i_ref[0]-1; i >= h->i_ref[0] - h->sh.i_mmco_remove_from_end; i-- )
1964 int diff = h->i_frame_num - h->fref[0][i]->i_frame_num;
1965 h->sh.mmco[h->sh.i_mmco_command_count].i_poc = h->fref[0][i]->i_poc;
1966 h->sh.mmco[h->sh.i_mmco_command_count++].i_difference_of_pic_nums = diff;
1969 x264_reference_check_reorder( h );
1971 h->i_ref[1] = X264_MIN( h->i_ref[1], h->frames.i_max_ref1 );
1972 h->i_ref[0] = X264_MIN( h->i_ref[0], h->frames.i_max_ref0 );
1973 h->i_ref[0] = X264_MIN( h->i_ref[0], h->param.i_frame_reference ); // if reconfig() has lowered the limit
1975 /* For Blu-ray compliance, don't reference frames outside of the minigop. */
1976 if( IS_X264_TYPE_B( h->fenc->i_type ) && h->param.b_bluray_compat )
1977 h->i_ref[0] = X264_MIN( h->i_ref[0], IS_X264_TYPE_B( h->fref[0][0]->i_type ) + 1 );
1979 /* add duplicates */
1980 if( h->fenc->i_type == X264_TYPE_P )
1983 if( h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE )
1986 w[1].weightfn = w[2].weightfn = NULL;
1987 if( h->param.rc.b_stat_read )
1988 x264_ratecontrol_set_weights( h, h->fenc );
1990 if( !h->fenc->weight[0][0].weightfn )
1992 h->fenc->weight[0][0].i_denom = 0;
1993 SET_WEIGHT( w[0], 1, 1, 0, -1 );
1994 idx = x264_weighted_reference_duplicate( h, 0, w );
1998 if( h->fenc->weight[0][0].i_scale == 1<<h->fenc->weight[0][0].i_denom )
2000 SET_WEIGHT( h->fenc->weight[0][0], 1, 1, 0, h->fenc->weight[0][0].i_offset );
2002 x264_weighted_reference_duplicate( h, 0, x264_weight_none );
2003 if( h->fenc->weight[0][0].i_offset > -128 )
2005 w[0] = h->fenc->weight[0][0];
2007 h->mc.weight_cache( h, &w[0] );
2008 idx = x264_weighted_reference_duplicate( h, 0, w );
2012 h->mb.ref_blind_dupe = idx;
2015 assert( h->i_ref[0] + h->i_ref[1] <= X264_REF_MAX );
2016 h->mb.pic.i_fref[0] = h->i_ref[0];
2017 h->mb.pic.i_fref[1] = h->i_ref[1];
2020 static void x264_fdec_filter_row( x264_t *h, int mb_y, int pass )
2022 /* mb_y is the mb to be encoded next, not the mb to be filtered here */
2023 int b_hpel = h->fdec->b_kept_as_ref;
2024 int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
2025 int b_end = mb_y == h->i_threadslice_end;
2026 int b_measure_quality = 1;
2027 int min_y = mb_y - (1 << SLICE_MBAFF);
2028 int b_start = min_y == h->i_threadslice_start;
2029 /* Even in interlaced mode, deblocking never modifies more than 4 pixels
2030 * above each MB, as bS=4 doesn't happen for the top of interlaced mbpairs. */
2031 int minpix_y = min_y*16 - 4 * !b_start;
2032 int maxpix_y = mb_y*16 - 4 * !b_end;
2033 b_deblock &= b_hpel || h->param.b_full_recon || h->param.psz_dump_yuv;
2034 if( h->param.b_sliced_threads )
2038 /* During encode: only do deblock if asked for */
2041 b_deblock &= h->param.b_full_recon;
2044 /* During post-encode pass: do deblock if not done yet, do hpel for all
2045 * rows except those between slices. */
2047 b_deblock &= !h->param.b_full_recon;
2048 b_hpel &= !(b_start && min_y > 0);
2049 b_measure_quality = 0;
2051 /* Final pass: do the rows between slices in sequence. */
2054 b_measure_quality = 0;
2058 if( mb_y & SLICE_MBAFF )
2060 if( min_y < h->i_threadslice_start )
2064 for( int y = min_y; y < mb_y; y += (1 << SLICE_MBAFF) )
2065 x264_frame_deblock_row( h, y );
2067 /* FIXME: Prediction requires different borders for interlaced/progressive mc,
2068 * but the actual image data is equivalent. For now, maintain this
2069 * consistency by copying deblocked pixels between planes. */
2070 if( PARAM_INTERLACED && (!h->param.b_sliced_threads || pass == 1) )
2071 for( int p = 0; p < h->fdec->i_plane; p++ )
2072 for( int i = minpix_y>>(CHROMA_V_SHIFT && p); i < maxpix_y>>(CHROMA_V_SHIFT && p); i++ )
2073 memcpy( h->fdec->plane_fld[p] + i*h->fdec->i_stride[p],
2074 h->fdec->plane[p] + i*h->fdec->i_stride[p],
2075 h->mb.i_mb_width*16*sizeof(pixel) );
2077 if( h->fdec->b_kept_as_ref && (!h->param.b_sliced_threads || pass == 1) )
2078 x264_frame_expand_border( h, h->fdec, min_y );
2081 int end = mb_y == h->mb.i_mb_height;
2082 /* Can't do hpel until the previous slice is done encoding. */
2083 if( h->param.analyse.i_subpel_refine )
2085 x264_frame_filter( h, h->fdec, min_y, end );
2086 x264_frame_expand_border_filtered( h, h->fdec, min_y, end );
2090 if( SLICE_MBAFF && pass == 0 )
2091 for( int i = 0; i < 3; i++ )
2093 XCHG( pixel *, h->intra_border_backup[0][i], h->intra_border_backup[3][i] );
2094 XCHG( pixel *, h->intra_border_backup[1][i], h->intra_border_backup[4][i] );
2097 if( h->i_thread_frames > 1 && h->fdec->b_kept_as_ref )
2098 x264_frame_cond_broadcast( h->fdec, mb_y*16 + (b_end ? 10000 : -(X264_THREAD_HEIGHT << SLICE_MBAFF)) );
2100 if( b_measure_quality )
2102 maxpix_y = X264_MIN( maxpix_y, h->param.i_height );
2103 if( h->param.analyse.b_psnr )
2105 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
2106 h->stat.frame.i_ssd[p] += x264_pixel_ssd_wxh( &h->pixf,
2107 h->fdec->plane[p] + minpix_y * h->fdec->i_stride[p], h->fdec->i_stride[p],
2108 h->fenc->plane[p] + minpix_y * h->fenc->i_stride[p], h->fenc->i_stride[p],
2109 h->param.i_width, maxpix_y-minpix_y );
2112 uint64_t ssd_u, ssd_v;
2113 int v_shift = CHROMA_V_SHIFT;
2114 x264_pixel_ssd_nv12( &h->pixf,
2115 h->fdec->plane[1] + (minpix_y>>v_shift) * h->fdec->i_stride[1], h->fdec->i_stride[1],
2116 h->fenc->plane[1] + (minpix_y>>v_shift) * h->fenc->i_stride[1], h->fenc->i_stride[1],
2117 h->param.i_width>>1, (maxpix_y-minpix_y)>>v_shift, &ssd_u, &ssd_v );
2118 h->stat.frame.i_ssd[1] += ssd_u;
2119 h->stat.frame.i_ssd[2] += ssd_v;
2123 if( h->param.analyse.b_ssim )
2127 /* offset by 2 pixels to avoid alignment of ssim blocks with dct blocks,
2128 * and overlap by 4 */
2129 minpix_y += b_start ? 2 : -6;
2130 h->stat.frame.f_ssim +=
2131 x264_pixel_ssim_wxh( &h->pixf,
2132 h->fdec->plane[0] + 2+minpix_y*h->fdec->i_stride[0], h->fdec->i_stride[0],
2133 h->fenc->plane[0] + 2+minpix_y*h->fenc->i_stride[0], h->fenc->i_stride[0],
2134 h->param.i_width-2, maxpix_y-minpix_y, h->scratch_buffer, &ssim_cnt );
2135 h->stat.frame.i_ssim_cnt += ssim_cnt;
2140 static inline int x264_reference_update( x264_t *h )
2142 if( !h->fdec->b_kept_as_ref )
2144 if( h->i_thread_frames > 1 )
2146 x264_frame_push_unused( h, h->fdec );
2147 h->fdec = x264_frame_pop_unused( h, 1 );
2154 /* apply mmco from previous frame. */
2155 for( int i = 0; i < h->sh.i_mmco_command_count; i++ )
2156 for( int j = 0; h->frames.reference[j]; j++ )
2157 if( h->frames.reference[j]->i_poc == h->sh.mmco[i].i_poc )
2158 x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[j] ) );
2160 /* move frame in the buffer */
2161 x264_frame_push( h->frames.reference, h->fdec );
2162 if( h->frames.reference[h->sps->i_num_ref_frames] )
2163 x264_frame_push_unused( h, x264_frame_shift( h->frames.reference ) );
2164 h->fdec = x264_frame_pop_unused( h, 1 );
2170 static inline void x264_reference_reset( x264_t *h )
2172 while( h->frames.reference[0] )
2173 x264_frame_push_unused( h, x264_frame_pop( h->frames.reference ) );
2178 static inline void x264_reference_hierarchy_reset( x264_t *h )
2181 int b_hasdelayframe = 0;
2183 /* look for delay frames -- chain must only contain frames that are disposable */
2184 for( int i = 0; h->frames.current[i] && IS_DISPOSABLE( h->frames.current[i]->i_type ); i++ )
2185 b_hasdelayframe |= h->frames.current[i]->i_coded
2186 != h->frames.current[i]->i_frame + h->sps->vui.i_num_reorder_frames;
2188 /* This function must handle b-pyramid and clear frames for open-gop */
2189 if( h->param.i_bframe_pyramid != X264_B_PYRAMID_STRICT && !b_hasdelayframe && h->frames.i_poc_last_open_gop == -1 )
2192 /* Remove last BREF. There will never be old BREFs in the
2193 * dpb during a BREF decode when pyramid == STRICT */
2194 for( ref = 0; h->frames.reference[ref]; ref++ )
2196 if( ( h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT
2197 && h->frames.reference[ref]->i_type == X264_TYPE_BREF )
2198 || ( h->frames.reference[ref]->i_poc < h->frames.i_poc_last_open_gop
2199 && h->sh.i_type != SLICE_TYPE_B ) )
2201 int diff = h->i_frame_num - h->frames.reference[ref]->i_frame_num;
2202 h->sh.mmco[h->sh.i_mmco_command_count].i_difference_of_pic_nums = diff;
2203 h->sh.mmco[h->sh.i_mmco_command_count++].i_poc = h->frames.reference[ref]->i_poc;
2204 x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[ref] ) );
2205 h->b_ref_reorder[0] = 1;
2210 /* Prepare room in the dpb for the delayed display time of the later b-frame's */
2211 if( h->param.i_bframe_pyramid )
2212 h->sh.i_mmco_remove_from_end = X264_MAX( ref + 2 - h->frames.i_max_dpb, 0 );
2215 static inline void x264_slice_init( x264_t *h, int i_nal_type, int i_global_qp )
2217 /* ------------------------ Create slice header ----------------------- */
2218 if( i_nal_type == NAL_SLICE_IDR )
2220 x264_slice_header_init( h, &h->sh, h->sps, h->pps, h->i_idr_pic_id, h->i_frame_num, i_global_qp );
2223 h->i_idr_pic_id ^= 1;
2227 x264_slice_header_init( h, &h->sh, h->sps, h->pps, -1, h->i_frame_num, i_global_qp );
2229 h->sh.i_num_ref_idx_l0_active = h->i_ref[0] <= 0 ? 1 : h->i_ref[0];
2230 h->sh.i_num_ref_idx_l1_active = h->i_ref[1] <= 0 ? 1 : h->i_ref[1];
2231 if( h->sh.i_num_ref_idx_l0_active != h->pps->i_num_ref_idx_l0_default_active ||
2232 (h->sh.i_type == SLICE_TYPE_B && h->sh.i_num_ref_idx_l1_active != h->pps->i_num_ref_idx_l1_default_active) )
2234 h->sh.b_num_ref_idx_override = 1;
2238 if( h->fenc->i_type == X264_TYPE_BREF && h->param.b_bluray_compat && h->sh.i_mmco_command_count )
2241 h->sh_backup = h->sh;
2244 h->fdec->i_frame_num = h->sh.i_frame_num;
2246 if( h->sps->i_poc_type == 0 )
2248 h->sh.i_poc = h->fdec->i_poc;
2249 if( PARAM_INTERLACED )
2251 h->sh.i_delta_poc_bottom = h->param.b_tff ? 1 : -1;
2252 h->sh.i_poc += h->sh.i_delta_poc_bottom == -1;
2255 h->sh.i_delta_poc_bottom = 0;
2256 h->fdec->i_delta_poc[0] = h->sh.i_delta_poc_bottom == -1;
2257 h->fdec->i_delta_poc[1] = h->sh.i_delta_poc_bottom == 1;
2261 /* Nothing to do ? */
2264 x264_macroblock_slice_init( h );
2270 uint8_t cabac_prevbyte;
2273 x264_frame_stat_t stat;
2276 int field_decoding_flag;
2279 static ALWAYS_INLINE void x264_bitstream_backup( x264_t *h, x264_bs_bak_t *bak, int i_skip, int full )
2283 bak->stat = h->stat.frame;
2284 bak->last_qp = h->mb.i_last_qp;
2285 bak->last_dqp = h->mb.i_last_dqp;
2286 bak->field_decoding_flag = h->mb.field_decoding_flag;
2290 bak->stat.i_mv_bits = h->stat.frame.i_mv_bits;
2291 bak->stat.i_tex_bits = h->stat.frame.i_tex_bits;
2293 /* In the per-MB backup, we don't need the contexts because flushing the CABAC
2294 * encoder has no context dependency and in this case, a slice is ended (and
2295 * thus the content of all contexts are thrown away). */
2296 if( h->param.b_cabac )
2299 memcpy( &bak->cabac, &h->cabac, sizeof(x264_cabac_t) );
2301 memcpy( &bak->cabac, &h->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
2302 /* x264's CABAC writer modifies the previous byte during carry, so it has to be
2304 bak->cabac_prevbyte = h->cabac.p[-1];
2308 bak->bs = h->out.bs;
2313 static ALWAYS_INLINE void x264_bitstream_restore( x264_t *h, x264_bs_bak_t *bak, int *skip, int full )
2317 h->stat.frame = bak->stat;
2318 h->mb.i_last_qp = bak->last_qp;
2319 h->mb.i_last_dqp = bak->last_dqp;
2320 h->mb.field_decoding_flag = bak->field_decoding_flag;
2324 h->stat.frame.i_mv_bits = bak->stat.i_mv_bits;
2325 h->stat.frame.i_tex_bits = bak->stat.i_tex_bits;
2327 if( h->param.b_cabac )
2330 memcpy( &h->cabac, &bak->cabac, sizeof(x264_cabac_t) );
2332 memcpy( &h->cabac, &bak->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
2333 h->cabac.p[-1] = bak->cabac_prevbyte;
2337 h->out.bs = bak->bs;
2342 static int x264_slice_write( x264_t *h )
2345 int mb_xy, i_mb_x, i_mb_y;
2346 /* NALUs other than the first use a 3-byte startcode.
2347 * Add one extra byte for the rbsp, and one more for the final CABAC putbyte.
2348 * Then add an extra 5 bytes just in case, to account for random NAL escapes and
2349 * other inaccuracies. */
2350 int overhead_guess = (NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal)) + 1 + h->param.b_cabac + 5;
2351 int slice_max_size = h->param.i_slice_max_size > 0 ? (h->param.i_slice_max_size-overhead_guess)*8 : 0;
2352 int back_up_bitstream = slice_max_size || (!h->param.b_cabac && h->sps->i_profile_idc < PROFILE_HIGH);
2353 int starting_bits = bs_pos(&h->out.bs);
2354 int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
2355 int b_hpel = h->fdec->b_kept_as_ref;
2356 int orig_last_mb = h->sh.i_last_mb;
2357 int thread_last_mb = h->i_threadslice_end * h->mb.i_mb_width - 1;
2358 uint8_t *last_emu_check;
2359 #define BS_BAK_SLICE_MAX_SIZE 0
2360 #define BS_BAK_SLICE_MIN_MBS 1
2361 #define BS_BAK_ROW_VBV 2
2362 x264_bs_bak_t bs_bak[3];
2363 b_deblock &= b_hpel || h->param.b_full_recon || h->param.psz_dump_yuv;
2364 bs_realign( &h->out.bs );
2367 x264_nal_start( h, h->i_nal_type, h->i_nal_ref_idc );
2368 h->out.nal[h->out.i_nal].i_first_mb = h->sh.i_first_mb;
2371 x264_macroblock_thread_init( h );
2373 /* Set the QP equal to the first QP in the slice for more accurate CABAC initialization. */
2374 h->mb.i_mb_xy = h->sh.i_first_mb;
2375 h->sh.i_qp = x264_ratecontrol_mb_qp( h );
2376 h->sh.i_qp = SPEC_QP( h->sh.i_qp );
2377 h->sh.i_qp_delta = h->sh.i_qp - h->pps->i_pic_init_qp;
2379 x264_slice_header_write( &h->out.bs, &h->sh, h->i_nal_ref_idc );
2380 if( h->param.b_cabac )
2382 /* alignment needed */
2383 bs_align_1( &h->out.bs );
2386 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 );
2387 x264_cabac_encode_init ( &h->cabac, h->out.bs.p, h->out.bs.p_end );
2388 last_emu_check = h->cabac.p;
2391 last_emu_check = h->out.bs.p;
2392 h->mb.i_last_qp = h->sh.i_qp;
2393 h->mb.i_last_dqp = 0;
2394 h->mb.field_decoding_flag = 0;
2396 i_mb_y = h->sh.i_first_mb / h->mb.i_mb_width;
2397 i_mb_x = h->sh.i_first_mb % h->mb.i_mb_width;
2402 mb_xy = i_mb_x + i_mb_y * h->mb.i_mb_width;
2403 int mb_spos = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
2407 if( x264_bitstream_check_buffer( h ) )
2409 if( !(i_mb_y & SLICE_MBAFF) && h->param.rc.i_vbv_buffer_size )
2410 x264_bitstream_backup( h, &bs_bak[BS_BAK_ROW_VBV], i_skip, 1 );
2411 if( !h->mb.b_reencode_mb )
2412 x264_fdec_filter_row( h, i_mb_y, 0 );
2415 if( !(i_mb_y & SLICE_MBAFF) && back_up_bitstream )
2417 x264_bitstream_backup( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], i_skip, 0 );
2418 if( slice_max_size && (thread_last_mb+1-mb_xy) == h->param.i_slice_min_mbs )
2419 x264_bitstream_backup( h, &bs_bak[BS_BAK_SLICE_MIN_MBS], i_skip, 0 );
2422 if( PARAM_INTERLACED )
2424 if( h->mb.b_adaptive_mbaff )
2428 /* FIXME: VSAD is fast but fairly poor at choosing the best interlace type. */
2429 h->mb.b_interlaced = x264_field_vsad( h, i_mb_x, i_mb_y );
2430 memcpy( &h->zigzagf, MB_INTERLACED ? &h->zigzagf_interlaced : &h->zigzagf_progressive, sizeof(h->zigzagf) );
2431 if( !MB_INTERLACED && (i_mb_y+2) == h->mb.i_mb_height )
2432 x264_expand_border_mbpair( h, i_mb_x, i_mb_y );
2435 h->mb.field[mb_xy] = MB_INTERLACED;
2440 x264_macroblock_cache_load_interlaced( h, i_mb_x, i_mb_y );
2442 x264_macroblock_cache_load_progressive( h, i_mb_x, i_mb_y );
2444 x264_macroblock_analyse( h );
2446 /* encode this macroblock -> be careful it can change the mb type to P_SKIP if needed */
2448 x264_macroblock_encode( h );
2450 if( h->param.b_cabac )
2452 if( mb_xy > h->sh.i_first_mb && !(SLICE_MBAFF && (i_mb_y&1)) )
2453 x264_cabac_encode_terminal( &h->cabac );
2455 if( IS_SKIP( h->mb.i_type ) )
2456 x264_cabac_mb_skip( h, 1 );
2459 if( h->sh.i_type != SLICE_TYPE_I )
2460 x264_cabac_mb_skip( h, 0 );
2461 x264_macroblock_write_cabac( h, &h->cabac );
2466 if( IS_SKIP( h->mb.i_type ) )
2470 if( h->sh.i_type != SLICE_TYPE_I )
2472 bs_write_ue( &h->out.bs, i_skip ); /* skip run */
2475 x264_macroblock_write_cavlc( h );
2476 /* If there was a CAVLC level code overflow, try again at a higher QP. */
2477 if( h->mb.b_overflow )
2479 h->mb.i_chroma_qp = h->chroma_qp_table[++h->mb.i_qp];
2480 h->mb.i_skip_intra = 0;
2481 h->mb.b_skip_mc = 0;
2482 h->mb.b_overflow = 0;
2483 x264_bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], &i_skip, 0 );
2489 int total_bits = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
2490 int mb_size = total_bits - mb_spos;
2492 if( slice_max_size && (!SLICE_MBAFF || (i_mb_y&1)) )
2494 /* Count the skip run, just in case. */
2495 if( !h->param.b_cabac )
2496 total_bits += bs_size_ue_big( i_skip );
2497 /* Check for escape bytes. */
2498 uint8_t *end = h->param.b_cabac ? h->cabac.p : h->out.bs.p;
2499 for( ; last_emu_check < end - 2; last_emu_check++ )
2500 if( last_emu_check[0] == 0 && last_emu_check[1] == 0 && last_emu_check[2] <= 3 )
2502 slice_max_size -= 8;
2505 /* We'll just re-encode this last macroblock if we go over the max slice size. */
2506 if( total_bits - starting_bits > slice_max_size && !h->mb.b_reencode_mb )
2508 if( !x264_frame_new_slice( h, h->fdec ) )
2510 /* Handle the most obnoxious slice-min-mbs edge case: we need to end the slice
2511 * because it's gone over the maximum size, but doing so would violate slice-min-mbs.
2512 * If possible, roll back to the last checkpoint and try again.
2513 * We could try raising QP, but that would break in the case where a slice spans multiple
2514 * rows, which the re-encoding infrastructure can't currently handle. */
2515 if( mb_xy <= thread_last_mb && (thread_last_mb+1-mb_xy) < h->param.i_slice_min_mbs )
2517 if( thread_last_mb-h->param.i_slice_min_mbs < h->sh.i_first_mb+h->param.i_slice_min_mbs )
2519 x264_log( h, X264_LOG_WARNING, "slice-max-size violated (frame %d, cause: slice-min-mbs)\n", h->i_frame );
2523 x264_bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MIN_MBS], &i_skip, 0 );
2524 h->mb.b_reencode_mb = 1;
2525 h->sh.i_last_mb = thread_last_mb-h->param.i_slice_min_mbs;
2528 if( mb_xy-SLICE_MBAFF*h->mb.i_mb_stride != h->sh.i_first_mb )
2530 x264_bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], &i_skip, 0 );
2531 h->mb.b_reencode_mb = 1;
2534 // set to bottom of previous mbpair
2536 h->sh.i_last_mb = mb_xy-1+h->mb.i_mb_stride*(!(i_mb_y&1));
2538 h->sh.i_last_mb = (i_mb_y-2+!(i_mb_y&1))*h->mb.i_mb_stride + h->mb.i_mb_width - 1;
2541 h->sh.i_last_mb = mb_xy-1;
2545 h->sh.i_last_mb = mb_xy;
2552 h->mb.b_reencode_mb = 0;
2555 if( h->param.b_visualize )
2556 x264_visualize_mb( h );
2560 x264_macroblock_cache_save( h );
2562 if( x264_ratecontrol_mb( h, mb_size ) < 0 )
2564 x264_bitstream_restore( h, &bs_bak[BS_BAK_ROW_VBV], &i_skip, 1 );
2565 h->mb.b_reencode_mb = 1;
2567 i_mb_y = i_mb_y - SLICE_MBAFF;
2568 h->mb.i_mb_prev_xy = i_mb_y * h->mb.i_mb_stride - 1;
2569 h->sh.i_last_mb = orig_last_mb;
2573 /* accumulate mb stats */
2574 h->stat.frame.i_mb_count[h->mb.i_type]++;
2576 int b_intra = IS_INTRA( h->mb.i_type );
2577 int b_skip = IS_SKIP( h->mb.i_type );
2578 if( h->param.i_log_level >= X264_LOG_INFO || h->param.rc.b_stat_write )
2580 if( !b_intra && !b_skip && !IS_DIRECT( h->mb.i_type ) )
2582 if( h->mb.i_partition != D_8x8 )
2583 h->stat.frame.i_mb_partition[h->mb.i_partition] += 4;
2585 for( int i = 0; i < 4; i++ )
2586 h->stat.frame.i_mb_partition[h->mb.i_sub_partition[i]] ++;
2587 if( h->param.i_frame_reference > 1 )
2588 for( int i_list = 0; i_list <= (h->sh.i_type == SLICE_TYPE_B); i_list++ )
2589 for( int i = 0; i < 4; i++ )
2591 int i_ref = h->mb.cache.ref[i_list][ x264_scan8[4*i] ];
2593 h->stat.frame.i_mb_count_ref[i_list][i_ref] ++;
2598 if( h->param.i_log_level >= X264_LOG_INFO )
2600 if( h->mb.i_cbp_luma | h->mb.i_cbp_chroma )
2604 for( int i = 0; i < 4; i++ )
2605 if( h->mb.i_cbp_luma & (1 << i) )
2606 for( int p = 0; p < 3; p++ )
2609 int nnz8x8 = M16( &h->mb.cache.non_zero_count[x264_scan8[s8]+0] )
2610 | M16( &h->mb.cache.non_zero_count[x264_scan8[s8]+8] );
2611 h->stat.frame.i_mb_cbp[!b_intra + p*2] += !!nnz8x8;
2616 int cbpsum = (h->mb.i_cbp_luma&1) + ((h->mb.i_cbp_luma>>1)&1)
2617 + ((h->mb.i_cbp_luma>>2)&1) + (h->mb.i_cbp_luma>>3);
2618 h->stat.frame.i_mb_cbp[!b_intra + 0] += cbpsum;
2619 h->stat.frame.i_mb_cbp[!b_intra + 2] += !!h->mb.i_cbp_chroma;
2620 h->stat.frame.i_mb_cbp[!b_intra + 4] += h->mb.i_cbp_chroma >> 1;
2623 if( h->mb.i_cbp_luma && !b_intra )
2625 h->stat.frame.i_mb_count_8x8dct[0] ++;
2626 h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
2628 if( b_intra && h->mb.i_type != I_PCM )
2630 if( h->mb.i_type == I_16x16 )
2631 h->stat.frame.i_mb_pred_mode[0][h->mb.i_intra16x16_pred_mode]++;
2632 else if( h->mb.i_type == I_8x8 )
2633 for( int i = 0; i < 16; i += 4 )
2634 h->stat.frame.i_mb_pred_mode[1][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
2635 else //if( h->mb.i_type == I_4x4 )
2636 for( int i = 0; i < 16; i++ )
2637 h->stat.frame.i_mb_pred_mode[2][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
2638 h->stat.frame.i_mb_pred_mode[3][x264_mb_chroma_pred_mode_fix[h->mb.i_chroma_pred_mode]]++;
2640 h->stat.frame.i_mb_field[b_intra?0:b_skip?2:1] += MB_INTERLACED;
2643 /* calculate deblock strength values (actual deblocking is done per-row along with hpel) */
2645 x264_macroblock_deblock_strength( h );
2647 if( mb_xy == h->sh.i_last_mb )
2652 i_mb_x += i_mb_y & 1;
2653 i_mb_y ^= i_mb_x < h->mb.i_mb_width;
2657 if( i_mb_x == h->mb.i_mb_width )
2663 if( h->sh.i_last_mb < h->sh.i_first_mb )
2666 h->out.nal[h->out.i_nal].i_last_mb = h->sh.i_last_mb;
2668 if( h->param.b_cabac )
2670 x264_cabac_encode_flush( h, &h->cabac );
2671 h->out.bs.p = h->cabac.p;
2676 bs_write_ue( &h->out.bs, i_skip ); /* last skip run */
2677 /* rbsp_slice_trailing_bits */
2678 bs_rbsp_trailing( &h->out.bs );
2679 bs_flush( &h->out.bs );
2681 if( x264_nal_end( h ) )
2684 if( h->sh.i_last_mb == (h->i_threadslice_end * h->mb.i_mb_width - 1) )
2686 h->stat.frame.i_misc_bits = bs_pos( &h->out.bs )
2687 + (h->out.i_nal*NALU_OVERHEAD * 8)
2688 - h->stat.frame.i_tex_bits
2689 - h->stat.frame.i_mv_bits;
2690 x264_fdec_filter_row( h, h->i_threadslice_end, 0 );
2692 if( h->param.b_sliced_threads )
2694 /* Tell the main thread we're done. */
2695 x264_threadslice_cond_broadcast( h, 1 );
2697 for( int mb_y = h->i_threadslice_start; mb_y <= h->i_threadslice_end; mb_y++ )
2698 x264_fdec_filter_row( h, mb_y, 1 );
2699 x264_threadslice_cond_broadcast( h, 2 );
2700 /* Do the first row of hpel, now that the previous slice is done */
2701 if( h->i_thread_idx > 0 )
2703 x264_threadslice_cond_wait( h->thread[h->i_thread_idx-1], 2 );
2704 x264_fdec_filter_row( h, h->i_threadslice_start + (1 << SLICE_MBAFF), 2 );
2708 /* Free mb info after the last thread's done using it */
2709 if( h->fdec->mb_info_free && (!h->param.b_sliced_threads || h->i_thread_idx == (h->param.i_threads-1)) )
2711 h->fdec->mb_info_free( h->fdec->mb_info );
2712 h->fdec->mb_info = NULL;
2713 h->fdec->mb_info_free = NULL;
2720 static void x264_thread_sync_context( x264_t *dst, x264_t *src )
2725 // reference counting
2726 for( x264_frame_t **f = src->frames.reference; *f; f++ )
2727 (*f)->i_reference_count++;
2728 for( x264_frame_t **f = dst->frames.reference; *f; f++ )
2729 x264_frame_push_unused( src, *f );
2730 src->fdec->i_reference_count++;
2731 x264_frame_push_unused( src, dst->fdec );
2733 // copy everything except the per-thread pointers and the constants.
2734 memcpy( &dst->i_frame, &src->i_frame, offsetof(x264_t, mb.type) - offsetof(x264_t, i_frame) );
2735 dst->param = src->param;
2736 dst->stat = src->stat;
2737 dst->pixf = src->pixf;
2740 static void x264_thread_sync_stat( x264_t *dst, x264_t *src )
2744 memcpy( &dst->stat.i_frame_count, &src->stat.i_frame_count, sizeof(dst->stat) - sizeof(dst->stat.frame) );
2747 static void *x264_slices_write( x264_t *h )
2749 int i_slice_num = 0;
2750 int last_thread_mb = h->sh.i_last_mb;
2753 if( h->param.b_visualize )
2754 if( x264_visualize_init( h ) )
2759 memset( &h->stat.frame, 0, sizeof(h->stat.frame) );
2760 h->mb.b_reencode_mb = 0;
2761 while( h->sh.i_first_mb + SLICE_MBAFF*h->mb.i_mb_stride <= last_thread_mb )
2763 h->sh.i_last_mb = last_thread_mb;
2764 if( !i_slice_num || !x264_frame_new_slice( h, h->fdec ) )
2766 if( h->param.i_slice_max_mbs )
2770 // convert first to mbaff form, add slice-max-mbs, then convert back to normal form
2771 int last_mbaff = 2*(h->sh.i_first_mb % h->mb.i_mb_width)
2772 + h->mb.i_mb_width*(h->sh.i_first_mb / h->mb.i_mb_width)
2773 + h->param.i_slice_max_mbs - 1;
2774 int last_x = (last_mbaff % (2*h->mb.i_mb_width))/2;
2775 int last_y = (last_mbaff / (2*h->mb.i_mb_width))*2 + 1;
2776 h->sh.i_last_mb = last_x + h->mb.i_mb_stride*last_y;
2780 h->sh.i_last_mb = h->sh.i_first_mb + h->param.i_slice_max_mbs - 1;
2781 if( h->sh.i_last_mb < last_thread_mb && last_thread_mb - h->sh.i_last_mb < h->param.i_slice_min_mbs )
2782 h->sh.i_last_mb = last_thread_mb - h->param.i_slice_min_mbs;
2786 else if( h->param.i_slice_count && !h->param.b_sliced_threads )
2788 int height = h->mb.i_mb_height >> PARAM_INTERLACED;
2789 int width = h->mb.i_mb_width << PARAM_INTERLACED;
2791 h->sh.i_last_mb = (height * i_slice_num + h->param.i_slice_count/2) / h->param.i_slice_count * width - 1;
2794 h->sh.i_last_mb = X264_MIN( h->sh.i_last_mb, last_thread_mb );
2795 if( x264_stack_align( x264_slice_write, h ) )
2797 h->sh.i_first_mb = h->sh.i_last_mb + 1;
2798 // if i_first_mb is not the last mb in a row then go to the next mb in MBAFF order
2799 if( SLICE_MBAFF && h->sh.i_first_mb % h->mb.i_mb_width )
2800 h->sh.i_first_mb -= h->mb.i_mb_stride;
2804 if( h->param.b_visualize )
2806 x264_visualize_show( h );
2807 x264_visualize_close( h );
2814 /* Tell other threads we're done, so they wouldn't wait for it */
2815 if( h->param.b_sliced_threads )
2816 x264_threadslice_cond_broadcast( h, 2 );
2820 static int x264_threaded_slices_write( x264_t *h )
2822 /* set first/last mb and sync contexts */
2823 for( int i = 0; i < h->param.i_threads; i++ )
2825 x264_t *t = h->thread[i];
2828 t->param = h->param;
2829 memcpy( &t->i_frame, &h->i_frame, offsetof(x264_t, rc) - offsetof(x264_t, i_frame) );
2831 int height = h->mb.i_mb_height >> PARAM_INTERLACED;
2832 t->i_threadslice_start = ((height * i + h->param.i_slice_count/2) / h->param.i_threads) << PARAM_INTERLACED;
2833 t->i_threadslice_end = ((height * (i+1) + h->param.i_slice_count/2) / h->param.i_threads) << PARAM_INTERLACED;
2834 t->sh.i_first_mb = t->i_threadslice_start * h->mb.i_mb_width;
2835 t->sh.i_last_mb = t->i_threadslice_end * h->mb.i_mb_width - 1;
2838 x264_stack_align( x264_analyse_weight_frame, h, h->mb.i_mb_height*16 + 16 );
2840 x264_threads_distribute_ratecontrol( h );
2843 for( int i = 0; i < h->param.i_threads; i++ )
2845 h->thread[i]->i_thread_idx = i;
2846 h->thread[i]->b_thread_active = 1;
2847 x264_threadslice_cond_broadcast( h->thread[i], 0 );
2850 for( int i = 0; i < h->param.i_threads; i++ )
2851 x264_threadpool_run( h->threadpool, (void*)x264_slices_write, h->thread[i] );
2853 for( int i = 0; i < h->param.i_threads; i++ )
2854 x264_threadslice_cond_wait( h->thread[i], 1 );
2856 x264_threads_merge_ratecontrol( h );
2858 for( int i = 1; i < h->param.i_threads; i++ )
2860 x264_t *t = h->thread[i];
2861 for( int j = 0; j < t->out.i_nal; j++ )
2863 h->out.nal[h->out.i_nal] = t->out.nal[j];
2865 x264_nal_check_buffer( h );
2867 /* All entries in stat.frame are ints except for ssd/ssim. */
2868 for( int j = 0; j < (offsetof(x264_t,stat.frame.i_ssd) - offsetof(x264_t,stat.frame.i_mv_bits)) / sizeof(int); j++ )
2869 ((int*)&h->stat.frame)[j] += ((int*)&t->stat.frame)[j];
2870 for( int j = 0; j < 3; j++ )
2871 h->stat.frame.i_ssd[j] += t->stat.frame.i_ssd[j];
2872 h->stat.frame.f_ssim += t->stat.frame.f_ssim;
2873 h->stat.frame.i_ssim_cnt += t->stat.frame.i_ssim_cnt;
2879 void x264_encoder_intra_refresh( x264_t *h )
2881 h = h->thread[h->i_thread_phase];
2882 h->b_queued_intra_refresh = 1;
2885 int x264_encoder_invalidate_reference( x264_t *h, int64_t pts )
2887 if( h->param.i_bframe )
2889 x264_log( h, X264_LOG_ERROR, "x264_encoder_invalidate_reference is not supported with B-frames enabled\n" );
2892 if( h->param.b_intra_refresh )
2894 x264_log( h, X264_LOG_ERROR, "x264_encoder_invalidate_reference is not supported with intra refresh enabled\n" );
2897 h = h->thread[h->i_thread_phase];
2898 if( pts >= h->i_last_idr_pts )
2900 for( int i = 0; h->frames.reference[i]; i++ )
2901 if( pts <= h->frames.reference[i]->i_pts )
2902 h->frames.reference[i]->b_corrupt = 1;
2903 if( pts <= h->fdec->i_pts )
2904 h->fdec->b_corrupt = 1;
2909 /****************************************************************************
2910 * x264_encoder_encode:
2911 * XXX: i_poc : is the poc of the current given picture
2912 * i_frame : is the number of the frame being coded
2913 * ex: type frame poc
2921 ****************************************************************************/
2922 int x264_encoder_encode( x264_t *h,
2923 x264_nal_t **pp_nal, int *pi_nal,
2924 x264_picture_t *pic_in,
2925 x264_picture_t *pic_out )
2927 x264_t *thread_current, *thread_prev, *thread_oldest;
2928 int i_nal_type, i_nal_ref_idc, i_global_qp;
2929 int overhead = NALU_OVERHEAD;
2932 if( h->opencl.b_fatal_error )
2936 if( h->i_thread_frames > 1 )
2938 thread_prev = h->thread[ h->i_thread_phase ];
2939 h->i_thread_phase = (h->i_thread_phase + 1) % h->i_thread_frames;
2940 thread_current = h->thread[ h->i_thread_phase ];
2941 thread_oldest = h->thread[ (h->i_thread_phase + 1) % h->i_thread_frames ];
2942 x264_thread_sync_context( thread_current, thread_prev );
2943 x264_thread_sync_ratecontrol( thread_current, thread_prev, thread_oldest );
2952 if( h->param.cpu&X264_CPU_SSE_MISALIGN )
2953 x264_cpu_mask_misalign_sse();
2955 h->i_cpb_delay_pir_offset = h->i_cpb_delay_pir_offset_next;
2961 /* ------------------- Setup new frame from picture -------------------- */
2962 if( pic_in != NULL )
2964 /* 1: Copy the picture to a frame and move it to a buffer */
2965 x264_frame_t *fenc = x264_frame_pop_unused( h, 0 );
2969 if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 )
2972 if( h->param.i_width != 16 * h->mb.i_mb_width ||
2973 h->param.i_height != 16 * h->mb.i_mb_height )
2974 x264_frame_expand_border_mod16( h, fenc );
2976 fenc->i_frame = h->frames.i_input++;
2978 if( fenc->i_frame == 0 )
2979 h->frames.i_first_pts = fenc->i_pts;
2980 if( h->frames.i_bframe_delay && fenc->i_frame == h->frames.i_bframe_delay )
2981 h->frames.i_bframe_delay_time = fenc->i_pts - h->frames.i_first_pts;
2983 if( h->param.b_vfr_input && fenc->i_pts <= h->frames.i_largest_pts )
2984 x264_log( h, X264_LOG_WARNING, "non-strictly-monotonic PTS\n" );
2986 h->frames.i_second_largest_pts = h->frames.i_largest_pts;
2987 h->frames.i_largest_pts = fenc->i_pts;
2989 if( (fenc->i_pic_struct < PIC_STRUCT_AUTO) || (fenc->i_pic_struct > PIC_STRUCT_TRIPLE) )
2990 fenc->i_pic_struct = PIC_STRUCT_AUTO;
2992 if( fenc->i_pic_struct == PIC_STRUCT_AUTO )
2995 int b_interlaced = fenc->param ? fenc->param->b_interlaced : h->param.b_interlaced;
2997 int b_interlaced = 0;
3001 int b_tff = fenc->param ? fenc->param->b_tff : h->param.b_tff;
3002 fenc->i_pic_struct = b_tff ? PIC_STRUCT_TOP_BOTTOM : PIC_STRUCT_BOTTOM_TOP;
3005 fenc->i_pic_struct = PIC_STRUCT_PROGRESSIVE;
3008 if( h->param.rc.b_mb_tree && h->param.rc.b_stat_read )
3010 if( x264_macroblock_tree_read( h, fenc, pic_in->prop.quant_offsets ) )
3014 x264_stack_align( x264_adaptive_quant_frame, h, fenc, pic_in->prop.quant_offsets );
3016 if( pic_in->prop.quant_offsets_free )
3017 pic_in->prop.quant_offsets_free( pic_in->prop.quant_offsets );
3019 if( h->frames.b_have_lowres )
3020 x264_frame_init_lowres( h, fenc );
3022 /* 2: Place the frame into the queue for its slice type decision */
3023 x264_lookahead_put_frame( h, fenc );
3025 if( h->frames.i_input <= h->frames.i_delay + 1 - h->i_thread_frames )
3027 /* Nothing yet to encode, waiting for filling of buffers */
3028 pic_out->i_type = X264_TYPE_AUTO;
3034 /* signal kills for lookahead thread */
3035 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
3036 h->lookahead->b_exit_thread = 1;
3037 x264_pthread_cond_broadcast( &h->lookahead->ifbuf.cv_fill );
3038 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
3042 /* 3: The picture is analyzed in the lookahead */
3043 if( !h->frames.current[0] )
3044 x264_lookahead_get_frames( h );
3046 if( !h->frames.current[0] && x264_lookahead_is_empty( h ) )
3047 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
3049 /* ------------------- Get frame to be encoded ------------------------- */
3050 /* 4: get picture to encode */
3051 h->fenc = x264_frame_shift( h->frames.current );
3053 /* If applicable, wait for previous frame reconstruction to finish */
3054 if( h->param.b_sliced_threads )
3055 if( x264_threadpool_wait_all( h ) < 0 )
3058 if( h->i_frame == h->i_thread_frames - 1 )
3059 h->i_reordered_pts_delay = h->fenc->i_reordered_pts;
3060 if( h->fenc->param )
3062 x264_encoder_reconfig( h, h->fenc->param );
3063 if( h->fenc->param->param_free )
3065 h->fenc->param->param_free( h->fenc->param );
3066 h->fenc->param = NULL;
3070 // ok to call this before encoding any frames, since the initial values of fdec have b_kept_as_ref=0
3071 if( x264_reference_update( h ) )
3073 h->fdec->i_lines_completed = -1;
3075 if( !IS_X264_TYPE_I( h->fenc->i_type ) )
3077 int valid_refs_left = 0;
3078 for( int i = 0; h->frames.reference[i]; i++ )
3079 if( !h->frames.reference[i]->b_corrupt )
3081 /* No valid reference frames left: force an IDR. */
3082 if( !valid_refs_left )
3084 h->fenc->b_keyframe = 1;
3085 h->fenc->i_type = X264_TYPE_IDR;
3089 if( h->fenc->b_keyframe )
3091 h->frames.i_last_keyframe = h->fenc->i_frame;
3092 if( h->fenc->i_type == X264_TYPE_IDR )
3095 h->frames.i_last_idr = h->fenc->i_frame;
3098 h->sh.i_mmco_command_count =
3099 h->sh.i_mmco_remove_from_end = 0;
3100 h->b_ref_reorder[0] =
3101 h->b_ref_reorder[1] = 0;
3103 h->fenc->i_poc = 2 * ( h->fenc->i_frame - X264_MAX( h->frames.i_last_idr, 0 ) );
3105 /* ------------------- Setup frame context ----------------------------- */
3106 /* 5: Init data dependent of frame type */
3107 if( h->fenc->i_type == X264_TYPE_IDR )
3109 /* reset ref pictures */
3110 i_nal_type = NAL_SLICE_IDR;
3111 i_nal_ref_idc = NAL_PRIORITY_HIGHEST;
3112 h->sh.i_type = SLICE_TYPE_I;
3113 x264_reference_reset( h );
3114 h->frames.i_poc_last_open_gop = -1;
3116 else if( h->fenc->i_type == X264_TYPE_I )
3118 i_nal_type = NAL_SLICE;
3119 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
3120 h->sh.i_type = SLICE_TYPE_I;
3121 x264_reference_hierarchy_reset( h );
3122 if( h->param.b_open_gop )
3123 h->frames.i_poc_last_open_gop = h->fenc->b_keyframe ? h->fenc->i_poc : -1;
3125 else if( h->fenc->i_type == X264_TYPE_P )
3127 i_nal_type = NAL_SLICE;
3128 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
3129 h->sh.i_type = SLICE_TYPE_P;
3130 x264_reference_hierarchy_reset( h );
3131 h->frames.i_poc_last_open_gop = -1;
3133 else if( h->fenc->i_type == X264_TYPE_BREF )
3135 i_nal_type = NAL_SLICE;
3136 i_nal_ref_idc = h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT ? NAL_PRIORITY_LOW : NAL_PRIORITY_HIGH;
3137 h->sh.i_type = SLICE_TYPE_B;
3138 x264_reference_hierarchy_reset( h );
3142 i_nal_type = NAL_SLICE;
3143 i_nal_ref_idc = NAL_PRIORITY_DISPOSABLE;
3144 h->sh.i_type = SLICE_TYPE_B;
3147 h->fdec->i_type = h->fenc->i_type;
3148 h->fdec->i_frame = h->fenc->i_frame;
3149 h->fenc->b_kept_as_ref =
3150 h->fdec->b_kept_as_ref = i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE && h->param.i_keyint_max > 1;
3152 h->fdec->mb_info = h->fenc->mb_info;
3153 h->fdec->mb_info_free = h->fenc->mb_info_free;
3154 h->fenc->mb_info = NULL;
3155 h->fenc->mb_info_free = NULL;
3157 h->fdec->i_pts = h->fenc->i_pts;
3158 if( h->frames.i_bframe_delay )
3160 int64_t *prev_reordered_pts = thread_current->frames.i_prev_reordered_pts;
3161 h->fdec->i_dts = h->i_frame > h->frames.i_bframe_delay
3162 ? prev_reordered_pts[ (h->i_frame - h->frames.i_bframe_delay) % h->frames.i_bframe_delay ]
3163 : h->fenc->i_reordered_pts - h->frames.i_bframe_delay_time;
3164 prev_reordered_pts[ h->i_frame % h->frames.i_bframe_delay ] = h->fenc->i_reordered_pts;
3167 h->fdec->i_dts = h->fenc->i_reordered_pts;
3168 if( h->fenc->i_type == X264_TYPE_IDR )
3169 h->i_last_idr_pts = h->fdec->i_pts;
3171 /* ------------------- Init ----------------------------- */
3172 /* build ref list 0/1 */
3173 x264_reference_build_list( h, h->fdec->i_poc );
3175 /* ---------------------- Write the bitstream -------------------------- */
3176 /* Init bitstream context */
3177 if( h->param.b_sliced_threads )
3179 for( int i = 0; i < h->param.i_threads; i++ )
3181 bs_init( &h->thread[i]->out.bs, h->thread[i]->out.p_bitstream, h->thread[i]->out.i_bitstream );
3182 h->thread[i]->out.i_nal = 0;
3187 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
3191 if( h->param.b_aud )
3195 if( h->sh.i_type == SLICE_TYPE_I )
3197 else if( h->sh.i_type == SLICE_TYPE_P )
3199 else if( h->sh.i_type == SLICE_TYPE_B )
3204 x264_nal_start( h, NAL_AUD, NAL_PRIORITY_DISPOSABLE );
3205 bs_write( &h->out.bs, 3, pic_type );
3206 bs_rbsp_trailing( &h->out.bs );
3207 if( x264_nal_end( h ) )
3209 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3212 h->i_nal_type = i_nal_type;
3213 h->i_nal_ref_idc = i_nal_ref_idc;
3215 if( h->param.b_intra_refresh )
3217 if( IS_X264_TYPE_I( h->fenc->i_type ) )
3219 h->fdec->i_frames_since_pir = 0;
3220 h->b_queued_intra_refresh = 0;
3221 /* PIR is currently only supported with ref == 1, so any intra frame effectively refreshes
3222 * the whole frame and counts as an intra refresh. */
3223 h->fdec->f_pir_position = h->mb.i_mb_width;
3225 else if( h->fenc->i_type == X264_TYPE_P )
3227 int pocdiff = (h->fdec->i_poc - h->fref[0][0]->i_poc)/2;
3228 float increment = X264_MAX( ((float)h->mb.i_mb_width-1) / h->param.i_keyint_max, 1 );
3229 h->fdec->f_pir_position = h->fref[0][0]->f_pir_position;
3230 h->fdec->i_frames_since_pir = h->fref[0][0]->i_frames_since_pir + pocdiff;
3231 if( h->fdec->i_frames_since_pir >= h->param.i_keyint_max ||
3232 (h->b_queued_intra_refresh && h->fdec->f_pir_position + 0.5 >= h->mb.i_mb_width) )
3234 h->fdec->f_pir_position = 0;
3235 h->fdec->i_frames_since_pir = 0;
3236 h->b_queued_intra_refresh = 0;
3237 h->fenc->b_keyframe = 1;
3239 h->fdec->i_pir_start_col = h->fdec->f_pir_position+0.5;
3240 h->fdec->f_pir_position += increment * pocdiff;
3241 h->fdec->i_pir_end_col = h->fdec->f_pir_position+0.5;
3242 /* If our intra refresh has reached the right side of the frame, we're done. */
3243 if( h->fdec->i_pir_end_col >= h->mb.i_mb_width - 1 )
3245 h->fdec->f_pir_position = h->mb.i_mb_width;
3246 h->fdec->i_pir_end_col = h->mb.i_mb_width - 1;
3251 if( h->fenc->b_keyframe )
3253 /* Write SPS and PPS */
3254 if( h->param.b_repeat_headers )
3256 /* generate sequence parameters */
3257 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
3258 x264_sps_write( &h->out.bs, h->sps );
3259 if( x264_nal_end( h ) )
3261 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3263 /* generate picture parameters */
3264 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
3265 x264_pps_write( &h->out.bs, h->sps, h->pps );
3266 if( x264_nal_end( h ) )
3268 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3271 /* when frame threading is used, buffering period sei is written in x264_encoder_frame_end */
3272 if( h->i_thread_frames == 1 && h->sps->vui.b_nal_hrd_parameters_present )
3274 x264_hrd_fullness( h );
3275 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3276 x264_sei_buffering_period_write( h, &h->out.bs );
3277 if( x264_nal_end( h ) )
3279 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3283 /* write extra sei */
3284 for( int i = 0; i < h->fenc->extra_sei.num_payloads; i++ )
3286 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3287 x264_sei_write( &h->out.bs, h->fenc->extra_sei.payloads[i].payload, h->fenc->extra_sei.payloads[i].payload_size,
3288 h->fenc->extra_sei.payloads[i].payload_type );
3289 if( x264_nal_end( h ) )
3291 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3292 if( h->fenc->extra_sei.sei_free )
3294 h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads[i].payload );
3295 h->fenc->extra_sei.payloads[i].payload = NULL;
3299 if( h->fenc->extra_sei.sei_free )
3301 h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads );
3302 h->fenc->extra_sei.payloads = NULL;
3303 h->fenc->extra_sei.sei_free = NULL;
3306 if( h->fenc->b_keyframe )
3308 if( h->param.b_repeat_headers && h->fenc->i_frame == 0 )
3310 /* identify ourself */
3311 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3312 if( x264_sei_version_write( h, &h->out.bs ) )
3314 if( x264_nal_end( h ) )
3316 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3319 if( h->fenc->i_type != X264_TYPE_IDR )
3321 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;
3322 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3323 x264_sei_recovery_point_write( h, &h->out.bs, time_to_recovery );
3324 if( x264_nal_end( h ) )
3326 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3329 if ( h->param.i_frame_packing >= 0 )
3331 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3332 x264_sei_frame_packing_write( h, &h->out.bs );
3333 if( x264_nal_end( h ) )
3335 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3339 /* generate sei pic timing */
3340 if( h->sps->vui.b_pic_struct_present || h->sps->vui.b_nal_hrd_parameters_present )
3342 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3343 x264_sei_pic_timing_write( h, &h->out.bs );
3344 if( x264_nal_end( h ) )
3346 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3349 /* As required by Blu-ray. */
3350 if( !IS_X264_TYPE_B( h->fenc->i_type ) && h->b_sh_backup )
3353 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3354 x264_sei_dec_ref_pic_marking_write( h, &h->out.bs );
3355 if( x264_nal_end( h ) )
3357 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal-1);
3360 if( h->fenc->b_keyframe && h->param.b_intra_refresh )
3361 h->i_cpb_delay_pir_offset_next = h->fenc->i_cpb_delay;
3363 /* Init the rate control */
3364 /* FIXME: Include slice header bit cost. */
3365 x264_ratecontrol_start( h, h->fenc->i_qpplus1, overhead*8 );
3366 i_global_qp = x264_ratecontrol_qp( h );
3368 pic_out->i_qpplus1 =
3369 h->fdec->i_qpplus1 = i_global_qp + 1;
3371 if( h->param.rc.b_stat_read && h->sh.i_type != SLICE_TYPE_I )
3373 x264_reference_build_list_optimal( h );
3374 x264_reference_check_reorder( h );
3378 h->fdec->i_poc_l0ref0 = h->fref[0][0]->i_poc;
3380 /* ------------------------ Create slice header ----------------------- */
3381 x264_slice_init( h, i_nal_type, i_global_qp );
3383 /*------------------------- Weights -------------------------------------*/
3384 if( h->sh.i_type == SLICE_TYPE_B )
3385 x264_macroblock_bipred_init( h );
3387 x264_weighted_pred_init( h );
3389 if( i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE )
3393 h->i_threadslice_start = 0;
3394 h->i_threadslice_end = h->mb.i_mb_height;
3395 if( h->i_thread_frames > 1 )
3397 x264_threadpool_run( h->threadpool, (void*)x264_slices_write, h );
3398 h->b_thread_active = 1;
3400 else if( h->param.b_sliced_threads )
3402 if( x264_threaded_slices_write( h ) )
3406 if( (intptr_t)x264_slices_write( h ) )
3409 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
3412 static int x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
3413 x264_nal_t **pp_nal, int *pi_nal,
3414 x264_picture_t *pic_out )
3416 char psz_message[80];
3418 if( !h->param.b_sliced_threads && h->b_thread_active )
3420 h->b_thread_active = 0;
3421 if( (intptr_t)x264_threadpool_wait( h->threadpool, h ) )
3426 pic_out->i_type = X264_TYPE_AUTO;
3432 /* generate buffering period sei and insert it into place */
3433 if( h->i_thread_frames > 1 && h->fenc->b_keyframe && h->sps->vui.b_nal_hrd_parameters_present )
3435 x264_hrd_fullness( h );
3436 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3437 x264_sei_buffering_period_write( h, &h->out.bs );
3438 if( x264_nal_end( h ) )
3440 /* buffering period sei must follow AUD, SPS and PPS and precede all other SEIs */
3442 while( h->out.nal[idx].i_type == NAL_AUD ||
3443 h->out.nal[idx].i_type == NAL_SPS ||
3444 h->out.nal[idx].i_type == NAL_PPS )
3446 x264_nal_t nal_tmp = h->out.nal[h->out.i_nal-1];
3447 memmove( &h->out.nal[idx+1], &h->out.nal[idx], (h->out.i_nal-idx-1)*sizeof(x264_nal_t) );
3448 h->out.nal[idx] = nal_tmp;
3451 int frame_size = x264_encoder_encapsulate_nals( h, 0 );
3452 if( frame_size < 0 )
3455 /* Set output picture properties */
3456 pic_out->i_type = h->fenc->i_type;
3458 pic_out->b_keyframe = h->fenc->b_keyframe;
3459 pic_out->i_pic_struct = h->fenc->i_pic_struct;
3461 pic_out->i_pts = h->fdec->i_pts;
3462 pic_out->i_dts = h->fdec->i_dts;
3464 if( pic_out->i_pts < pic_out->i_dts )
3465 x264_log( h, X264_LOG_WARNING, "invalid DTS: PTS is less than DTS\n" );
3467 pic_out->opaque = h->fenc->opaque;
3469 pic_out->img.i_csp = h->fdec->i_csp;
3471 pic_out->img.i_csp |= X264_CSP_HIGH_DEPTH;
3473 pic_out->img.i_plane = h->fdec->i_plane;
3474 for( int i = 0; i < pic_out->img.i_plane; i++ )
3476 pic_out->img.i_stride[i] = h->fdec->i_stride[i] * sizeof(pixel);
3477 pic_out->img.plane[i] = (uint8_t*)h->fdec->plane[i];
3480 x264_frame_push_unused( thread_current, h->fenc );
3482 /* ---------------------- Update encoder state ------------------------- */
3486 if( x264_ratecontrol_end( h, frame_size * 8, &filler ) < 0 )
3489 pic_out->hrd_timing = h->fenc->hrd_timing;
3490 pic_out->prop.f_crf_avg = h->fdec->f_crf_avg;
3495 overhead = (FILLER_OVERHEAD - h->param.b_annexb);
3496 if( h->param.i_slice_max_size && filler > h->param.i_slice_max_size )
3498 int next_size = filler - h->param.i_slice_max_size;
3499 int overflow = X264_MAX( overhead - next_size, 0 );
3500 f = h->param.i_slice_max_size - overhead - overflow;
3503 f = X264_MAX( 0, filler - overhead );
3505 if( x264_bitstream_check_buffer_filler( h, f ) )
3507 x264_nal_start( h, NAL_FILLER, NAL_PRIORITY_DISPOSABLE );
3508 x264_filler_write( h, &h->out.bs, f );
3509 if( x264_nal_end( h ) )
3511 int total_size = x264_encoder_encapsulate_nals( h, h->out.i_nal-1 );
3512 if( total_size < 0 )
3514 frame_size += total_size;
3515 filler -= total_size;
3518 /* End bitstream, set output */
3519 *pi_nal = h->out.i_nal;
3520 *pp_nal = h->out.nal;
3524 x264_noise_reduction_update( h );
3526 /* ---------------------- Compute/Print statistics --------------------- */
3527 x264_thread_sync_stat( h, h->thread[0] );
3530 h->stat.i_frame_count[h->sh.i_type]++;
3531 h->stat.i_frame_size[h->sh.i_type] += frame_size;
3532 h->stat.f_frame_qp[h->sh.i_type] += h->fdec->f_qp_avg_aq;
3534 for( int i = 0; i < X264_MBTYPE_MAX; i++ )
3535 h->stat.i_mb_count[h->sh.i_type][i] += h->stat.frame.i_mb_count[i];
3536 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
3537 h->stat.i_mb_partition[h->sh.i_type][i] += h->stat.frame.i_mb_partition[i];
3538 for( int i = 0; i < 2; i++ )
3539 h->stat.i_mb_count_8x8dct[i] += h->stat.frame.i_mb_count_8x8dct[i];
3540 for( int i = 0; i < 6; i++ )
3541 h->stat.i_mb_cbp[i] += h->stat.frame.i_mb_cbp[i];
3542 for( int i = 0; i < 4; i++ )
3543 for( int j = 0; j < 13; j++ )
3544 h->stat.i_mb_pred_mode[i][j] += h->stat.frame.i_mb_pred_mode[i][j];
3545 if( h->sh.i_type != SLICE_TYPE_I )
3546 for( int i_list = 0; i_list < 2; i_list++ )
3547 for( int i = 0; i < X264_REF_MAX*2; i++ )
3548 h->stat.i_mb_count_ref[h->sh.i_type][i_list][i] += h->stat.frame.i_mb_count_ref[i_list][i];
3549 for( int i = 0; i < 3; i++ )
3550 h->stat.i_mb_field[i] += h->stat.frame.i_mb_field[i];
3551 if( h->sh.i_type == SLICE_TYPE_P && h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE )
3553 h->stat.i_wpred[0] += !!h->sh.weight[0][0].weightfn;
3554 h->stat.i_wpred[1] += !!h->sh.weight[0][1].weightfn || !!h->sh.weight[0][2].weightfn;
3556 if( h->sh.i_type == SLICE_TYPE_B )
3558 h->stat.i_direct_frames[ h->sh.b_direct_spatial_mv_pred ] ++;
3559 if( h->mb.b_direct_auto_write )
3561 //FIXME somewhat arbitrary time constants
3562 if( h->stat.i_direct_score[0] + h->stat.i_direct_score[1] > h->mb.i_mb_count )
3563 for( int i = 0; i < 2; i++ )
3564 h->stat.i_direct_score[i] = h->stat.i_direct_score[i] * 9/10;
3565 for( int i = 0; i < 2; i++ )
3566 h->stat.i_direct_score[i] += h->stat.frame.i_direct_score[i];
3570 h->stat.i_consecutive_bframes[h->fenc->i_bframes]++;
3572 psz_message[0] = '\0';
3573 double dur = h->fenc->f_duration;
3574 h->stat.f_frame_duration[h->sh.i_type] += dur;
3575 if( h->param.analyse.b_psnr )
3579 h->stat.frame.i_ssd[0],
3580 h->stat.frame.i_ssd[1],
3581 h->stat.frame.i_ssd[2],
3583 int luma_size = h->param.i_width * h->param.i_height;
3584 int chroma_size = CHROMA_SIZE( luma_size );
3585 pic_out->prop.f_psnr[0] = x264_psnr( ssd[0], luma_size );
3586 pic_out->prop.f_psnr[1] = x264_psnr( ssd[1], chroma_size );
3587 pic_out->prop.f_psnr[2] = x264_psnr( ssd[2], chroma_size );
3588 pic_out->prop.f_psnr_avg = x264_psnr( ssd[0] + ssd[1] + ssd[2], luma_size + chroma_size*2 );
3590 h->stat.f_ssd_global[h->sh.i_type] += dur * (ssd[0] + ssd[1] + ssd[2]);
3591 h->stat.f_psnr_average[h->sh.i_type] += dur * pic_out->prop.f_psnr_avg;
3592 h->stat.f_psnr_mean_y[h->sh.i_type] += dur * pic_out->prop.f_psnr[0];
3593 h->stat.f_psnr_mean_u[h->sh.i_type] += dur * pic_out->prop.f_psnr[1];
3594 h->stat.f_psnr_mean_v[h->sh.i_type] += dur * pic_out->prop.f_psnr[2];
3596 snprintf( psz_message, 80, " PSNR Y:%5.2f U:%5.2f V:%5.2f", pic_out->prop.f_psnr[0],
3597 pic_out->prop.f_psnr[1],
3598 pic_out->prop.f_psnr[2] );
3601 if( h->param.analyse.b_ssim )
3603 pic_out->prop.f_ssim = h->stat.frame.f_ssim / h->stat.frame.i_ssim_cnt;
3604 h->stat.f_ssim_mean_y[h->sh.i_type] += pic_out->prop.f_ssim * dur;
3605 snprintf( psz_message + strlen(psz_message), 80 - strlen(psz_message),
3606 " SSIM Y:%.5f", pic_out->prop.f_ssim );
3608 psz_message[79] = '\0';
3610 x264_log( h, X264_LOG_DEBUG,
3611 "frame=%4d QP=%.2f NAL=%d Slice:%c Poc:%-3d I:%-4d P:%-4d SKIP:%-4d size=%d bytes%s\n",
3613 h->fdec->f_qp_avg_aq,
3615 h->sh.i_type == SLICE_TYPE_I ? 'I' : (h->sh.i_type == SLICE_TYPE_P ? 'P' : 'B' ),
3617 h->stat.frame.i_mb_count_i,
3618 h->stat.frame.i_mb_count_p,
3619 h->stat.frame.i_mb_count_skip,
3623 // keep stats all in one place
3624 x264_thread_sync_stat( h->thread[0], h );
3625 // for the use of the next frame
3626 x264_thread_sync_stat( thread_current, h );
3628 #ifdef DEBUG_MB_TYPE
3630 static const char mb_chars[] = { 'i', 'i', 'I', 'C', 'P', '8', 'S',
3631 'D', '<', 'X', 'B', 'X', '>', 'B', 'B', 'B', 'B', '8', 'S' };
3632 for( int mb_xy = 0; mb_xy < h->mb.i_mb_width * h->mb.i_mb_height; mb_xy++ )
3634 if( h->mb.type[mb_xy] < X264_MBTYPE_MAX && h->mb.type[mb_xy] >= 0 )
3635 fprintf( stderr, "%c ", mb_chars[ h->mb.type[mb_xy] ] );
3637 fprintf( stderr, "? " );
3639 if( (mb_xy+1) % h->mb.i_mb_width == 0 )
3640 fprintf( stderr, "\n" );
3645 /* Remove duplicates, must be done near the end as breaks h->fref0 array
3646 * by freeing some of its pointers. */
3647 for( int i = 0; i < h->i_ref[0]; i++ )
3648 if( h->fref[0][i] && h->fref[0][i]->b_duplicate )
3650 x264_frame_push_blank_unused( h, h->fref[0][i] );
3654 if( h->param.psz_dump_yuv )
3655 x264_frame_dump( h );
3661 static void x264_print_intra( int64_t *i_mb_count, double i_count, int b_print_pcm, char *intra )
3663 intra += sprintf( intra, "I16..4%s: %4.1f%% %4.1f%% %4.1f%%",
3664 b_print_pcm ? "..PCM" : "",
3665 i_mb_count[I_16x16]/ i_count,
3666 i_mb_count[I_8x8] / i_count,
3667 i_mb_count[I_4x4] / i_count );
3669 sprintf( intra, " %4.1f%%", i_mb_count[I_PCM] / i_count );
3672 /****************************************************************************
3673 * x264_encoder_close:
3674 ****************************************************************************/
3675 void x264_encoder_close ( x264_t *h )
3677 int64_t i_yuv_size = FRAME_SIZE( h->param.i_width * h->param.i_height );
3678 int64_t i_mb_count_size[2][7] = {{0}};
3680 int b_print_pcm = h->stat.i_mb_count[SLICE_TYPE_I][I_PCM]
3681 || h->stat.i_mb_count[SLICE_TYPE_P][I_PCM]
3682 || h->stat.i_mb_count[SLICE_TYPE_B][I_PCM];
3684 x264_lookahead_delete( h );
3687 x264_opencl_lookahead_delete( h );
3688 x264_opencl_function_t *ocl = h->opencl.ocl;
3691 if( h->param.b_sliced_threads )
3692 x264_threadpool_wait_all( h );
3693 if( h->param.i_threads > 1 )
3694 x264_threadpool_delete( h->threadpool );
3695 if( h->param.i_lookahead_threads > 1 )
3696 x264_threadpool_delete( h->lookaheadpool );
3697 if( h->i_thread_frames > 1 )
3699 for( int i = 0; i < h->i_thread_frames; i++ )
3700 if( h->thread[i]->b_thread_active )
3702 assert( h->thread[i]->fenc->i_reference_count == 1 );
3703 x264_frame_delete( h->thread[i]->fenc );
3706 x264_t *thread_prev = h->thread[h->i_thread_phase];
3707 x264_thread_sync_ratecontrol( h, thread_prev, h );
3708 x264_thread_sync_ratecontrol( thread_prev, thread_prev, h );
3709 h->i_frame = thread_prev->i_frame + 1 - h->i_thread_frames;
3713 /* Slices used and PSNR */
3714 for( int i = 0; i < 3; i++ )
3716 static const uint8_t slice_order[] = { SLICE_TYPE_I, SLICE_TYPE_P, SLICE_TYPE_B };
3717 int i_slice = slice_order[i];
3719 if( h->stat.i_frame_count[i_slice] > 0 )
3721 int i_count = h->stat.i_frame_count[i_slice];
3722 double dur = h->stat.f_frame_duration[i_slice];
3723 if( h->param.analyse.b_psnr )
3725 x264_log( h, X264_LOG_INFO,
3726 "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",
3727 slice_type_to_char[i_slice],
3729 h->stat.f_frame_qp[i_slice] / i_count,
3730 (double)h->stat.i_frame_size[i_slice] / i_count,
3731 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,
3732 h->stat.f_psnr_average[i_slice] / dur,
3733 x264_psnr( h->stat.f_ssd_global[i_slice], dur * i_yuv_size ) );
3737 x264_log( h, X264_LOG_INFO,
3738 "frame %c:%-5d Avg QP:%5.2f size:%6.0f\n",
3739 slice_type_to_char[i_slice],
3741 h->stat.f_frame_qp[i_slice] / i_count,
3742 (double)h->stat.i_frame_size[i_slice] / i_count );
3746 if( h->param.i_bframe && h->stat.i_frame_count[SLICE_TYPE_B] )
3750 // weight by number of frames (including the I/P-frames) that are in a sequence of N B-frames
3751 for( int i = 0; i <= h->param.i_bframe; i++ )
3752 den += (i+1) * h->stat.i_consecutive_bframes[i];
3753 for( int i = 0; i <= h->param.i_bframe; i++ )
3754 p += sprintf( p, " %4.1f%%", 100. * (i+1) * h->stat.i_consecutive_bframes[i] / den );
3755 x264_log( h, X264_LOG_INFO, "consecutive B-frames:%s\n", buf );
3758 for( int i_type = 0; i_type < 2; i_type++ )
3759 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
3761 if( i == D_DIRECT_8x8 ) continue; /* direct is counted as its own type */
3762 i_mb_count_size[i_type][x264_mb_partition_pixel_table[i]] += h->stat.i_mb_partition[i_type][i];
3766 if( h->stat.i_frame_count[SLICE_TYPE_I] > 0 )
3768 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_I];
3769 double i_count = h->stat.i_frame_count[SLICE_TYPE_I] * h->mb.i_mb_count / 100.0;
3770 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
3771 x264_log( h, X264_LOG_INFO, "mb I %s\n", buf );
3773 if( h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
3775 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_P];
3776 double i_count = h->stat.i_frame_count[SLICE_TYPE_P] * h->mb.i_mb_count / 100.0;
3777 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_P];
3778 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
3779 x264_log( h, X264_LOG_INFO,
3780 "mb P %s P16..4: %4.1f%% %4.1f%% %4.1f%% %4.1f%% %4.1f%% skip:%4.1f%%\n",
3782 i_mb_size[PIXEL_16x16] / (i_count*4),
3783 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
3784 i_mb_size[PIXEL_8x8] / (i_count*4),
3785 (i_mb_size[PIXEL_8x4] + i_mb_size[PIXEL_4x8]) / (i_count*4),
3786 i_mb_size[PIXEL_4x4] / (i_count*4),
3787 i_mb_count[P_SKIP] / i_count );
3789 if( h->stat.i_frame_count[SLICE_TYPE_B] > 0 )
3791 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_B];
3792 double i_count = h->stat.i_frame_count[SLICE_TYPE_B] * h->mb.i_mb_count / 100.0;
3793 double i_mb_list_count;
3794 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_B];
3795 int64_t list_count[3] = {0}; /* 0 == L0, 1 == L1, 2 == BI */
3796 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
3797 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
3798 for( int j = 0; j < 2; j++ )
3800 int l0 = x264_mb_type_list_table[i][0][j];
3801 int l1 = x264_mb_type_list_table[i][1][j];
3803 list_count[l1+l0*l1] += h->stat.i_mb_count[SLICE_TYPE_B][i] * 2;
3805 list_count[0] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L0_8x8];
3806 list_count[1] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L1_8x8];
3807 list_count[2] += h->stat.i_mb_partition[SLICE_TYPE_B][D_BI_8x8];
3808 i_mb_count[B_DIRECT] += (h->stat.i_mb_partition[SLICE_TYPE_B][D_DIRECT_8x8]+2)/4;
3809 i_mb_list_count = (list_count[0] + list_count[1] + list_count[2]) / 100.0;
3810 sprintf( buf + strlen(buf), " B16..8: %4.1f%% %4.1f%% %4.1f%% direct:%4.1f%% skip:%4.1f%%",
3811 i_mb_size[PIXEL_16x16] / (i_count*4),
3812 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
3813 i_mb_size[PIXEL_8x8] / (i_count*4),
3814 i_mb_count[B_DIRECT] / i_count,
3815 i_mb_count[B_SKIP] / i_count );
3816 if( i_mb_list_count != 0 )
3817 sprintf( buf + strlen(buf), " L0:%4.1f%% L1:%4.1f%% BI:%4.1f%%",
3818 list_count[0] / i_mb_list_count,
3819 list_count[1] / i_mb_list_count,
3820 list_count[2] / i_mb_list_count );
3821 x264_log( h, X264_LOG_INFO, "mb B %s\n", buf );
3824 x264_ratecontrol_summary( h );
3826 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 )
3828 #define SUM3(p) (p[SLICE_TYPE_I] + p[SLICE_TYPE_P] + p[SLICE_TYPE_B])
3829 #define SUM3b(p,o) (p[SLICE_TYPE_I][o] + p[SLICE_TYPE_P][o] + p[SLICE_TYPE_B][o])
3830 int64_t i_i8x8 = SUM3b( h->stat.i_mb_count, I_8x8 );
3831 int64_t i_intra = i_i8x8 + SUM3b( h->stat.i_mb_count, I_4x4 )
3832 + SUM3b( h->stat.i_mb_count, I_16x16 );
3833 int64_t i_all_intra = i_intra + SUM3b( h->stat.i_mb_count, I_PCM);
3834 int64_t i_skip = SUM3b( h->stat.i_mb_count, P_SKIP )
3835 + SUM3b( h->stat.i_mb_count, B_SKIP );
3836 const int i_count = h->stat.i_frame_count[SLICE_TYPE_I] +
3837 h->stat.i_frame_count[SLICE_TYPE_P] +
3838 h->stat.i_frame_count[SLICE_TYPE_B];
3839 int64_t i_mb_count = (int64_t)i_count * h->mb.i_mb_count;
3840 int64_t i_inter = i_mb_count - i_skip - i_intra;
3841 const double duration = h->stat.f_frame_duration[SLICE_TYPE_I] +
3842 h->stat.f_frame_duration[SLICE_TYPE_P] +
3843 h->stat.f_frame_duration[SLICE_TYPE_B];
3844 float f_bitrate = SUM3(h->stat.i_frame_size) / duration / 125;
3846 if( PARAM_INTERLACED )
3848 char *fieldstats = buf;
3851 fieldstats += sprintf( fieldstats, " inter:%.1f%%", h->stat.i_mb_field[1] * 100.0 / i_inter );
3853 fieldstats += sprintf( fieldstats, " skip:%.1f%%", h->stat.i_mb_field[2] * 100.0 / i_skip );
3854 x264_log( h, X264_LOG_INFO, "field mbs: intra: %.1f%%%s\n",
3855 h->stat.i_mb_field[0] * 100.0 / i_intra, buf );
3858 if( h->pps->b_transform_8x8_mode )
3861 if( h->stat.i_mb_count_8x8dct[0] )
3862 sprintf( buf, " inter:%.1f%%", 100. * h->stat.i_mb_count_8x8dct[1] / h->stat.i_mb_count_8x8dct[0] );
3863 x264_log( h, X264_LOG_INFO, "8x8 transform intra:%.1f%%%s\n", 100. * i_i8x8 / i_intra, buf );
3866 if( (h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO ||
3867 (h->stat.i_direct_frames[0] && h->stat.i_direct_frames[1]))
3868 && h->stat.i_frame_count[SLICE_TYPE_B] )
3870 x264_log( h, X264_LOG_INFO, "direct mvs spatial:%.1f%% temporal:%.1f%%\n",
3871 h->stat.i_direct_frames[1] * 100. / h->stat.i_frame_count[SLICE_TYPE_B],
3872 h->stat.i_direct_frames[0] * 100. / h->stat.i_frame_count[SLICE_TYPE_B] );
3876 int csize = CHROMA444 ? 4 : 1;
3877 if( i_mb_count != i_all_intra )
3878 sprintf( buf, " inter: %.1f%% %.1f%% %.1f%%",
3879 h->stat.i_mb_cbp[1] * 100.0 / ((i_mb_count - i_all_intra)*4),
3880 h->stat.i_mb_cbp[3] * 100.0 / ((i_mb_count - i_all_intra)*csize),
3881 h->stat.i_mb_cbp[5] * 100.0 / ((i_mb_count - i_all_intra)*csize) );
3882 x264_log( h, X264_LOG_INFO, "coded y,%s,%s intra: %.1f%% %.1f%% %.1f%%%s\n",
3883 CHROMA444?"u":"uvDC", CHROMA444?"v":"uvAC",
3884 h->stat.i_mb_cbp[0] * 100.0 / (i_all_intra*4),
3885 h->stat.i_mb_cbp[2] * 100.0 / (i_all_intra*csize),
3886 h->stat.i_mb_cbp[4] * 100.0 / (i_all_intra*csize), buf );
3888 int64_t fixed_pred_modes[4][9] = {{0}};
3889 int64_t sum_pred_modes[4] = {0};
3890 for( int i = 0; i <= I_PRED_16x16_DC_128; i++ )
3892 fixed_pred_modes[0][x264_mb_pred_mode16x16_fix[i]] += h->stat.i_mb_pred_mode[0][i];
3893 sum_pred_modes[0] += h->stat.i_mb_pred_mode[0][i];
3895 if( sum_pred_modes[0] )
3896 x264_log( h, X264_LOG_INFO, "i16 v,h,dc,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
3897 fixed_pred_modes[0][0] * 100.0 / sum_pred_modes[0],
3898 fixed_pred_modes[0][1] * 100.0 / sum_pred_modes[0],
3899 fixed_pred_modes[0][2] * 100.0 / sum_pred_modes[0],
3900 fixed_pred_modes[0][3] * 100.0 / sum_pred_modes[0] );
3901 for( int i = 1; i <= 2; i++ )
3903 for( int j = 0; j <= I_PRED_8x8_DC_128; j++ )
3905 fixed_pred_modes[i][x264_mb_pred_mode4x4_fix(j)] += h->stat.i_mb_pred_mode[i][j];
3906 sum_pred_modes[i] += h->stat.i_mb_pred_mode[i][j];
3908 if( sum_pred_modes[i] )
3909 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,
3910 fixed_pred_modes[i][0] * 100.0 / sum_pred_modes[i],
3911 fixed_pred_modes[i][1] * 100.0 / sum_pred_modes[i],
3912 fixed_pred_modes[i][2] * 100.0 / sum_pred_modes[i],
3913 fixed_pred_modes[i][3] * 100.0 / sum_pred_modes[i],
3914 fixed_pred_modes[i][4] * 100.0 / sum_pred_modes[i],
3915 fixed_pred_modes[i][5] * 100.0 / sum_pred_modes[i],
3916 fixed_pred_modes[i][6] * 100.0 / sum_pred_modes[i],
3917 fixed_pred_modes[i][7] * 100.0 / sum_pred_modes[i],
3918 fixed_pred_modes[i][8] * 100.0 / sum_pred_modes[i] );
3920 for( int i = 0; i <= I_PRED_CHROMA_DC_128; i++ )
3922 fixed_pred_modes[3][x264_mb_chroma_pred_mode_fix[i]] += h->stat.i_mb_pred_mode[3][i];
3923 sum_pred_modes[3] += h->stat.i_mb_pred_mode[3][i];
3925 if( sum_pred_modes[3] && !CHROMA444 )
3926 x264_log( h, X264_LOG_INFO, "i8c dc,h,v,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
3927 fixed_pred_modes[3][0] * 100.0 / sum_pred_modes[3],
3928 fixed_pred_modes[3][1] * 100.0 / sum_pred_modes[3],
3929 fixed_pred_modes[3][2] * 100.0 / sum_pred_modes[3],
3930 fixed_pred_modes[3][3] * 100.0 / sum_pred_modes[3] );
3932 if( h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE && h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
3933 x264_log( h, X264_LOG_INFO, "Weighted P-Frames: Y:%.1f%% UV:%.1f%%\n",
3934 h->stat.i_wpred[0] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P],
3935 h->stat.i_wpred[1] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P] );
3937 for( int i_list = 0; i_list < 2; i_list++ )
3938 for( int i_slice = 0; i_slice < 2; i_slice++ )
3943 for( int i = 0; i < X264_REF_MAX*2; i++ )
3944 if( h->stat.i_mb_count_ref[i_slice][i_list][i] )
3946 i_den += h->stat.i_mb_count_ref[i_slice][i_list][i];
3951 for( int i = 0; i <= i_max; i++ )
3952 p += sprintf( p, " %4.1f%%", 100. * h->stat.i_mb_count_ref[i_slice][i_list][i] / i_den );
3953 x264_log( h, X264_LOG_INFO, "ref %c L%d:%s\n", "PB"[i_slice], i_list, buf );
3956 if( h->param.analyse.b_ssim )
3958 float ssim = SUM3( h->stat.f_ssim_mean_y ) / duration;
3959 x264_log( h, X264_LOG_INFO, "SSIM Mean Y:%.7f (%6.3fdb)\n", ssim, x264_ssim( ssim ) );
3961 if( h->param.analyse.b_psnr )
3963 x264_log( h, X264_LOG_INFO,
3964 "PSNR Mean Y:%6.3f U:%6.3f V:%6.3f Avg:%6.3f Global:%6.3f kb/s:%.2f\n",
3965 SUM3( h->stat.f_psnr_mean_y ) / duration,
3966 SUM3( h->stat.f_psnr_mean_u ) / duration,
3967 SUM3( h->stat.f_psnr_mean_v ) / duration,
3968 SUM3( h->stat.f_psnr_average ) / duration,
3969 x264_psnr( SUM3( h->stat.f_ssd_global ), duration * i_yuv_size ),
3973 x264_log( h, X264_LOG_INFO, "kb/s:%.2f\n", f_bitrate );
3977 x264_ratecontrol_delete( h );
3980 if( h->param.rc.psz_stat_out )
3981 free( h->param.rc.psz_stat_out );
3982 if( h->param.rc.psz_stat_in )
3983 free( h->param.rc.psz_stat_in );
3985 x264_cqm_delete( h );
3986 x264_free( h->nal_buffer );
3987 x264_analyse_free_costs( h );
3989 if( h->i_thread_frames > 1 )
3990 h = h->thread[h->i_thread_phase];
3993 x264_frame_delete_list( h->frames.unused[0] );
3994 x264_frame_delete_list( h->frames.unused[1] );
3995 x264_frame_delete_list( h->frames.current );
3996 x264_frame_delete_list( h->frames.blank_unused );
4000 for( int i = 0; i < h->i_thread_frames; i++ )
4001 if( h->thread[i]->b_thread_active )
4002 for( int j = 0; j < h->thread[i]->i_ref[0]; j++ )
4003 if( h->thread[i]->fref[0][j] && h->thread[i]->fref[0][j]->b_duplicate )
4004 x264_frame_delete( h->thread[i]->fref[0][j] );
4006 if( h->param.i_lookahead_threads > 1 )
4007 for( int i = 0; i < h->param.i_lookahead_threads; i++ )
4008 x264_free( h->lookahead_thread[i] );
4010 for( int i = h->param.i_threads - 1; i >= 0; i-- )
4012 x264_frame_t **frame;
4014 if( !h->param.b_sliced_threads || i == 0 )
4016 for( frame = h->thread[i]->frames.reference; *frame; frame++ )
4018 assert( (*frame)->i_reference_count > 0 );
4019 (*frame)->i_reference_count--;
4020 if( (*frame)->i_reference_count == 0 )
4021 x264_frame_delete( *frame );
4023 frame = &h->thread[i]->fdec;
4026 assert( (*frame)->i_reference_count > 0 );
4027 (*frame)->i_reference_count--;
4028 if( (*frame)->i_reference_count == 0 )
4029 x264_frame_delete( *frame );
4031 x264_macroblock_cache_free( h->thread[i] );
4033 x264_macroblock_thread_free( h->thread[i], 0 );
4034 x264_free( h->thread[i]->out.p_bitstream );
4035 x264_free( h->thread[i]->out.nal );
4036 x264_pthread_mutex_destroy( &h->thread[i]->mutex );
4037 x264_pthread_cond_destroy( &h->thread[i]->cv );
4038 x264_free( h->thread[i] );
4041 x264_opencl_close_library( ocl );
4045 int x264_encoder_delayed_frames( x264_t *h )
4047 int delayed_frames = 0;
4048 if( h->i_thread_frames > 1 )
4050 for( int i = 0; i < h->i_thread_frames; i++ )
4051 delayed_frames += h->thread[i]->b_thread_active;
4052 h = h->thread[h->i_thread_phase];
4054 for( int i = 0; h->frames.current[i]; i++ )
4056 x264_pthread_mutex_lock( &h->lookahead->ofbuf.mutex );
4057 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
4058 x264_pthread_mutex_lock( &h->lookahead->next.mutex );
4059 delayed_frames += h->lookahead->ifbuf.i_size + h->lookahead->next.i_size + h->lookahead->ofbuf.i_size;
4060 x264_pthread_mutex_unlock( &h->lookahead->next.mutex );
4061 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
4062 x264_pthread_mutex_unlock( &h->lookahead->ofbuf.mutex );
4063 return delayed_frames;
4066 int x264_encoder_maximum_delayed_frames( x264_t *h )
4068 return h->frames.i_delay;