1 /*****************************************************************************
2 * encoder.c: top-level encoder functions
3 *****************************************************************************
4 * Copyright (C) 2003-2015 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"
35 #if HAVE_INTEL_DISPATCHER
36 #include "extras/intel_dispatcher.h"
39 //#define DEBUG_MB_TYPE
41 #define bs_write_ue bs_write_ue_big
43 static int x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
44 x264_nal_t **pp_nal, int *pi_nal,
45 x264_picture_t *pic_out );
47 /****************************************************************************
49 ******************************* x264 libs **********************************
51 ****************************************************************************/
52 static double x264_psnr( double sqe, double size )
54 double mse = sqe / (PIXEL_MAX*PIXEL_MAX * size);
55 if( mse <= 0.0000000001 ) /* Max 100dB */
58 return -10.0 * log10( mse );
61 static double x264_ssim( double ssim )
63 double inv_ssim = 1 - ssim;
64 if( inv_ssim <= 0.0000000001 ) /* Max 100dB */
67 return -10.0 * log10( inv_ssim );
70 static int x264_threadpool_wait_all( x264_t *h )
72 for( int i = 0; i < h->param.i_threads; i++ )
73 if( h->thread[i]->b_thread_active )
75 h->thread[i]->b_thread_active = 0;
76 if( (intptr_t)x264_threadpool_wait( h->threadpool, h->thread[i] ) < 0 )
82 static void x264_frame_dump( x264_t *h )
84 FILE *f = x264_fopen( h->param.psz_dump_yuv, "r+b" );
88 /* Wait for the threads to finish deblocking */
89 if( h->param.b_sliced_threads )
90 x264_threadpool_wait_all( h );
92 /* Write the frame in display order */
93 int frame_size = FRAME_SIZE( h->param.i_height * h->param.i_width * sizeof(pixel) );
94 fseek( f, (uint64_t)h->fdec->i_frame * frame_size, SEEK_SET );
95 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
96 for( int y = 0; y < h->param.i_height; y++ )
97 fwrite( &h->fdec->plane[p][y*h->fdec->i_stride[p]], sizeof(pixel), h->param.i_width, f );
100 int cw = h->param.i_width>>1;
101 int ch = h->param.i_height>>CHROMA_V_SHIFT;
102 pixel *planeu = x264_malloc( (cw*ch*2+32)*sizeof(pixel) );
105 pixel *planev = planeu + cw*ch + 16;
106 h->mc.plane_copy_deinterleave( planeu, cw, planev, cw, h->fdec->plane[1], h->fdec->i_stride[1], cw, ch );
107 fwrite( planeu, 1, cw*ch*sizeof(pixel), f );
108 fwrite( planev, 1, cw*ch*sizeof(pixel), f );
115 /* Fill "default" values */
116 static void x264_slice_header_init( x264_t *h, x264_slice_header_t *sh,
117 x264_sps_t *sps, x264_pps_t *pps,
118 int i_idr_pic_id, int i_frame, int i_qp )
120 x264_param_t *param = &h->param;
122 /* First we fill all fields */
127 sh->i_last_mb = h->mb.i_mb_count - 1;
128 sh->i_pps_id = pps->i_id;
130 sh->i_frame_num = i_frame;
132 sh->b_mbaff = PARAM_INTERLACED;
133 sh->b_field_pic = 0; /* no field support for now */
134 sh->b_bottom_field = 0; /* not yet used */
136 sh->i_idr_pic_id = i_idr_pic_id;
138 /* poc stuff, fixed later */
140 sh->i_delta_poc_bottom = 0;
141 sh->i_delta_poc[0] = 0;
142 sh->i_delta_poc[1] = 0;
144 sh->i_redundant_pic_cnt = 0;
146 h->mb.b_direct_auto_write = h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO
148 && ( h->param.rc.b_stat_write || !h->param.rc.b_stat_read );
150 if( !h->mb.b_direct_auto_read && sh->i_type == SLICE_TYPE_B )
152 if( h->fref[1][0]->i_poc_l0ref0 == h->fref[0][0]->i_poc )
154 if( h->mb.b_direct_auto_write )
155 sh->b_direct_spatial_mv_pred = ( h->stat.i_direct_score[1] > h->stat.i_direct_score[0] );
157 sh->b_direct_spatial_mv_pred = ( param->analyse.i_direct_mv_pred == X264_DIRECT_PRED_SPATIAL );
161 h->mb.b_direct_auto_write = 0;
162 sh->b_direct_spatial_mv_pred = 1;
165 /* else b_direct_spatial_mv_pred was read from the 2pass statsfile */
167 sh->b_num_ref_idx_override = 0;
168 sh->i_num_ref_idx_l0_active = 1;
169 sh->i_num_ref_idx_l1_active = 1;
171 sh->b_ref_pic_list_reordering[0] = h->b_ref_reorder[0];
172 sh->b_ref_pic_list_reordering[1] = h->b_ref_reorder[1];
174 /* If the ref list isn't in the default order, construct reordering header */
175 for( int list = 0; list < 2; list++ )
177 if( sh->b_ref_pic_list_reordering[list] )
179 int pred_frame_num = i_frame;
180 for( int i = 0; i < h->i_ref[list]; i++ )
182 int diff = h->fref[list][i]->i_frame_num - pred_frame_num;
183 sh->ref_pic_list_order[list][i].idc = ( diff > 0 );
184 sh->ref_pic_list_order[list][i].arg = (abs(diff) - 1) & ((1 << sps->i_log2_max_frame_num) - 1);
185 pred_frame_num = h->fref[list][i]->i_frame_num;
190 sh->i_cabac_init_idc = param->i_cabac_init_idc;
192 sh->i_qp = SPEC_QP(i_qp);
193 sh->i_qp_delta = sh->i_qp - pps->i_pic_init_qp;
194 sh->b_sp_for_swidth = 0;
197 int deblock_thresh = i_qp + 2 * X264_MIN(param->i_deblocking_filter_alphac0, param->i_deblocking_filter_beta);
198 /* If effective qp <= 15, deblocking would have no effect anyway */
199 if( param->b_deblocking_filter && (h->mb.b_variable_qp || 15 < deblock_thresh ) )
200 sh->i_disable_deblocking_filter_idc = param->b_sliced_threads ? 2 : 0;
202 sh->i_disable_deblocking_filter_idc = 1;
203 sh->i_alpha_c0_offset = param->i_deblocking_filter_alphac0 << 1;
204 sh->i_beta_offset = param->i_deblocking_filter_beta << 1;
207 static void x264_slice_header_write( bs_t *s, x264_slice_header_t *sh, int i_nal_ref_idc )
211 int first_x = sh->i_first_mb % sh->sps->i_mb_width;
212 int first_y = sh->i_first_mb / sh->sps->i_mb_width;
213 assert( (first_y&1) == 0 );
214 bs_write_ue( s, (2*first_x + sh->sps->i_mb_width*(first_y&~1) + (first_y&1)) >> 1 );
217 bs_write_ue( s, sh->i_first_mb );
219 bs_write_ue( s, sh->i_type + 5 ); /* same type things */
220 bs_write_ue( s, sh->i_pps_id );
221 bs_write( s, sh->sps->i_log2_max_frame_num, sh->i_frame_num & ((1<<sh->sps->i_log2_max_frame_num)-1) );
223 if( !sh->sps->b_frame_mbs_only )
225 bs_write1( s, sh->b_field_pic );
226 if( sh->b_field_pic )
227 bs_write1( s, sh->b_bottom_field );
230 if( sh->i_idr_pic_id >= 0 ) /* NAL IDR */
231 bs_write_ue( s, sh->i_idr_pic_id );
233 if( sh->sps->i_poc_type == 0 )
235 bs_write( s, sh->sps->i_log2_max_poc_lsb, sh->i_poc & ((1<<sh->sps->i_log2_max_poc_lsb)-1) );
236 if( sh->pps->b_pic_order && !sh->b_field_pic )
237 bs_write_se( s, sh->i_delta_poc_bottom );
240 if( sh->pps->b_redundant_pic_cnt )
241 bs_write_ue( s, sh->i_redundant_pic_cnt );
243 if( sh->i_type == SLICE_TYPE_B )
244 bs_write1( s, sh->b_direct_spatial_mv_pred );
246 if( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_B )
248 bs_write1( s, sh->b_num_ref_idx_override );
249 if( sh->b_num_ref_idx_override )
251 bs_write_ue( s, sh->i_num_ref_idx_l0_active - 1 );
252 if( sh->i_type == SLICE_TYPE_B )
253 bs_write_ue( s, sh->i_num_ref_idx_l1_active - 1 );
257 /* ref pic list reordering */
258 if( sh->i_type != SLICE_TYPE_I )
260 bs_write1( s, sh->b_ref_pic_list_reordering[0] );
261 if( sh->b_ref_pic_list_reordering[0] )
263 for( int i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
265 bs_write_ue( s, sh->ref_pic_list_order[0][i].idc );
266 bs_write_ue( s, sh->ref_pic_list_order[0][i].arg );
271 if( sh->i_type == SLICE_TYPE_B )
273 bs_write1( s, sh->b_ref_pic_list_reordering[1] );
274 if( sh->b_ref_pic_list_reordering[1] )
276 for( int i = 0; i < sh->i_num_ref_idx_l1_active; i++ )
278 bs_write_ue( s, sh->ref_pic_list_order[1][i].idc );
279 bs_write_ue( s, sh->ref_pic_list_order[1][i].arg );
285 sh->b_weighted_pred = 0;
286 if( sh->pps->b_weighted_pred && sh->i_type == SLICE_TYPE_P )
288 sh->b_weighted_pred = sh->weight[0][0].weightfn || sh->weight[0][1].weightfn || sh->weight[0][2].weightfn;
289 /* pred_weight_table() */
290 bs_write_ue( s, sh->weight[0][0].i_denom );
291 bs_write_ue( s, sh->weight[0][1].i_denom );
292 for( int i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
294 int luma_weight_l0_flag = !!sh->weight[i][0].weightfn;
295 int chroma_weight_l0_flag = !!sh->weight[i][1].weightfn || !!sh->weight[i][2].weightfn;
296 bs_write1( s, luma_weight_l0_flag );
297 if( luma_weight_l0_flag )
299 bs_write_se( s, sh->weight[i][0].i_scale );
300 bs_write_se( s, sh->weight[i][0].i_offset );
302 bs_write1( s, chroma_weight_l0_flag );
303 if( chroma_weight_l0_flag )
305 for( int j = 1; j < 3; j++ )
307 bs_write_se( s, sh->weight[i][j].i_scale );
308 bs_write_se( s, sh->weight[i][j].i_offset );
313 else if( sh->pps->b_weighted_bipred == 1 && sh->i_type == SLICE_TYPE_B )
318 if( i_nal_ref_idc != 0 )
320 if( sh->i_idr_pic_id >= 0 )
322 bs_write1( s, 0 ); /* no output of prior pics flag */
323 bs_write1( s, 0 ); /* long term reference flag */
327 bs_write1( s, sh->i_mmco_command_count > 0 ); /* adaptive_ref_pic_marking_mode_flag */
328 if( sh->i_mmco_command_count > 0 )
330 for( int i = 0; i < sh->i_mmco_command_count; i++ )
332 bs_write_ue( s, 1 ); /* mark short term ref as unused */
333 bs_write_ue( s, sh->mmco[i].i_difference_of_pic_nums - 1 );
335 bs_write_ue( s, 0 ); /* end command list */
340 if( sh->pps->b_cabac && sh->i_type != SLICE_TYPE_I )
341 bs_write_ue( s, sh->i_cabac_init_idc );
343 bs_write_se( s, sh->i_qp_delta ); /* slice qp delta */
345 if( sh->pps->b_deblocking_filter_control )
347 bs_write_ue( s, sh->i_disable_deblocking_filter_idc );
348 if( sh->i_disable_deblocking_filter_idc != 1 )
350 bs_write_se( s, sh->i_alpha_c0_offset >> 1 );
351 bs_write_se( s, sh->i_beta_offset >> 1 );
356 /* If we are within a reasonable distance of the end of the memory allocated for the bitstream, */
357 /* reallocate, adding an arbitrary amount of space. */
358 static int x264_bitstream_check_buffer_internal( x264_t *h, int size, int b_cabac, int i_nal )
360 if( (b_cabac && (h->cabac.p_end - h->cabac.p < size)) ||
361 (h->out.bs.p_end - h->out.bs.p < size) )
363 int buf_size = h->out.i_bitstream + size;
364 uint8_t *buf = x264_malloc( buf_size );
367 int aligned_size = h->out.i_bitstream & ~15;
368 h->mc.memcpy_aligned( buf, h->out.p_bitstream, aligned_size );
369 memcpy( buf + aligned_size, h->out.p_bitstream + aligned_size, h->out.i_bitstream - aligned_size );
371 intptr_t delta = buf - h->out.p_bitstream;
373 h->out.bs.p_start += delta;
374 h->out.bs.p += delta;
375 h->out.bs.p_end = buf + buf_size;
377 h->cabac.p_start += delta;
379 h->cabac.p_end = buf + buf_size;
381 for( int i = 0; i <= i_nal; i++ )
382 h->out.nal[i].p_payload += delta;
384 x264_free( h->out.p_bitstream );
385 h->out.p_bitstream = buf;
386 h->out.i_bitstream = buf_size;
391 static int x264_bitstream_check_buffer( x264_t *h )
393 int max_row_size = (2500 << SLICE_MBAFF) * h->mb.i_mb_width;
394 return x264_bitstream_check_buffer_internal( h, max_row_size, h->param.b_cabac, h->out.i_nal );
397 static int x264_bitstream_check_buffer_filler( x264_t *h, int filler )
399 filler += 32; // add padding for safety
400 return x264_bitstream_check_buffer_internal( h, filler, 0, -1 );
404 static void x264_encoder_thread_init( x264_t *h )
406 if( h->param.i_sync_lookahead )
407 x264_lower_thread_priority( 10 );
411 /****************************************************************************
413 ****************************************************************************
414 ****************************** External API*********************************
415 ****************************************************************************
417 ****************************************************************************/
419 static int x264_validate_parameters( x264_t *h, int b_open )
421 if( !h->param.pf_log )
423 x264_log( NULL, X264_LOG_ERROR, "pf_log not set! did you forget to call x264_param_default?\n" );
430 int cpuflags = x264_cpu_detect();
433 if( !(cpuflags & X264_CPU_SSE) )
435 x264_log( h, X264_LOG_ERROR, "your cpu does not support SSE1, but x264 was compiled with asm\n");
439 if( !(cpuflags & X264_CPU_MMX2) )
441 x264_log( h, X264_LOG_ERROR, "your cpu does not support MMXEXT, but x264 was compiled with asm\n");
445 if( !fail && !(cpuflags & X264_CPU_CMOV) )
447 x264_log( h, X264_LOG_ERROR, "your cpu does not support CMOV, but x264 was compiled with asm\n");
452 x264_log( h, X264_LOG_ERROR, "to run x264, recompile without asm (configure --disable-asm)\n");
459 h->param.b_interlaced = !!PARAM_INTERLACED;
461 if( h->param.b_interlaced )
463 x264_log( h, X264_LOG_ERROR, "not compiled with interlaced support\n" );
468 if( h->param.i_width <= 0 || h->param.i_height <= 0 )
470 x264_log( h, X264_LOG_ERROR, "invalid width x height (%dx%d)\n",
471 h->param.i_width, h->param.i_height );
475 int i_csp = h->param.i_csp & X264_CSP_MASK;
476 #if X264_CHROMA_FORMAT
477 if( CHROMA_FORMAT != CHROMA_420 && i_csp >= X264_CSP_I420 && i_csp < X264_CSP_I422 )
479 x264_log( h, X264_LOG_ERROR, "not compiled with 4:2:0 support\n" );
482 else if( CHROMA_FORMAT != CHROMA_422 && i_csp >= X264_CSP_I422 && i_csp < X264_CSP_I444 )
484 x264_log( h, X264_LOG_ERROR, "not compiled with 4:2:2 support\n" );
487 else if( CHROMA_FORMAT != CHROMA_444 && i_csp >= X264_CSP_I444 && i_csp <= X264_CSP_RGB )
489 x264_log( h, X264_LOG_ERROR, "not compiled with 4:4:4 support\n" );
493 if( i_csp <= X264_CSP_NONE || i_csp >= X264_CSP_MAX )
495 x264_log( h, X264_LOG_ERROR, "invalid CSP (only I420/YV12/NV12/NV21/I422/YV16/NV16/I444/YV24/BGR/BGRA/RGB supported)\n" );
499 int w_mod = i_csp < X264_CSP_I444 ? 2 : 1;
500 int h_mod = (i_csp < X264_CSP_I422 ? 2 : 1) << PARAM_INTERLACED;
501 if( h->param.i_width % w_mod )
503 x264_log( h, X264_LOG_ERROR, "width not divisible by %d (%dx%d)\n",
504 w_mod, h->param.i_width, h->param.i_height );
507 if( h->param.i_height % h_mod )
509 x264_log( h, X264_LOG_ERROR, "height not divisible by %d (%dx%d)\n",
510 h_mod, h->param.i_width, h->param.i_height );
514 if( h->param.crop_rect.i_left >= h->param.i_width ||
515 h->param.crop_rect.i_right >= h->param.i_width ||
516 h->param.crop_rect.i_top >= h->param.i_height ||
517 h->param.crop_rect.i_bottom >= h->param.i_height ||
518 h->param.crop_rect.i_left + h->param.crop_rect.i_right >= h->param.i_width ||
519 h->param.crop_rect.i_top + h->param.crop_rect.i_bottom >= h->param.i_height )
521 x264_log( h, X264_LOG_ERROR, "invalid crop-rect %u,%u,%u,%u\n", h->param.crop_rect.i_left,
522 h->param.crop_rect.i_top, h->param.crop_rect.i_right, h->param.crop_rect.i_bottom );
525 if( h->param.crop_rect.i_left % w_mod || h->param.crop_rect.i_right % w_mod ||
526 h->param.crop_rect.i_top % h_mod || h->param.crop_rect.i_bottom % h_mod )
528 x264_log( h, X264_LOG_ERROR, "crop-rect %u,%u,%u,%u not divisible by %dx%d\n", h->param.crop_rect.i_left,
529 h->param.crop_rect.i_top, h->param.crop_rect.i_right, h->param.crop_rect.i_bottom, w_mod, h_mod );
533 if( h->param.vui.i_sar_width <= 0 || h->param.vui.i_sar_height <= 0 )
535 h->param.vui.i_sar_width = 0;
536 h->param.vui.i_sar_height = 0;
539 if( h->param.i_threads == X264_THREADS_AUTO )
541 h->param.i_threads = x264_cpu_num_processors() * (h->param.b_sliced_threads?2:3)/2;
542 /* Avoid too many threads as they don't improve performance and
543 * complicate VBV. Capped at an arbitrary 2 rows per thread. */
544 int max_threads = X264_MAX( 1, (h->param.i_height+15)/16 / 2 );
545 h->param.i_threads = X264_MIN( h->param.i_threads, max_threads );
547 int max_sliced_threads = X264_MAX( 1, (h->param.i_height+15)/16 / 4 );
548 if( h->param.i_threads > 1 )
551 x264_log( h, X264_LOG_WARNING, "not compiled with thread support!\n");
552 h->param.i_threads = 1;
554 /* Avoid absurdly small thread slices as they can reduce performance
555 * and VBV compliance. Capped at an arbitrary 4 rows per thread. */
556 if( h->param.b_sliced_threads )
557 h->param.i_threads = X264_MIN( h->param.i_threads, max_sliced_threads );
559 h->param.i_threads = x264_clip3( h->param.i_threads, 1, X264_THREAD_MAX );
560 if( h->param.i_threads == 1 )
562 h->param.b_sliced_threads = 0;
563 h->param.i_lookahead_threads = 1;
565 h->i_thread_frames = h->param.b_sliced_threads ? 1 : h->param.i_threads;
566 if( h->i_thread_frames > 1 )
567 h->param.nalu_process = NULL;
569 if( h->param.b_opencl )
572 x264_log( h, X264_LOG_WARNING, "OpenCL: not compiled with OpenCL support, disabling\n" );
573 h->param.b_opencl = 0;
575 x264_log( h, X264_LOG_WARNING, "OpenCL lookahead does not support high bit depth, disabling opencl\n" );
576 h->param.b_opencl = 0;
578 if( h->param.i_width < 32 || h->param.i_height < 32 )
580 x264_log( h, X264_LOG_WARNING, "OpenCL: frame size is too small, disabling opencl\n" );
581 h->param.b_opencl = 0;
584 if( h->param.opencl_device_id && h->param.i_opencl_device )
586 x264_log( h, X264_LOG_WARNING, "OpenCL: device id and device skip count configured; dropping skip\n" );
587 h->param.i_opencl_device = 0;
591 h->param.i_keyint_max = x264_clip3( h->param.i_keyint_max, 1, X264_KEYINT_MAX_INFINITE );
592 if( h->param.i_keyint_max == 1 )
594 h->param.b_intra_refresh = 0;
595 h->param.analyse.i_weighted_pred = 0;
596 h->param.i_frame_reference = 1;
597 h->param.i_dpb_size = 1;
600 if( h->param.i_frame_packing < -1 || h->param.i_frame_packing > 7 )
602 x264_log( h, X264_LOG_WARNING, "ignoring unknown frame packing value\n" );
603 h->param.i_frame_packing = -1;
605 if( h->param.i_frame_packing == 7 &&
606 ((h->param.i_width - h->param.crop_rect.i_left - h->param.crop_rect.i_right) % 3 ||
607 (h->param.i_height - h->param.crop_rect.i_top - h->param.crop_rect.i_bottom) % 3) )
609 x264_log( h, X264_LOG_ERROR, "cropped resolution %dx%d not compatible with tile format frame packing\n",
610 h->param.i_width - h->param.crop_rect.i_left - h->param.crop_rect.i_right,
611 h->param.i_height - h->param.crop_rect.i_top - h->param.crop_rect.i_bottom );
615 /* Detect default ffmpeg settings and terminate with an error. */
619 score += h->param.analyse.i_me_range == 0;
620 score += h->param.rc.i_qp_step == 3;
621 score += h->param.i_keyint_max == 12;
622 score += h->param.rc.i_qp_min == 2;
623 score += h->param.rc.i_qp_max == 31;
624 score += h->param.rc.f_qcompress == 0.5;
625 score += fabs(h->param.rc.f_ip_factor - 1.25) < 0.01;
626 score += fabs(h->param.rc.f_pb_factor - 1.25) < 0.01;
627 score += h->param.analyse.inter == 0 && h->param.analyse.i_subpel_refine == 8;
630 x264_log( h, X264_LOG_ERROR, "broken ffmpeg default settings detected\n" );
631 x264_log( h, X264_LOG_ERROR, "use an encoding preset (e.g. -vpre medium)\n" );
632 x264_log( h, X264_LOG_ERROR, "preset usage: -vpre <speed> -vpre <profile>\n" );
633 x264_log( h, X264_LOG_ERROR, "speed presets are listed in x264 --help\n" );
634 x264_log( h, X264_LOG_ERROR, "profile is optional; x264 defaults to high\n" );
639 if( h->param.rc.i_rc_method < 0 || h->param.rc.i_rc_method > 2 )
641 x264_log( h, X264_LOG_ERROR, "no ratecontrol method specified\n" );
645 if( PARAM_INTERLACED )
646 h->param.b_pic_struct = 1;
648 if( h->param.i_avcintra_class )
650 if( BIT_DEPTH != 10 )
652 x264_log( h, X264_LOG_ERROR, "%2d-bit AVC-Intra is not widely compatible\n", BIT_DEPTH );
653 x264_log( h, X264_LOG_ERROR, "10-bit x264 is required to encode AVC-Intra\n" );
657 int type = h->param.i_avcintra_class == 200 ? 2 :
658 h->param.i_avcintra_class == 100 ? 1 :
659 h->param.i_avcintra_class == 50 ? 0 : -1;
662 x264_log( h, X264_LOG_ERROR, "Invalid AVC-Intra class\n" );
666 /* [50/100/200][res][fps] */
673 const uint8_t *cqm_4ic;
674 const uint8_t *cqm_8iy;
675 } avcintra_lut[3][2][7] =
677 {{{ 60000, 1001, 0, 912, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
678 { 50, 1, 0, 1100, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
679 { 30000, 1001, 0, 912, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
680 { 25, 1, 0, 1100, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
681 { 24000, 1001, 0, 912, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy }},
682 {{ 30000, 1001, 1, 1820, x264_cqm_avci50_4ic, x264_cqm_avci50_1080i_8iy },
683 { 25, 1, 1, 2196, x264_cqm_avci50_4ic, x264_cqm_avci50_1080i_8iy },
684 { 60000, 1001, 0, 1820, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
685 { 30000, 1001, 0, 1820, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
686 { 50, 1, 0, 2196, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
687 { 25, 1, 0, 2196, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
688 { 24000, 1001, 0, 1820, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy }}},
689 {{{ 60000, 1001, 0, 1848, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
690 { 50, 1, 0, 2224, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
691 { 30000, 1001, 0, 1848, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
692 { 25, 1, 0, 2224, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
693 { 24000, 1001, 0, 1848, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy }},
694 {{ 30000, 1001, 1, 3692, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080i_8iy },
695 { 25, 1, 1, 4444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080i_8iy },
696 { 60000, 1001, 0, 3692, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
697 { 30000, 1001, 0, 3692, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
698 { 50, 1, 0, 4444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
699 { 25, 1, 0, 4444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
700 { 24000, 1001, 0, 3692, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy }}},
701 {{{ 60000, 1001, 0, 3724, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
702 { 50, 1, 0, 4472, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy }},
703 {{ 30000, 1001, 1, 7444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080i_8iy },
704 { 25, 1, 1, 8940, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080i_8iy },
705 { 60000, 1001, 0, 7444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
706 { 30000, 1001, 0, 7444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
707 { 50, 1, 0, 8940, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
708 { 25, 1, 0, 8940, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
709 { 24000, 1001, 0, 7444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy }}}
713 if( i_csp >= X264_CSP_I420 && i_csp < X264_CSP_I422 && !type )
715 if( h->param.i_width == 1440 && h->param.i_height == 1080 ) res = 1;
716 else if( h->param.i_width == 960 && h->param.i_height == 720 ) res = 0;
718 else if( i_csp >= X264_CSP_I422 && i_csp < X264_CSP_I444 && type )
720 if( h->param.i_width == 1920 && h->param.i_height == 1080 ) res = 1;
721 else if( h->param.i_width == 1280 && h->param.i_height == 720 ) res = 0;
725 x264_log( h, X264_LOG_ERROR, "Invalid colorspace for AVC-Intra %d\n", h->param.i_avcintra_class );
731 x264_log( h, X264_LOG_ERROR, "Resolution %dx%d invalid for AVC-Intra %d\n",
732 h->param.i_width, h->param.i_height, h->param.i_avcintra_class );
736 if( h->param.nalu_process )
738 x264_log( h, X264_LOG_ERROR, "nalu_process is not supported in AVC-Intra mode\n" );
742 if( !h->param.b_repeat_headers )
744 x264_log( h, X264_LOG_ERROR, "Separate headers not supported in AVC-Intra mode\n" );
749 uint32_t fps_num = h->param.i_fps_num, fps_den = h->param.i_fps_den;
750 x264_reduce_fraction( &fps_num, &fps_den );
751 for( i = 0; i < 7; i++ )
753 if( avcintra_lut[type][res][i].fps_num == fps_num &&
754 avcintra_lut[type][res][i].fps_den == fps_den &&
755 avcintra_lut[type][res][i].interlaced == PARAM_INTERLACED )
762 x264_log( h, X264_LOG_ERROR, "FPS %d/%d%c not compatible with AVC-Intra\n",
763 h->param.i_fps_num, h->param.i_fps_den, PARAM_INTERLACED ? 'i' : 'p' );
767 h->param.i_keyint_max = 1;
768 h->param.b_intra_refresh = 0;
769 h->param.analyse.i_weighted_pred = 0;
770 h->param.i_frame_reference = 1;
771 h->param.i_dpb_size = 1;
773 h->param.b_bluray_compat = 0;
774 h->param.b_vfr_input = 0;
776 h->param.vui.i_chroma_loc = 0;
777 h->param.i_nal_hrd = X264_NAL_HRD_NONE;
778 h->param.b_deblocking_filter = 0;
779 h->param.b_stitchable = 1;
780 h->param.b_pic_struct = 0;
781 h->param.analyse.b_transform_8x8 = 1;
782 h->param.analyse.intra = X264_ANALYSE_I8x8;
783 h->param.analyse.i_chroma_qp_offset = res && type ? 3 : 4;
784 h->param.b_cabac = !type;
785 h->param.rc.i_vbv_buffer_size = avcintra_lut[type][res][i].frame_size;
786 h->param.rc.i_vbv_max_bitrate =
787 h->param.rc.i_bitrate = h->param.rc.i_vbv_buffer_size * fps_num / fps_den;
788 h->param.rc.i_rc_method = X264_RC_ABR;
789 h->param.rc.f_vbv_buffer_init = 1.0;
790 h->param.rc.b_filler = 1;
791 h->param.i_cqm_preset = X264_CQM_CUSTOM;
792 memcpy( h->param.cqm_4iy, x264_cqm_jvt4i, sizeof(h->param.cqm_4iy) );
793 memcpy( h->param.cqm_4ic, avcintra_lut[type][res][i].cqm_4ic, sizeof(h->param.cqm_4ic) );
794 memcpy( h->param.cqm_8iy, avcintra_lut[type][res][i].cqm_8iy, sizeof(h->param.cqm_8iy) );
796 /* Need exactly 10 slices of equal MB count... why? $deity knows... */
797 h->param.i_slice_max_mbs = ((h->param.i_width + 15) / 16) * ((h->param.i_height + 15) / 16) / 10;
798 h->param.i_slice_max_size = 0;
799 /* The slice structure only allows a maximum of 2 threads for 1080i/p
800 * and 1 or 5 threads for 720p */
801 if( h->param.b_sliced_threads )
804 h->param.i_threads = X264_MIN( 2, h->param.i_threads );
807 h->param.i_threads = X264_MIN( 5, h->param.i_threads );
808 if( h->param.i_threads < 5 )
809 h->param.i_threads = 1;
814 h->param.vui.i_sar_width = h->param.vui.i_sar_height = 1;
817 h->param.vui.i_sar_width = 4;
818 h->param.vui.i_sar_height = 3;
821 /* Official encoder doesn't appear to go under 13
822 * and Avid cannot handle negative QPs */
823 h->param.rc.i_qp_min = X264_MAX( h->param.rc.i_qp_min, QP_BD_OFFSET + 1 );
826 h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, -QP_BD_OFFSET, 51 );
827 h->param.rc.f_rf_constant_max = x264_clip3f( h->param.rc.f_rf_constant_max, -QP_BD_OFFSET, 51 );
828 h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, 0, QP_MAX );
829 h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 11 );
830 h->param.rc.f_ip_factor = X264_MAX( h->param.rc.f_ip_factor, 0.01f );
831 h->param.rc.f_pb_factor = X264_MAX( h->param.rc.f_pb_factor, 0.01f );
832 if( h->param.rc.i_rc_method == X264_RC_CRF )
834 h->param.rc.i_qp_constant = h->param.rc.f_rf_constant + QP_BD_OFFSET;
835 h->param.rc.i_bitrate = 0;
837 if( b_open && (h->param.rc.i_rc_method == X264_RC_CQP || h->param.rc.i_rc_method == X264_RC_CRF)
838 && h->param.rc.i_qp_constant == 0 )
840 h->mb.b_lossless = 1;
841 h->param.i_cqm_preset = X264_CQM_FLAT;
842 h->param.psz_cqm_file = NULL;
843 h->param.rc.i_rc_method = X264_RC_CQP;
844 h->param.rc.f_ip_factor = 1;
845 h->param.rc.f_pb_factor = 1;
846 h->param.analyse.b_psnr = 0;
847 h->param.analyse.b_ssim = 0;
848 h->param.analyse.i_chroma_qp_offset = 0;
849 h->param.analyse.i_trellis = 0;
850 h->param.analyse.b_fast_pskip = 0;
851 h->param.analyse.i_noise_reduction = 0;
852 h->param.analyse.b_psy = 0;
853 h->param.i_bframe = 0;
854 /* 8x8dct is not useful without RD in CAVLC lossless */
855 if( !h->param.b_cabac && h->param.analyse.i_subpel_refine < 6 )
856 h->param.analyse.b_transform_8x8 = 0;
857 h->param.analyse.inter &= ~X264_ANALYSE_I8x8;
858 h->param.analyse.intra &= ~X264_ANALYSE_I8x8;
860 if( h->param.rc.i_rc_method == X264_RC_CQP )
862 float qp_p = h->param.rc.i_qp_constant;
863 float qp_i = qp_p - 6*log2f( h->param.rc.f_ip_factor );
864 float qp_b = qp_p + 6*log2f( h->param.rc.f_pb_factor );
865 h->param.rc.i_qp_min = x264_clip3( (int)(X264_MIN3( qp_p, qp_i, qp_b )), 0, QP_MAX );
866 h->param.rc.i_qp_max = x264_clip3( (int)(X264_MAX3( qp_p, qp_i, qp_b ) + .999), 0, QP_MAX );
867 h->param.rc.i_aq_mode = 0;
868 h->param.rc.b_mb_tree = 0;
869 h->param.rc.i_bitrate = 0;
871 h->param.rc.i_qp_max = x264_clip3( h->param.rc.i_qp_max, 0, QP_MAX );
872 h->param.rc.i_qp_min = x264_clip3( h->param.rc.i_qp_min, 0, h->param.rc.i_qp_max );
873 h->param.rc.i_qp_step = x264_clip3( h->param.rc.i_qp_step, 2, QP_MAX );
874 h->param.rc.i_bitrate = x264_clip3( h->param.rc.i_bitrate, 0, 2000000 );
875 if( h->param.rc.i_rc_method == X264_RC_ABR && !h->param.rc.i_bitrate )
877 x264_log( h, X264_LOG_ERROR, "bitrate not specified\n" );
880 h->param.rc.i_vbv_buffer_size = x264_clip3( h->param.rc.i_vbv_buffer_size, 0, 2000000 );
881 h->param.rc.i_vbv_max_bitrate = x264_clip3( h->param.rc.i_vbv_max_bitrate, 0, 2000000 );
882 h->param.rc.f_vbv_buffer_init = x264_clip3f( h->param.rc.f_vbv_buffer_init, 0, 2000000 );
883 if( h->param.rc.i_vbv_buffer_size )
885 if( h->param.rc.i_rc_method == X264_RC_CQP )
887 x264_log( h, X264_LOG_WARNING, "VBV is incompatible with constant QP, ignored.\n" );
888 h->param.rc.i_vbv_max_bitrate = 0;
889 h->param.rc.i_vbv_buffer_size = 0;
891 else if( h->param.rc.i_vbv_max_bitrate == 0 )
893 if( h->param.rc.i_rc_method == X264_RC_ABR )
895 x264_log( h, X264_LOG_WARNING, "VBV maxrate unspecified, assuming CBR\n" );
896 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
900 x264_log( h, X264_LOG_WARNING, "VBV bufsize set but maxrate unspecified, ignored\n" );
901 h->param.rc.i_vbv_buffer_size = 0;
904 else if( h->param.rc.i_vbv_max_bitrate < h->param.rc.i_bitrate &&
905 h->param.rc.i_rc_method == X264_RC_ABR )
907 x264_log( h, X264_LOG_WARNING, "max bitrate less than average bitrate, assuming CBR\n" );
908 h->param.rc.i_bitrate = h->param.rc.i_vbv_max_bitrate;
911 else if( h->param.rc.i_vbv_max_bitrate )
913 x264_log( h, X264_LOG_WARNING, "VBV maxrate specified, but no bufsize, ignored\n" );
914 h->param.rc.i_vbv_max_bitrate = 0;
917 h->param.i_slice_max_size = X264_MAX( h->param.i_slice_max_size, 0 );
918 h->param.i_slice_max_mbs = X264_MAX( h->param.i_slice_max_mbs, 0 );
919 h->param.i_slice_min_mbs = X264_MAX( h->param.i_slice_min_mbs, 0 );
920 if( h->param.i_slice_max_mbs )
921 h->param.i_slice_min_mbs = X264_MIN( h->param.i_slice_min_mbs, h->param.i_slice_max_mbs/2 );
922 else if( !h->param.i_slice_max_size )
923 h->param.i_slice_min_mbs = 0;
924 if( PARAM_INTERLACED && h->param.i_slice_min_mbs )
926 x264_log( h, X264_LOG_WARNING, "interlace + slice-min-mbs is not implemented\n" );
927 h->param.i_slice_min_mbs = 0;
929 int mb_width = (h->param.i_width+15)/16;
930 if( h->param.i_slice_min_mbs > mb_width )
932 x264_log( h, X264_LOG_WARNING, "slice-min-mbs > row mb size (%d) not implemented\n", mb_width );
933 h->param.i_slice_min_mbs = mb_width;
936 int max_slices = (h->param.i_height+((16<<PARAM_INTERLACED)-1))/(16<<PARAM_INTERLACED);
937 if( h->param.b_sliced_threads )
938 h->param.i_slice_count = x264_clip3( h->param.i_threads, 0, max_slices );
941 h->param.i_slice_count = x264_clip3( h->param.i_slice_count, 0, max_slices );
942 if( h->param.i_slice_max_mbs || h->param.i_slice_max_size )
943 h->param.i_slice_count = 0;
945 if( h->param.i_slice_count_max > 0 )
946 h->param.i_slice_count_max = X264_MAX( h->param.i_slice_count, h->param.i_slice_count_max );
948 if( h->param.b_bluray_compat )
950 h->param.i_bframe_pyramid = X264_MIN( X264_B_PYRAMID_STRICT, h->param.i_bframe_pyramid );
951 h->param.i_bframe = X264_MIN( h->param.i_bframe, 3 );
953 h->param.i_nal_hrd = X264_MAX( h->param.i_nal_hrd, X264_NAL_HRD_VBR );
954 h->param.i_slice_max_size = 0;
955 h->param.i_slice_max_mbs = 0;
956 h->param.b_intra_refresh = 0;
957 h->param.i_frame_reference = X264_MIN( h->param.i_frame_reference, 6 );
958 h->param.i_dpb_size = X264_MIN( h->param.i_dpb_size, 6 );
959 /* Don't use I-frames, because Blu-ray treats them the same as IDR. */
960 h->param.i_keyint_min = 1;
961 /* Due to the proliferation of broken players that don't handle dupes properly. */
962 h->param.analyse.i_weighted_pred = X264_MIN( h->param.analyse.i_weighted_pred, X264_WEIGHTP_SIMPLE );
963 if( h->param.b_fake_interlaced )
964 h->param.b_pic_struct = 1;
967 h->param.i_frame_reference = x264_clip3( h->param.i_frame_reference, 1, X264_REF_MAX );
968 h->param.i_dpb_size = x264_clip3( h->param.i_dpb_size, 1, X264_REF_MAX );
969 if( h->param.i_scenecut_threshold < 0 )
970 h->param.i_scenecut_threshold = 0;
971 h->param.analyse.i_direct_mv_pred = x264_clip3( h->param.analyse.i_direct_mv_pred, X264_DIRECT_PRED_NONE, X264_DIRECT_PRED_AUTO );
972 if( !h->param.analyse.i_subpel_refine && h->param.analyse.i_direct_mv_pred > X264_DIRECT_PRED_SPATIAL )
974 x264_log( h, X264_LOG_WARNING, "subme=0 + direct=temporal is not supported\n" );
975 h->param.analyse.i_direct_mv_pred = X264_DIRECT_PRED_SPATIAL;
977 h->param.i_bframe = x264_clip3( h->param.i_bframe, 0, X264_MIN( X264_BFRAME_MAX, h->param.i_keyint_max-1 ) );
978 h->param.i_bframe_bias = x264_clip3( h->param.i_bframe_bias, -90, 100 );
979 if( h->param.i_bframe <= 1 )
980 h->param.i_bframe_pyramid = X264_B_PYRAMID_NONE;
981 h->param.i_bframe_pyramid = x264_clip3( h->param.i_bframe_pyramid, X264_B_PYRAMID_NONE, X264_B_PYRAMID_NORMAL );
982 h->param.i_bframe_adaptive = x264_clip3( h->param.i_bframe_adaptive, X264_B_ADAPT_NONE, X264_B_ADAPT_TRELLIS );
983 if( !h->param.i_bframe )
985 h->param.i_bframe_adaptive = X264_B_ADAPT_NONE;
986 h->param.analyse.i_direct_mv_pred = 0;
987 h->param.analyse.b_weighted_bipred = 0;
988 h->param.b_open_gop = 0;
990 if( h->param.b_intra_refresh && h->param.i_bframe_pyramid == X264_B_PYRAMID_NORMAL )
992 x264_log( h, X264_LOG_WARNING, "b-pyramid normal + intra-refresh is not supported\n" );
993 h->param.i_bframe_pyramid = X264_B_PYRAMID_STRICT;
995 if( h->param.b_intra_refresh && (h->param.i_frame_reference > 1 || h->param.i_dpb_size > 1) )
997 x264_log( h, X264_LOG_WARNING, "ref > 1 + intra-refresh is not supported\n" );
998 h->param.i_frame_reference = 1;
999 h->param.i_dpb_size = 1;
1001 if( h->param.b_intra_refresh && h->param.b_open_gop )
1003 x264_log( h, X264_LOG_WARNING, "intra-refresh is not compatible with open-gop\n" );
1004 h->param.b_open_gop = 0;
1006 if( !h->param.i_fps_num || !h->param.i_fps_den )
1008 h->param.i_fps_num = 25;
1009 h->param.i_fps_den = 1;
1011 float fps = (float) h->param.i_fps_num / h->param.i_fps_den;
1012 if( h->param.i_keyint_min == X264_KEYINT_MIN_AUTO )
1013 h->param.i_keyint_min = X264_MIN( h->param.i_keyint_max / 10, fps );
1014 h->param.i_keyint_min = x264_clip3( h->param.i_keyint_min, 1, h->param.i_keyint_max/2+1 );
1015 h->param.rc.i_lookahead = x264_clip3( h->param.rc.i_lookahead, 0, X264_LOOKAHEAD_MAX );
1017 int maxrate = X264_MAX( h->param.rc.i_vbv_max_bitrate, h->param.rc.i_bitrate );
1018 float bufsize = maxrate ? (float)h->param.rc.i_vbv_buffer_size / maxrate : 0;
1019 h->param.rc.i_lookahead = X264_MIN( h->param.rc.i_lookahead, X264_MAX( h->param.i_keyint_max, bufsize*fps ) );
1022 if( !h->param.i_timebase_num || !h->param.i_timebase_den || !(h->param.b_vfr_input || h->param.b_pulldown) )
1024 h->param.i_timebase_num = h->param.i_fps_den;
1025 h->param.i_timebase_den = h->param.i_fps_num;
1028 h->param.rc.f_qcompress = x264_clip3f( h->param.rc.f_qcompress, 0.0, 1.0 );
1029 if( h->param.i_keyint_max == 1 || h->param.rc.f_qcompress == 1 )
1030 h->param.rc.b_mb_tree = 0;
1031 if( (!h->param.b_intra_refresh && h->param.i_keyint_max != X264_KEYINT_MAX_INFINITE) &&
1032 !h->param.rc.i_lookahead && h->param.rc.b_mb_tree )
1034 x264_log( h, X264_LOG_WARNING, "lookaheadless mb-tree requires intra refresh or infinite keyint\n" );
1035 h->param.rc.b_mb_tree = 0;
1037 if( b_open && h->param.rc.b_stat_read )
1038 h->param.rc.i_lookahead = 0;
1040 if( h->param.i_sync_lookahead < 0 )
1041 h->param.i_sync_lookahead = h->param.i_bframe + 1;
1042 h->param.i_sync_lookahead = X264_MIN( h->param.i_sync_lookahead, X264_LOOKAHEAD_MAX );
1043 if( h->param.rc.b_stat_read || h->i_thread_frames == 1 )
1044 h->param.i_sync_lookahead = 0;
1046 h->param.i_sync_lookahead = 0;
1049 h->param.i_deblocking_filter_alphac0 = x264_clip3( h->param.i_deblocking_filter_alphac0, -6, 6 );
1050 h->param.i_deblocking_filter_beta = x264_clip3( h->param.i_deblocking_filter_beta, -6, 6 );
1051 h->param.analyse.i_luma_deadzone[0] = x264_clip3( h->param.analyse.i_luma_deadzone[0], 0, 32 );
1052 h->param.analyse.i_luma_deadzone[1] = x264_clip3( h->param.analyse.i_luma_deadzone[1], 0, 32 );
1054 h->param.i_cabac_init_idc = x264_clip3( h->param.i_cabac_init_idc, 0, 2 );
1056 if( h->param.i_cqm_preset < X264_CQM_FLAT || h->param.i_cqm_preset > X264_CQM_CUSTOM )
1057 h->param.i_cqm_preset = X264_CQM_FLAT;
1059 if( h->param.analyse.i_me_method < X264_ME_DIA ||
1060 h->param.analyse.i_me_method > X264_ME_TESA )
1061 h->param.analyse.i_me_method = X264_ME_HEX;
1062 h->param.analyse.i_me_range = x264_clip3( h->param.analyse.i_me_range, 4, 1024 );
1063 if( h->param.analyse.i_me_range > 16 && h->param.analyse.i_me_method <= X264_ME_HEX )
1064 h->param.analyse.i_me_range = 16;
1065 if( h->param.analyse.i_me_method == X264_ME_TESA &&
1066 (h->mb.b_lossless || h->param.analyse.i_subpel_refine <= 1) )
1067 h->param.analyse.i_me_method = X264_ME_ESA;
1068 h->param.analyse.b_mixed_references = h->param.analyse.b_mixed_references && h->param.i_frame_reference > 1;
1069 h->param.analyse.inter &= X264_ANALYSE_PSUB16x16|X264_ANALYSE_PSUB8x8|X264_ANALYSE_BSUB16x16|
1070 X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
1071 h->param.analyse.intra &= X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
1072 if( !(h->param.analyse.inter & X264_ANALYSE_PSUB16x16) )
1073 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
1074 if( !h->param.analyse.b_transform_8x8 )
1076 h->param.analyse.inter &= ~X264_ANALYSE_I8x8;
1077 h->param.analyse.intra &= ~X264_ANALYSE_I8x8;
1079 h->param.analyse.i_trellis = x264_clip3( h->param.analyse.i_trellis, 0, 2 );
1080 h->param.rc.i_aq_mode = x264_clip3( h->param.rc.i_aq_mode, 0, 3 );
1081 h->param.rc.f_aq_strength = x264_clip3f( h->param.rc.f_aq_strength, 0, 3 );
1082 if( h->param.rc.f_aq_strength == 0 )
1083 h->param.rc.i_aq_mode = 0;
1085 if( h->param.i_log_level < X264_LOG_INFO )
1087 h->param.analyse.b_psnr = 0;
1088 h->param.analyse.b_ssim = 0;
1090 /* Warn users trying to measure PSNR/SSIM with psy opts on. */
1091 if( b_open && (h->param.analyse.b_psnr || h->param.analyse.b_ssim) )
1095 if( h->param.analyse.b_psy )
1097 s = h->param.analyse.b_psnr ? "psnr" : "ssim";
1098 x264_log( h, X264_LOG_WARNING, "--%s used with psy on: results will be invalid!\n", s );
1100 else if( !h->param.rc.i_aq_mode && h->param.analyse.b_ssim )
1102 x264_log( h, X264_LOG_WARNING, "--ssim used with AQ off: results will be invalid!\n" );
1105 else if( h->param.rc.i_aq_mode && h->param.analyse.b_psnr )
1107 x264_log( h, X264_LOG_WARNING, "--psnr used with AQ on: results will be invalid!\n" );
1111 x264_log( h, X264_LOG_WARNING, "--tune %s should be used if attempting to benchmark %s!\n", s, s );
1114 if( !h->param.analyse.b_psy )
1116 h->param.analyse.f_psy_rd = 0;
1117 h->param.analyse.f_psy_trellis = 0;
1119 h->param.analyse.f_psy_rd = x264_clip3f( h->param.analyse.f_psy_rd, 0, 10 );
1120 h->param.analyse.f_psy_trellis = x264_clip3f( h->param.analyse.f_psy_trellis, 0, 10 );
1121 h->mb.i_psy_rd = h->param.analyse.i_subpel_refine >= 6 ? FIX8( h->param.analyse.f_psy_rd ) : 0;
1122 h->mb.i_psy_trellis = h->param.analyse.i_trellis ? FIX8( h->param.analyse.f_psy_trellis / 4 ) : 0;
1123 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -32, 32);
1124 /* In 4:4:4 mode, chroma gets twice as much resolution, so we can halve its quality. */
1125 if( b_open && i_csp >= X264_CSP_I444 && i_csp < X264_CSP_BGR && h->param.analyse.b_psy )
1126 h->param.analyse.i_chroma_qp_offset += 6;
1127 /* Psy RDO increases overall quantizers to improve the quality of luma--this indirectly hurts chroma quality */
1128 /* so we lower the chroma QP offset to compensate */
1129 if( b_open && h->mb.i_psy_rd && !h->param.i_avcintra_class )
1130 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_rd < 0.25 ? 1 : 2;
1131 /* Psy trellis has a similar effect. */
1132 if( b_open && h->mb.i_psy_trellis && !h->param.i_avcintra_class )
1133 h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_trellis < 0.25 ? 1 : 2;
1134 h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
1135 /* MB-tree requires AQ to be on, even if the strength is zero. */
1136 if( !h->param.rc.i_aq_mode && h->param.rc.b_mb_tree )
1138 h->param.rc.i_aq_mode = 1;
1139 h->param.rc.f_aq_strength = 0;
1141 h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
1142 if( h->param.analyse.i_subpel_refine >= 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
1143 h->param.analyse.i_subpel_refine = 9;
1146 const x264_level_t *l = x264_levels;
1147 if( h->param.i_level_idc < 0 )
1149 int maxrate_bak = h->param.rc.i_vbv_max_bitrate;
1150 if( h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.i_vbv_buffer_size <= 0 )
1151 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate * 2;
1152 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1153 do h->param.i_level_idc = l->level_idc;
1154 while( l[1].level_idc && x264_validate_levels( h, 0 ) && l++ );
1155 h->param.rc.i_vbv_max_bitrate = maxrate_bak;
1159 while( l->level_idc && l->level_idc != h->param.i_level_idc )
1161 if( l->level_idc == 0 )
1163 x264_log( h, X264_LOG_ERROR, "invalid level_idc: %d\n", h->param.i_level_idc );
1167 if( h->param.analyse.i_mv_range <= 0 )
1168 h->param.analyse.i_mv_range = l->mv_range >> PARAM_INTERLACED;
1170 h->param.analyse.i_mv_range = x264_clip3(h->param.analyse.i_mv_range, 32, 512 >> PARAM_INTERLACED);
1173 h->param.analyse.i_weighted_pred = x264_clip3( h->param.analyse.i_weighted_pred, X264_WEIGHTP_NONE, X264_WEIGHTP_SMART );
1175 if( h->param.i_lookahead_threads == X264_THREADS_AUTO )
1177 if( h->param.b_sliced_threads )
1178 h->param.i_lookahead_threads = h->param.i_threads;
1181 /* If we're using much slower lookahead settings than encoding settings, it helps a lot to use
1182 * more lookahead threads. This typically happens in the first pass of a two-pass encode, so
1183 * try to guess at this sort of case.
1185 * Tuned by a little bit of real encoding with the various presets. */
1186 int badapt = h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS;
1187 int subme = X264_MIN( h->param.analyse.i_subpel_refine / 3, 3 ) + (h->param.analyse.i_subpel_refine > 1);
1188 int bframes = X264_MIN( (h->param.i_bframe - 1) / 3, 3 );
1190 /* [b-adapt 0/1 vs 2][quantized subme][quantized bframes] */
1191 static const uint8_t lookahead_thread_div[2][5][4] =
1192 {{{6,6,6,6}, {3,3,3,3}, {4,4,4,4}, {6,6,6,6}, {12,12,12,12}},
1193 {{3,2,1,1}, {2,1,1,1}, {4,3,2,1}, {6,4,3,2}, {12, 9, 6, 4}}};
1195 h->param.i_lookahead_threads = h->param.i_threads / lookahead_thread_div[badapt][subme][bframes];
1196 /* Since too many lookahead threads significantly degrades lookahead accuracy, limit auto
1197 * lookahead threads to about 8 macroblock rows high each at worst. This number is chosen
1198 * pretty much arbitrarily. */
1199 h->param.i_lookahead_threads = X264_MIN( h->param.i_lookahead_threads, h->param.i_height / 128 );
1202 h->param.i_lookahead_threads = x264_clip3( h->param.i_lookahead_threads, 1, X264_MIN( max_sliced_threads, X264_LOOKAHEAD_THREAD_MAX ) );
1204 if( PARAM_INTERLACED )
1206 if( h->param.analyse.i_me_method >= X264_ME_ESA )
1208 x264_log( h, X264_LOG_WARNING, "interlace + me=esa is not implemented\n" );
1209 h->param.analyse.i_me_method = X264_ME_UMH;
1211 if( h->param.analyse.i_weighted_pred > 0 )
1213 x264_log( h, X264_LOG_WARNING, "interlace + weightp is not implemented\n" );
1214 h->param.analyse.i_weighted_pred = X264_WEIGHTP_NONE;
1218 if( !h->param.analyse.i_weighted_pred && h->param.rc.b_mb_tree && h->param.analyse.b_psy )
1219 h->param.analyse.i_weighted_pred = X264_WEIGHTP_FAKE;
1221 if( h->i_thread_frames > 1 )
1223 int r = h->param.analyse.i_mv_range_thread;
1227 // half of the available space is reserved and divided evenly among the threads,
1228 // the rest is allocated to whichever thread is far enough ahead to use it.
1229 // reserving more space increases quality for some videos, but costs more time
1230 // in thread synchronization.
1231 int max_range = (h->param.i_height + X264_THREAD_HEIGHT) / h->i_thread_frames - X264_THREAD_HEIGHT;
1234 r = X264_MAX( r, h->param.analyse.i_me_range );
1235 r = X264_MIN( r, h->param.analyse.i_mv_range );
1236 // round up to use the whole mb row
1237 r2 = (r & ~15) + ((-X264_THREAD_HEIGHT) & 15);
1240 x264_log( h, X264_LOG_DEBUG, "using mv_range_thread = %d\n", r2 );
1241 h->param.analyse.i_mv_range_thread = r2;
1244 if( h->param.rc.f_rate_tolerance < 0 )
1245 h->param.rc.f_rate_tolerance = 0;
1246 if( h->param.rc.f_qblur < 0 )
1247 h->param.rc.f_qblur = 0;
1248 if( h->param.rc.f_complexity_blur < 0 )
1249 h->param.rc.f_complexity_blur = 0;
1251 h->param.i_sps_id &= 31;
1253 h->param.i_nal_hrd = x264_clip3( h->param.i_nal_hrd, X264_NAL_HRD_NONE, X264_NAL_HRD_CBR );
1255 if( h->param.i_nal_hrd && !h->param.rc.i_vbv_buffer_size )
1257 x264_log( h, X264_LOG_WARNING, "NAL HRD parameters require VBV parameters\n" );
1258 h->param.i_nal_hrd = X264_NAL_HRD_NONE;
1261 if( h->param.i_nal_hrd == X264_NAL_HRD_CBR &&
1262 (h->param.rc.i_bitrate != h->param.rc.i_vbv_max_bitrate || !h->param.rc.i_vbv_max_bitrate) )
1264 x264_log( h, X264_LOG_WARNING, "CBR HRD requires constant bitrate\n" );
1265 h->param.i_nal_hrd = X264_NAL_HRD_VBR;
1268 if( h->param.i_nal_hrd == X264_NAL_HRD_CBR )
1269 h->param.rc.b_filler = 1;
1271 /* ensure the booleans are 0 or 1 so they can be used in math */
1272 #define BOOLIFY(x) h->param.x = !!h->param.x
1274 BOOLIFY( b_constrained_intra );
1275 BOOLIFY( b_deblocking_filter );
1276 BOOLIFY( b_deterministic );
1277 BOOLIFY( b_sliced_threads );
1278 BOOLIFY( b_interlaced );
1279 BOOLIFY( b_intra_refresh );
1281 BOOLIFY( b_repeat_headers );
1282 BOOLIFY( b_annexb );
1283 BOOLIFY( b_vfr_input );
1284 BOOLIFY( b_pulldown );
1286 BOOLIFY( b_pic_struct );
1287 BOOLIFY( b_fake_interlaced );
1288 BOOLIFY( b_open_gop );
1289 BOOLIFY( b_bluray_compat );
1290 BOOLIFY( b_stitchable );
1291 BOOLIFY( b_full_recon );
1292 BOOLIFY( b_opencl );
1293 BOOLIFY( analyse.b_transform_8x8 );
1294 BOOLIFY( analyse.b_weighted_bipred );
1295 BOOLIFY( analyse.b_chroma_me );
1296 BOOLIFY( analyse.b_mixed_references );
1297 BOOLIFY( analyse.b_fast_pskip );
1298 BOOLIFY( analyse.b_dct_decimate );
1299 BOOLIFY( analyse.b_psy );
1300 BOOLIFY( analyse.b_psnr );
1301 BOOLIFY( analyse.b_ssim );
1302 BOOLIFY( rc.b_stat_write );
1303 BOOLIFY( rc.b_stat_read );
1304 BOOLIFY( rc.b_mb_tree );
1305 BOOLIFY( rc.b_filler );
1311 static void mbcmp_init( x264_t *h )
1313 int satd = !h->mb.b_lossless && h->param.analyse.i_subpel_refine > 1;
1314 memcpy( h->pixf.mbcmp, satd ? h->pixf.satd : h->pixf.sad_aligned, sizeof(h->pixf.mbcmp) );
1315 memcpy( h->pixf.mbcmp_unaligned, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.mbcmp_unaligned) );
1316 h->pixf.intra_mbcmp_x3_16x16 = satd ? h->pixf.intra_satd_x3_16x16 : h->pixf.intra_sad_x3_16x16;
1317 h->pixf.intra_mbcmp_x3_8x16c = satd ? h->pixf.intra_satd_x3_8x16c : h->pixf.intra_sad_x3_8x16c;
1318 h->pixf.intra_mbcmp_x3_8x8c = satd ? h->pixf.intra_satd_x3_8x8c : h->pixf.intra_sad_x3_8x8c;
1319 h->pixf.intra_mbcmp_x3_8x8 = satd ? h->pixf.intra_sa8d_x3_8x8 : h->pixf.intra_sad_x3_8x8;
1320 h->pixf.intra_mbcmp_x3_4x4 = satd ? h->pixf.intra_satd_x3_4x4 : h->pixf.intra_sad_x3_4x4;
1321 h->pixf.intra_mbcmp_x9_4x4 = h->param.b_cpu_independent || h->mb.b_lossless ? NULL
1322 : satd ? h->pixf.intra_satd_x9_4x4 : h->pixf.intra_sad_x9_4x4;
1323 h->pixf.intra_mbcmp_x9_8x8 = h->param.b_cpu_independent || h->mb.b_lossless ? NULL
1324 : satd ? h->pixf.intra_sa8d_x9_8x8 : h->pixf.intra_sad_x9_8x8;
1325 satd &= h->param.analyse.i_me_method == X264_ME_TESA;
1326 memcpy( h->pixf.fpelcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.fpelcmp) );
1327 memcpy( h->pixf.fpelcmp_x3, satd ? h->pixf.satd_x3 : h->pixf.sad_x3, sizeof(h->pixf.fpelcmp_x3) );
1328 memcpy( h->pixf.fpelcmp_x4, satd ? h->pixf.satd_x4 : h->pixf.sad_x4, sizeof(h->pixf.fpelcmp_x4) );
1331 static void chroma_dsp_init( x264_t *h )
1333 memcpy( h->luma2chroma_pixel, x264_luma2chroma_pixel[CHROMA_FORMAT], sizeof(h->luma2chroma_pixel) );
1335 switch( CHROMA_FORMAT )
1338 memcpy( h->predict_chroma, h->predict_8x8c, sizeof(h->predict_chroma) );
1339 h->mc.prefetch_fenc = h->mc.prefetch_fenc_420;
1340 h->loopf.deblock_chroma[0] = h->loopf.deblock_h_chroma_420;
1341 h->loopf.deblock_chroma_intra[0] = h->loopf.deblock_h_chroma_420_intra;
1342 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_chroma_420_mbaff;
1343 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_chroma_420_intra_mbaff;
1344 h->pixf.intra_mbcmp_x3_chroma = h->pixf.intra_mbcmp_x3_8x8c;
1345 h->quantf.coeff_last[DCT_CHROMA_DC] = h->quantf.coeff_last4;
1346 h->quantf.coeff_level_run[DCT_CHROMA_DC] = h->quantf.coeff_level_run4;
1349 memcpy( h->predict_chroma, h->predict_8x16c, sizeof(h->predict_chroma) );
1350 h->mc.prefetch_fenc = h->mc.prefetch_fenc_422;
1351 h->loopf.deblock_chroma[0] = h->loopf.deblock_h_chroma_422;
1352 h->loopf.deblock_chroma_intra[0] = h->loopf.deblock_h_chroma_422_intra;
1353 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_chroma_422_mbaff;
1354 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_chroma_422_intra_mbaff;
1355 h->pixf.intra_mbcmp_x3_chroma = h->pixf.intra_mbcmp_x3_8x16c;
1356 h->quantf.coeff_last[DCT_CHROMA_DC] = h->quantf.coeff_last8;
1357 h->quantf.coeff_level_run[DCT_CHROMA_DC] = h->quantf.coeff_level_run8;
1360 h->mc.prefetch_fenc = h->mc.prefetch_fenc_422; /* FIXME: doesn't cover V plane */
1361 h->loopf.deblock_chroma_mbaff = h->loopf.deblock_luma_mbaff;
1362 h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_luma_intra_mbaff;
1367 static void x264_set_aspect_ratio( x264_t *h, x264_param_t *param, int initial )
1370 if( param->vui.i_sar_width > 0 && param->vui.i_sar_height > 0 )
1372 uint32_t i_w = param->vui.i_sar_width;
1373 uint32_t i_h = param->vui.i_sar_height;
1374 uint32_t old_w = h->param.vui.i_sar_width;
1375 uint32_t old_h = h->param.vui.i_sar_height;
1377 x264_reduce_fraction( &i_w, &i_h );
1379 while( i_w > 65535 || i_h > 65535 )
1385 x264_reduce_fraction( &i_w, &i_h );
1387 if( i_w != old_w || i_h != old_h || initial )
1389 h->param.vui.i_sar_width = 0;
1390 h->param.vui.i_sar_height = 0;
1391 if( i_w == 0 || i_h == 0 )
1392 x264_log( h, X264_LOG_WARNING, "cannot create valid sample aspect ratio\n" );
1395 x264_log( h, initial?X264_LOG_INFO:X264_LOG_DEBUG, "using SAR=%d/%d\n", i_w, i_h );
1396 h->param.vui.i_sar_width = i_w;
1397 h->param.vui.i_sar_height = i_h;
1403 /****************************************************************************
1404 * x264_encoder_open:
1405 ****************************************************************************/
1406 x264_t *x264_encoder_open( x264_param_t *param )
1410 int qp, i_slicetype_length;
1412 CHECKED_MALLOCZERO( h, sizeof(x264_t) );
1414 /* Create a copy of param */
1415 memcpy( &h->param, param, sizeof(x264_param_t) );
1417 if( param->param_free )
1418 param->param_free( param );
1420 #if HAVE_INTEL_DISPATCHER
1421 x264_intel_dispatcher_override();
1424 if( x264_threading_init() )
1426 x264_log( h, X264_LOG_ERROR, "unable to initialize threading\n" );
1430 if( x264_validate_parameters( h, 1 ) < 0 )
1433 if( h->param.psz_cqm_file )
1434 if( x264_cqm_parse_file( h, h->param.psz_cqm_file ) < 0 )
1437 if( h->param.rc.psz_stat_out )
1438 h->param.rc.psz_stat_out = strdup( h->param.rc.psz_stat_out );
1439 if( h->param.rc.psz_stat_in )
1440 h->param.rc.psz_stat_in = strdup( h->param.rc.psz_stat_in );
1442 x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den );
1443 x264_reduce_fraction( &h->param.i_timebase_num, &h->param.i_timebase_den );
1449 if( h->param.i_avcintra_class )
1450 h->i_idr_pic_id = 5;
1452 h->i_idr_pic_id = 0;
1454 if( (uint64_t)h->param.i_timebase_den * 2 > UINT32_MAX )
1456 x264_log( h, X264_LOG_ERROR, "Effective timebase denominator %u exceeds H.264 maximum\n", h->param.i_timebase_den );
1460 x264_set_aspect_ratio( h, &h->param, 1 );
1462 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1463 x264_pps_init( h->pps, h->param.i_sps_id, &h->param, h->sps );
1465 x264_validate_levels( h, 1 );
1467 h->chroma_qp_table = i_chroma_qp_table + 12 + h->pps->i_chroma_qp_index_offset;
1469 if( x264_cqm_init( h ) < 0 )
1472 h->mb.i_mb_width = h->sps->i_mb_width;
1473 h->mb.i_mb_height = h->sps->i_mb_height;
1474 h->mb.i_mb_count = h->mb.i_mb_width * h->mb.i_mb_height;
1476 h->mb.chroma_h_shift = CHROMA_FORMAT == CHROMA_420 || CHROMA_FORMAT == CHROMA_422;
1477 h->mb.chroma_v_shift = CHROMA_FORMAT == CHROMA_420;
1479 /* Adaptive MBAFF and subme 0 are not supported as we require halving motion
1480 * vectors during prediction, resulting in hpel mvs.
1481 * The chosen solution is to make MBAFF non-adaptive in this case. */
1482 h->mb.b_adaptive_mbaff = PARAM_INTERLACED && h->param.analyse.i_subpel_refine;
1485 if( h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS && !h->param.rc.b_stat_read )
1486 h->frames.i_delay = X264_MAX(h->param.i_bframe,3)*4;
1488 h->frames.i_delay = h->param.i_bframe;
1489 if( h->param.rc.b_mb_tree || h->param.rc.i_vbv_buffer_size )
1490 h->frames.i_delay = X264_MAX( h->frames.i_delay, h->param.rc.i_lookahead );
1491 i_slicetype_length = h->frames.i_delay;
1492 h->frames.i_delay += h->i_thread_frames - 1;
1493 h->frames.i_delay += h->param.i_sync_lookahead;
1494 h->frames.i_delay += h->param.b_vfr_input;
1495 h->frames.i_bframe_delay = h->param.i_bframe ? (h->param.i_bframe_pyramid ? 2 : 1) : 0;
1497 h->frames.i_max_ref0 = h->param.i_frame_reference;
1498 h->frames.i_max_ref1 = X264_MIN( h->sps->vui.i_num_reorder_frames, h->param.i_frame_reference );
1499 h->frames.i_max_dpb = h->sps->vui.i_max_dec_frame_buffering;
1500 h->frames.b_have_lowres = !h->param.rc.b_stat_read
1501 && ( h->param.rc.i_rc_method == X264_RC_ABR
1502 || h->param.rc.i_rc_method == X264_RC_CRF
1503 || h->param.i_bframe_adaptive
1504 || h->param.i_scenecut_threshold
1505 || h->param.rc.b_mb_tree
1506 || h->param.analyse.i_weighted_pred );
1507 h->frames.b_have_lowres |= h->param.rc.b_stat_read && h->param.rc.i_vbv_buffer_size > 0;
1508 h->frames.b_have_sub8x8_esa = !!(h->param.analyse.inter & X264_ANALYSE_PSUB8x8);
1510 h->frames.i_last_idr =
1511 h->frames.i_last_keyframe = - h->param.i_keyint_max;
1512 h->frames.i_input = 0;
1513 h->frames.i_largest_pts = h->frames.i_second_largest_pts = -1;
1514 h->frames.i_poc_last_open_gop = -1;
1516 CHECKED_MALLOCZERO( h->frames.unused[0], (h->frames.i_delay + 3) * sizeof(x264_frame_t *) );
1517 /* Allocate room for max refs plus a few extra just in case. */
1518 CHECKED_MALLOCZERO( h->frames.unused[1], (h->i_thread_frames + X264_REF_MAX + 4) * sizeof(x264_frame_t *) );
1519 CHECKED_MALLOCZERO( h->frames.current, (h->param.i_sync_lookahead + h->param.i_bframe
1520 + h->i_thread_frames + 3) * sizeof(x264_frame_t *) );
1521 if( h->param.analyse.i_weighted_pred > 0 )
1522 CHECKED_MALLOCZERO( h->frames.blank_unused, h->i_thread_frames * 4 * sizeof(x264_frame_t *) );
1523 h->i_ref[0] = h->i_ref[1] = 0;
1524 h->i_cpb_delay = h->i_coded_fields = h->i_disp_fields = 0;
1525 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);
1526 h->i_disp_fields_last_frame = -1;
1529 /* init CPU functions */
1530 x264_predict_16x16_init( h->param.cpu, h->predict_16x16 );
1531 x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c );
1532 x264_predict_8x16c_init( h->param.cpu, h->predict_8x16c );
1533 x264_predict_8x8_init( h->param.cpu, h->predict_8x8, &h->predict_8x8_filter );
1534 x264_predict_4x4_init( h->param.cpu, h->predict_4x4 );
1535 x264_pixel_init( h->param.cpu, &h->pixf );
1536 x264_dct_init( h->param.cpu, &h->dctf );
1537 x264_zigzag_init( h->param.cpu, &h->zigzagf_progressive, &h->zigzagf_interlaced );
1538 memcpy( &h->zigzagf, PARAM_INTERLACED ? &h->zigzagf_interlaced : &h->zigzagf_progressive, sizeof(h->zigzagf) );
1539 x264_mc_init( h->param.cpu, &h->mc, h->param.b_cpu_independent );
1540 x264_quant_init( h, h->param.cpu, &h->quantf );
1541 x264_deblock_init( h->param.cpu, &h->loopf, PARAM_INTERLACED );
1542 x264_bitstream_init( h->param.cpu, &h->bsf );
1543 if( h->param.b_cabac )
1544 x264_cabac_init( h );
1546 x264_stack_align( x264_cavlc_init, h );
1549 chroma_dsp_init( h );
1551 p = buf + sprintf( buf, "using cpu capabilities:" );
1552 for( int i = 0; x264_cpu_names[i].flags; i++ )
1554 if( !strcmp(x264_cpu_names[i].name, "SSE")
1555 && h->param.cpu & (X264_CPU_SSE2) )
1557 if( !strcmp(x264_cpu_names[i].name, "SSE2")
1558 && h->param.cpu & (X264_CPU_SSE2_IS_FAST|X264_CPU_SSE2_IS_SLOW) )
1560 if( !strcmp(x264_cpu_names[i].name, "SSE3")
1561 && (h->param.cpu & X264_CPU_SSSE3 || !(h->param.cpu & X264_CPU_CACHELINE_64)) )
1563 if( !strcmp(x264_cpu_names[i].name, "SSE4.1")
1564 && (h->param.cpu & X264_CPU_SSE42) )
1566 if( !strcmp(x264_cpu_names[i].name, "BMI1")
1567 && (h->param.cpu & X264_CPU_BMI2) )
1569 if( (h->param.cpu & x264_cpu_names[i].flags) == x264_cpu_names[i].flags
1570 && (!i || x264_cpu_names[i].flags != x264_cpu_names[i-1].flags) )
1571 p += sprintf( p, " %s", x264_cpu_names[i].name );
1574 p += sprintf( p, " none!" );
1575 x264_log( h, X264_LOG_INFO, "%s\n", buf );
1577 float *logs = x264_analyse_prepare_costs( h );
1580 for( qp = X264_MIN( h->param.rc.i_qp_min, QP_MAX_SPEC ); qp <= h->param.rc.i_qp_max; qp++ )
1581 if( x264_analyse_init_costs( h, logs, qp ) )
1583 if( x264_analyse_init_costs( h, logs, X264_LOOKAHEAD_QP ) )
1587 static const uint16_t cost_mv_correct[7] = { 24, 47, 95, 189, 379, 757, 1515 };
1588 /* Checks for known miscompilation issues. */
1589 if( h->cost_mv[X264_LOOKAHEAD_QP][2013] != cost_mv_correct[BIT_DEPTH-8] )
1591 x264_log( h, X264_LOG_ERROR, "MV cost test failed: x264 has been miscompiled!\n" );
1595 /* Must be volatile or else GCC will optimize it out. */
1596 volatile int temp = 392;
1597 if( x264_clz( temp ) != 23 )
1599 x264_log( h, X264_LOG_ERROR, "CLZ test failed: x264 has been miscompiled!\n" );
1600 #if ARCH_X86 || ARCH_X86_64
1601 x264_log( h, X264_LOG_ERROR, "Are you attempting to run an SSE4a/LZCNT-targeted build on a CPU that\n" );
1602 x264_log( h, X264_LOG_ERROR, "doesn't support it?\n" );
1608 h->out.i_bitstream = X264_MAX( 1000000, h->param.i_width * h->param.i_height * 4
1609 * ( h->param.rc.i_rc_method == X264_RC_ABR ? pow( 0.95, h->param.rc.i_qp_min )
1610 : pow( 0.95, h->param.rc.i_qp_constant ) * X264_MAX( 1, h->param.rc.f_ip_factor )));
1612 h->nal_buffer_size = h->out.i_bitstream * 3/2 + 4 + 64; /* +4 for startcode, +64 for nal_escape assembly padding */
1613 CHECKED_MALLOC( h->nal_buffer, h->nal_buffer_size );
1615 CHECKED_MALLOC( h->reconfig_h, sizeof(x264_t) );
1617 if( h->param.i_threads > 1 &&
1618 x264_threadpool_init( &h->threadpool, h->param.i_threads, (void*)x264_encoder_thread_init, h ) )
1620 if( h->param.i_lookahead_threads > 1 &&
1621 x264_threadpool_init( &h->lookaheadpool, h->param.i_lookahead_threads, NULL, NULL ) )
1625 if( h->param.b_opencl )
1627 h->opencl.ocl = x264_opencl_load_library();
1628 if( !h->opencl.ocl )
1630 x264_log( h, X264_LOG_WARNING, "failed to load OpenCL\n" );
1631 h->param.b_opencl = 0;
1637 for( int i = 1; i < h->param.i_threads + !!h->param.i_sync_lookahead; i++ )
1638 CHECKED_MALLOC( h->thread[i], sizeof(x264_t) );
1639 if( h->param.i_lookahead_threads > 1 )
1640 for( int i = 0; i < h->param.i_lookahead_threads; i++ )
1642 CHECKED_MALLOC( h->lookahead_thread[i], sizeof(x264_t) );
1643 *h->lookahead_thread[i] = *h;
1645 *h->reconfig_h = *h;
1647 for( int i = 0; i < h->param.i_threads; i++ )
1649 int init_nal_count = h->param.i_slice_count + 3;
1650 int allocate_threadlocal_data = !h->param.b_sliced_threads || !i;
1654 if( x264_pthread_mutex_init( &h->thread[i]->mutex, NULL ) )
1656 if( x264_pthread_cond_init( &h->thread[i]->cv, NULL ) )
1659 if( allocate_threadlocal_data )
1661 h->thread[i]->fdec = x264_frame_pop_unused( h, 1 );
1662 if( !h->thread[i]->fdec )
1666 h->thread[i]->fdec = h->thread[0]->fdec;
1668 CHECKED_MALLOC( h->thread[i]->out.p_bitstream, h->out.i_bitstream );
1669 /* Start each thread with room for init_nal_count NAL units; it'll realloc later if needed. */
1670 CHECKED_MALLOC( h->thread[i]->out.nal, init_nal_count*sizeof(x264_nal_t) );
1671 h->thread[i]->out.i_nals_allocated = init_nal_count;
1673 if( allocate_threadlocal_data && x264_macroblock_cache_allocate( h->thread[i] ) < 0 )
1678 if( h->param.b_opencl && x264_opencl_lookahead_init( h ) < 0 )
1679 h->param.b_opencl = 0;
1682 if( x264_lookahead_init( h, i_slicetype_length ) )
1685 for( int i = 0; i < h->param.i_threads; i++ )
1686 if( x264_macroblock_thread_allocate( h->thread[i], 0 ) < 0 )
1689 if( x264_ratecontrol_new( h ) < 0 )
1692 if( h->param.i_nal_hrd )
1694 x264_log( h, X264_LOG_DEBUG, "HRD bitrate: %i bits/sec\n", h->sps->vui.hrd.i_bit_rate_unscaled );
1695 x264_log( h, X264_LOG_DEBUG, "CPB size: %i bits\n", h->sps->vui.hrd.i_cpb_size_unscaled );
1698 if( h->param.psz_dump_yuv )
1700 /* create or truncate the reconstructed video file */
1701 FILE *f = x264_fopen( h->param.psz_dump_yuv, "w" );
1704 x264_log( h, X264_LOG_ERROR, "dump_yuv: can't write to %s\n", h->param.psz_dump_yuv );
1707 else if( !x264_is_regular_file( f ) )
1709 x264_log( h, X264_LOG_ERROR, "dump_yuv: incompatible with non-regular file %s\n", h->param.psz_dump_yuv );
1716 const char *profile = h->sps->i_profile_idc == PROFILE_BASELINE ? "Constrained Baseline" :
1717 h->sps->i_profile_idc == PROFILE_MAIN ? "Main" :
1718 h->sps->i_profile_idc == PROFILE_HIGH ? "High" :
1719 h->sps->i_profile_idc == PROFILE_HIGH10 ? (h->sps->b_constraint_set3 == 1 ? "High 10 Intra" : "High 10") :
1720 h->sps->i_profile_idc == PROFILE_HIGH422 ? (h->sps->b_constraint_set3 == 1 ? "High 4:2:2 Intra" : "High 4:2:2") :
1721 h->sps->b_constraint_set3 == 1 ? "High 4:4:4 Intra" : "High 4:4:4 Predictive";
1723 snprintf( level, sizeof(level), "%d.%d", h->sps->i_level_idc/10, h->sps->i_level_idc%10 );
1724 if( h->sps->i_level_idc == 9 || ( h->sps->i_level_idc == 11 && h->sps->b_constraint_set3 &&
1725 (h->sps->i_profile_idc == PROFILE_BASELINE || h->sps->i_profile_idc == PROFILE_MAIN) ) )
1726 strcpy( level, "1b" );
1728 if( h->sps->i_profile_idc < PROFILE_HIGH10 )
1730 x264_log( h, X264_LOG_INFO, "profile %s, level %s\n",
1735 static const char * const subsampling[4] = { "4:0:0", "4:2:0", "4:2:2", "4:4:4" };
1736 x264_log( h, X264_LOG_INFO, "profile %s, level %s, %s %d-bit\n",
1737 profile, level, subsampling[CHROMA_FORMAT], BIT_DEPTH );
1746 /****************************************************************************/
1747 static int x264_encoder_try_reconfig( x264_t *h, x264_param_t *param, int *rc_reconfig )
1750 x264_set_aspect_ratio( h, param, 0 );
1751 #define COPY(var) h->param.var = param->var
1752 COPY( i_frame_reference ); // but never uses more refs than initially specified
1753 COPY( i_bframe_bias );
1754 if( h->param.i_scenecut_threshold )
1755 COPY( i_scenecut_threshold ); // can't turn it on or off, only vary the threshold
1756 COPY( b_deblocking_filter );
1757 COPY( i_deblocking_filter_alphac0 );
1758 COPY( i_deblocking_filter_beta );
1759 COPY( i_frame_packing );
1760 COPY( analyse.inter );
1761 COPY( analyse.intra );
1762 COPY( analyse.i_direct_mv_pred );
1763 /* Scratch buffer prevents me_range from being increased for esa/tesa */
1764 if( h->param.analyse.i_me_method < X264_ME_ESA || param->analyse.i_me_range < h->param.analyse.i_me_range )
1765 COPY( analyse.i_me_range );
1766 COPY( analyse.i_noise_reduction );
1767 /* We can't switch out of subme=0 during encoding. */
1768 if( h->param.analyse.i_subpel_refine )
1769 COPY( analyse.i_subpel_refine );
1770 COPY( analyse.i_trellis );
1771 COPY( analyse.b_chroma_me );
1772 COPY( analyse.b_dct_decimate );
1773 COPY( analyse.b_fast_pskip );
1774 COPY( analyse.b_mixed_references );
1775 COPY( analyse.f_psy_rd );
1776 COPY( analyse.f_psy_trellis );
1778 // can only twiddle these if they were enabled to begin with:
1779 if( h->param.analyse.i_me_method >= X264_ME_ESA || param->analyse.i_me_method < X264_ME_ESA )
1780 COPY( analyse.i_me_method );
1781 if( h->param.analyse.i_me_method >= X264_ME_ESA && !h->frames.b_have_sub8x8_esa )
1782 h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
1783 if( h->pps->b_transform_8x8_mode )
1784 COPY( analyse.b_transform_8x8 );
1785 if( h->frames.i_max_ref1 > 1 )
1786 COPY( i_bframe_pyramid );
1787 COPY( i_slice_max_size );
1788 COPY( i_slice_max_mbs );
1789 COPY( i_slice_min_mbs );
1790 COPY( i_slice_count );
1791 COPY( i_slice_count_max );
1794 /* VBV can't be turned on if it wasn't on to begin with */
1795 if( h->param.rc.i_vbv_max_bitrate > 0 && h->param.rc.i_vbv_buffer_size > 0 &&
1796 param->rc.i_vbv_max_bitrate > 0 && param->rc.i_vbv_buffer_size > 0 )
1798 *rc_reconfig |= h->param.rc.i_vbv_max_bitrate != param->rc.i_vbv_max_bitrate;
1799 *rc_reconfig |= h->param.rc.i_vbv_buffer_size != param->rc.i_vbv_buffer_size;
1800 *rc_reconfig |= h->param.rc.i_bitrate != param->rc.i_bitrate;
1801 COPY( rc.i_vbv_max_bitrate );
1802 COPY( rc.i_vbv_buffer_size );
1803 COPY( rc.i_bitrate );
1805 *rc_reconfig |= h->param.rc.f_rf_constant != param->rc.f_rf_constant;
1806 *rc_reconfig |= h->param.rc.f_rf_constant_max != param->rc.f_rf_constant_max;
1807 COPY( rc.f_rf_constant );
1808 COPY( rc.f_rf_constant_max );
1811 return x264_validate_parameters( h, 0 );
1814 int x264_encoder_reconfig_apply( x264_t *h, x264_param_t *param )
1817 int ret = x264_encoder_try_reconfig( h, param, &rc_reconfig );
1821 x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1823 /* Supported reconfiguration options (1-pass only):
1827 * bitrate (CBR only) */
1828 if( !ret && rc_reconfig )
1829 x264_ratecontrol_init_reconfigurable( h, 0 );
1834 /****************************************************************************
1835 * x264_encoder_reconfig:
1836 ****************************************************************************/
1837 int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
1839 h = h->thread[h->thread[0]->i_thread_phase];
1840 x264_param_t param_save = h->reconfig_h->param;
1841 h->reconfig_h->param = h->param;
1844 int ret = x264_encoder_try_reconfig( h->reconfig_h, param, &rc_reconfig );
1848 h->reconfig_h->param = param_save;
1853 /****************************************************************************
1854 * x264_encoder_parameters:
1855 ****************************************************************************/
1856 void x264_encoder_parameters( x264_t *h, x264_param_t *param )
1858 memcpy( param, &h->thread[h->i_thread_phase]->param, sizeof(x264_param_t) );
1861 /* internal usage */
1862 static void x264_nal_start( x264_t *h, int i_type, int i_ref_idc )
1864 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1866 nal->i_ref_idc = i_ref_idc;
1867 nal->i_type = i_type;
1868 nal->b_long_startcode = 1;
1871 nal->p_payload= &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1875 /* if number of allocated nals is not enough, re-allocate a larger one. */
1876 static int x264_nal_check_buffer( x264_t *h )
1878 if( h->out.i_nal >= h->out.i_nals_allocated )
1880 x264_nal_t *new_out = x264_malloc( sizeof(x264_nal_t) * (h->out.i_nals_allocated*2) );
1883 memcpy( new_out, h->out.nal, sizeof(x264_nal_t) * (h->out.i_nals_allocated) );
1884 x264_free( h->out.nal );
1885 h->out.nal = new_out;
1886 h->out.i_nals_allocated *= 2;
1891 static int x264_nal_end( x264_t *h )
1893 x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1894 uint8_t *end = &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1895 nal->i_payload = end - nal->p_payload;
1896 /* Assembly implementation of nal_escape reads past the end of the input.
1897 * While undefined padding wouldn't actually affect the output, it makes valgrind unhappy. */
1898 memset( end, 0xff, 64 );
1899 if( h->param.nalu_process )
1900 h->param.nalu_process( h, nal, h->fenc->opaque );
1903 return x264_nal_check_buffer( h );
1906 static int x264_check_encapsulated_buffer( x264_t *h, x264_t *h0, int start,
1907 int previous_nal_size, int necessary_size )
1909 if( h0->nal_buffer_size < necessary_size )
1911 necessary_size *= 2;
1912 uint8_t *buf = x264_malloc( necessary_size );
1915 if( previous_nal_size )
1916 memcpy( buf, h0->nal_buffer, previous_nal_size );
1918 intptr_t delta = buf - h0->nal_buffer;
1919 for( int i = 0; i < start; i++ )
1920 h->out.nal[i].p_payload += delta;
1922 x264_free( h0->nal_buffer );
1923 h0->nal_buffer = buf;
1924 h0->nal_buffer_size = necessary_size;
1930 static int x264_encoder_encapsulate_nals( x264_t *h, int start )
1932 x264_t *h0 = h->thread[0];
1933 int nal_size = 0, previous_nal_size = 0;
1935 if( h->param.nalu_process )
1937 for( int i = start; i < h->out.i_nal; i++ )
1938 nal_size += h->out.nal[i].i_payload;
1942 for( int i = 0; i < start; i++ )
1943 previous_nal_size += h->out.nal[i].i_payload;
1945 for( int i = start; i < h->out.i_nal; i++ )
1946 nal_size += h->out.nal[i].i_payload;
1948 /* Worst-case NAL unit escaping: reallocate the buffer if it's too small. */
1949 int necessary_size = previous_nal_size + nal_size * 3/2 + h->out.i_nal * 4 + 4 + 64;
1950 for( int i = start; i < h->out.i_nal; i++ )
1951 necessary_size += h->out.nal[i].i_padding;
1952 if( x264_check_encapsulated_buffer( h, h0, start, previous_nal_size, necessary_size ) )
1955 uint8_t *nal_buffer = h0->nal_buffer + previous_nal_size;
1957 for( int i = start; i < h->out.i_nal; i++ )
1959 int old_payload_len = h->out.nal[i].i_payload;
1960 h->out.nal[i].b_long_startcode = !i || h->out.nal[i].i_type == NAL_SPS || h->out.nal[i].i_type == NAL_PPS ||
1961 h->param.i_avcintra_class;
1962 x264_nal_encode( h, nal_buffer, &h->out.nal[i] );
1963 nal_buffer += h->out.nal[i].i_payload;
1964 if( h->param.i_avcintra_class )
1966 h->out.nal[i].i_padding -= h->out.nal[i].i_payload - (old_payload_len + NALU_OVERHEAD);
1967 if( h->out.nal[i].i_padding > 0 )
1969 memset( nal_buffer, 0, h->out.nal[i].i_padding );
1970 nal_buffer += h->out.nal[i].i_padding;
1971 h->out.nal[i].i_payload += h->out.nal[i].i_padding;
1973 h->out.nal[i].i_padding = X264_MAX( h->out.nal[i].i_padding, 0 );
1979 return nal_buffer - (h0->nal_buffer + previous_nal_size);
1982 /****************************************************************************
1983 * x264_encoder_headers:
1984 ****************************************************************************/
1985 int x264_encoder_headers( x264_t *h, x264_nal_t **pp_nal, int *pi_nal )
1988 /* init bitstream context */
1990 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
1992 /* Write SEI, SPS and PPS. */
1994 /* generate sequence parameters */
1995 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
1996 x264_sps_write( &h->out.bs, h->sps );
1997 if( x264_nal_end( h ) )
2000 /* generate picture parameters */
2001 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
2002 x264_pps_write( &h->out.bs, h->sps, h->pps );
2003 if( x264_nal_end( h ) )
2006 /* identify ourselves */
2007 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
2008 if( x264_sei_version_write( h, &h->out.bs ) )
2010 if( x264_nal_end( h ) )
2013 frame_size = x264_encoder_encapsulate_nals( h, 0 );
2014 if( frame_size < 0 )
2018 *pi_nal = h->out.i_nal;
2019 *pp_nal = &h->out.nal[0];
2025 /* Check to see whether we have chosen a reference list ordering different
2026 * from the standard's default. */
2027 static inline void x264_reference_check_reorder( x264_t *h )
2029 /* The reorder check doesn't check for missing frames, so just
2030 * force a reorder if one of the reference list is corrupt. */
2031 for( int i = 0; h->frames.reference[i]; i++ )
2032 if( h->frames.reference[i]->b_corrupt )
2034 h->b_ref_reorder[0] = 1;
2037 for( int list = 0; list <= (h->sh.i_type == SLICE_TYPE_B); list++ )
2038 for( int i = 0; i < h->i_ref[list] - 1; i++ )
2040 int framenum_diff = h->fref[list][i+1]->i_frame_num - h->fref[list][i]->i_frame_num;
2041 int poc_diff = h->fref[list][i+1]->i_poc - h->fref[list][i]->i_poc;
2042 /* P and B-frames use different default orders. */
2043 if( h->sh.i_type == SLICE_TYPE_P ? framenum_diff > 0 : list == 1 ? poc_diff < 0 : poc_diff > 0 )
2045 h->b_ref_reorder[list] = 1;
2051 /* return -1 on failure, else return the index of the new reference frame */
2052 int x264_weighted_reference_duplicate( x264_t *h, int i_ref, const x264_weight_t *w )
2054 int i = h->i_ref[0];
2056 x264_frame_t *newframe;
2057 if( i <= 1 ) /* empty list, definitely can't duplicate frame */
2060 //Duplication is only used in X264_WEIGHTP_SMART
2061 if( h->param.analyse.i_weighted_pred != X264_WEIGHTP_SMART )
2064 /* Duplication is a hack to compensate for crappy rounding in motion compensation.
2065 * With high bit depth, it's not worth doing, so turn it off except in the case of
2066 * unweighted dupes. */
2067 if( BIT_DEPTH > 8 && w != x264_weight_none )
2070 newframe = x264_frame_pop_blank_unused( h );
2074 //FIXME: probably don't need to copy everything
2075 *newframe = *h->fref[0][i_ref];
2076 newframe->i_reference_count = 1;
2077 newframe->orig = h->fref[0][i_ref];
2078 newframe->b_duplicate = 1;
2079 memcpy( h->fenc->weight[j], w, sizeof(h->fenc->weight[i]) );
2081 /* shift the frames to make space for the dupe. */
2082 h->b_ref_reorder[0] = 1;
2083 if( h->i_ref[0] < X264_REF_MAX )
2085 h->fref[0][X264_REF_MAX-1] = NULL;
2086 x264_frame_unshift( &h->fref[0][j], newframe );
2091 static void x264_weighted_pred_init( x264_t *h )
2093 /* for now no analysis and set all weights to nothing */
2094 for( int i_ref = 0; i_ref < h->i_ref[0]; i_ref++ )
2095 h->fenc->weighted[i_ref] = h->fref[0][i_ref]->filtered[0][0];
2097 // FIXME: This only supports weighting of one reference frame
2098 // and duplicates of that frame.
2099 h->fenc->i_lines_weighted = 0;
2101 for( int i_ref = 0; i_ref < (h->i_ref[0] << SLICE_MBAFF); i_ref++ )
2102 for( int i = 0; i < 3; i++ )
2103 h->sh.weight[i_ref][i].weightfn = NULL;
2106 if( h->sh.i_type != SLICE_TYPE_P || h->param.analyse.i_weighted_pred <= 0 )
2109 int i_padv = PADV << PARAM_INTERLACED;
2111 int weightplane[2] = { 0, 0 };
2112 int buffer_next = 0;
2113 for( int i = 0; i < 3; i++ )
2115 for( int j = 0; j < h->i_ref[0]; j++ )
2117 if( h->fenc->weight[j][i].weightfn )
2119 h->sh.weight[j][i] = h->fenc->weight[j][i];
2120 // if weight is useless, don't write it to stream
2121 if( h->sh.weight[j][i].i_scale == 1<<h->sh.weight[j][i].i_denom && h->sh.weight[j][i].i_offset == 0 )
2122 h->sh.weight[j][i].weightfn = NULL;
2125 if( !weightplane[!!i] )
2127 weightplane[!!i] = 1;
2128 h->sh.weight[0][!!i].i_denom = denom = h->sh.weight[j][i].i_denom;
2129 assert( x264_clip3( denom, 0, 7 ) == denom );
2132 assert( h->sh.weight[j][i].i_denom == denom );
2135 h->fenc->weighted[j] = h->mb.p_weight_buf[buffer_next++] + h->fenc->i_stride[0] * i_padv + PADH;
2136 //scale full resolution frame
2137 if( h->param.i_threads == 1 )
2139 pixel *src = h->fref[0][j]->filtered[0][0] - h->fref[0][j]->i_stride[0]*i_padv - PADH;
2140 pixel *dst = h->fenc->weighted[j] - h->fenc->i_stride[0]*i_padv - PADH;
2141 int stride = h->fenc->i_stride[0];
2142 int width = h->fenc->i_width[0] + PADH*2;
2143 int height = h->fenc->i_lines[0] + i_padv*2;
2144 x264_weight_scale_plane( h, dst, stride, src, stride, width, height, &h->sh.weight[j][0] );
2145 h->fenc->i_lines_weighted = height;
2153 if( weightplane[1] )
2154 for( int i = 0; i < h->i_ref[0]; i++ )
2156 if( h->sh.weight[i][1].weightfn && !h->sh.weight[i][2].weightfn )
2158 h->sh.weight[i][2].i_scale = 1 << h->sh.weight[0][1].i_denom;
2159 h->sh.weight[i][2].i_offset = 0;
2161 else if( h->sh.weight[i][2].weightfn && !h->sh.weight[i][1].weightfn )
2163 h->sh.weight[i][1].i_scale = 1 << h->sh.weight[0][1].i_denom;
2164 h->sh.weight[i][1].i_offset = 0;
2168 if( !weightplane[0] )
2169 h->sh.weight[0][0].i_denom = 0;
2170 if( !weightplane[1] )
2171 h->sh.weight[0][1].i_denom = 0;
2172 h->sh.weight[0][2].i_denom = h->sh.weight[0][1].i_denom;
2175 static inline int x264_reference_distance( x264_t *h, x264_frame_t *frame )
2177 if( h->param.i_frame_packing == 5 )
2178 return abs((h->fenc->i_frame&~1) - (frame->i_frame&~1)) +
2179 ((h->fenc->i_frame&1) != (frame->i_frame&1));
2181 return abs(h->fenc->i_frame - frame->i_frame);
2184 static inline void x264_reference_build_list( x264_t *h, int i_poc )
2188 /* build ref list 0/1 */
2189 h->mb.pic.i_fref[0] = h->i_ref[0] = 0;
2190 h->mb.pic.i_fref[1] = h->i_ref[1] = 0;
2191 if( h->sh.i_type == SLICE_TYPE_I )
2194 for( int i = 0; h->frames.reference[i]; i++ )
2196 if( h->frames.reference[i]->b_corrupt )
2198 if( h->frames.reference[i]->i_poc < i_poc )
2199 h->fref[0][h->i_ref[0]++] = h->frames.reference[i];
2200 else if( h->frames.reference[i]->i_poc > i_poc )
2201 h->fref[1][h->i_ref[1]++] = h->frames.reference[i];
2204 if( h->sh.i_mmco_remove_from_end )
2206 /* Order ref0 for MMCO remove */
2210 for( int i = 0; i < h->i_ref[0] - 1; i++ )
2212 if( h->fref[0][i]->i_frame < h->fref[0][i+1]->i_frame )
2214 XCHG( x264_frame_t*, h->fref[0][i], h->fref[0][i+1] );
2221 for( int i = h->i_ref[0]-1; i >= h->i_ref[0] - h->sh.i_mmco_remove_from_end; i-- )
2223 int diff = h->i_frame_num - h->fref[0][i]->i_frame_num;
2224 h->sh.mmco[h->sh.i_mmco_command_count].i_poc = h->fref[0][i]->i_poc;
2225 h->sh.mmco[h->sh.i_mmco_command_count++].i_difference_of_pic_nums = diff;
2229 /* Order reference lists by distance from the current frame. */
2230 for( int list = 0; list < 2; list++ )
2232 h->fref_nearest[list] = h->fref[list][0];
2236 for( int i = 0; i < h->i_ref[list] - 1; i++ )
2238 if( list ? h->fref[list][i+1]->i_poc < h->fref_nearest[list]->i_poc
2239 : h->fref[list][i+1]->i_poc > h->fref_nearest[list]->i_poc )
2240 h->fref_nearest[list] = h->fref[list][i+1];
2241 if( x264_reference_distance( h, h->fref[list][i] ) > x264_reference_distance( h, h->fref[list][i+1] ) )
2243 XCHG( x264_frame_t*, h->fref[list][i], h->fref[list][i+1] );
2251 x264_reference_check_reorder( h );
2253 h->i_ref[1] = X264_MIN( h->i_ref[1], h->frames.i_max_ref1 );
2254 h->i_ref[0] = X264_MIN( h->i_ref[0], h->frames.i_max_ref0 );
2255 h->i_ref[0] = X264_MIN( h->i_ref[0], h->param.i_frame_reference ); // if reconfig() has lowered the limit
2257 /* For Blu-ray compliance, don't reference frames outside of the minigop. */
2258 if( IS_X264_TYPE_B( h->fenc->i_type ) && h->param.b_bluray_compat )
2259 h->i_ref[0] = X264_MIN( h->i_ref[0], IS_X264_TYPE_B( h->fref[0][0]->i_type ) + 1 );
2261 /* add duplicates */
2262 if( h->fenc->i_type == X264_TYPE_P )
2265 if( h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE )
2268 w[1].weightfn = w[2].weightfn = NULL;
2269 if( h->param.rc.b_stat_read )
2270 x264_ratecontrol_set_weights( h, h->fenc );
2272 if( !h->fenc->weight[0][0].weightfn )
2274 h->fenc->weight[0][0].i_denom = 0;
2275 SET_WEIGHT( w[0], 1, 1, 0, -1 );
2276 idx = x264_weighted_reference_duplicate( h, 0, w );
2280 if( h->fenc->weight[0][0].i_scale == 1<<h->fenc->weight[0][0].i_denom )
2282 SET_WEIGHT( h->fenc->weight[0][0], 1, 1, 0, h->fenc->weight[0][0].i_offset );
2284 x264_weighted_reference_duplicate( h, 0, x264_weight_none );
2285 if( h->fenc->weight[0][0].i_offset > -128 )
2287 w[0] = h->fenc->weight[0][0];
2289 h->mc.weight_cache( h, &w[0] );
2290 idx = x264_weighted_reference_duplicate( h, 0, w );
2294 h->mb.ref_blind_dupe = idx;
2297 assert( h->i_ref[0] + h->i_ref[1] <= X264_REF_MAX );
2298 h->mb.pic.i_fref[0] = h->i_ref[0];
2299 h->mb.pic.i_fref[1] = h->i_ref[1];
2302 static void x264_fdec_filter_row( x264_t *h, int mb_y, int pass )
2304 /* mb_y is the mb to be encoded next, not the mb to be filtered here */
2305 int b_hpel = h->fdec->b_kept_as_ref;
2306 int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
2307 int b_end = mb_y == h->i_threadslice_end;
2308 int b_measure_quality = 1;
2309 int min_y = mb_y - (1 << SLICE_MBAFF);
2310 int b_start = min_y == h->i_threadslice_start;
2311 /* Even in interlaced mode, deblocking never modifies more than 4 pixels
2312 * above each MB, as bS=4 doesn't happen for the top of interlaced mbpairs. */
2313 int minpix_y = min_y*16 - 4 * !b_start;
2314 int maxpix_y = mb_y*16 - 4 * !b_end;
2315 b_deblock &= b_hpel || h->param.b_full_recon || h->param.psz_dump_yuv;
2316 if( h->param.b_sliced_threads )
2320 /* During encode: only do deblock if asked for */
2323 b_deblock &= h->param.b_full_recon;
2326 /* During post-encode pass: do deblock if not done yet, do hpel for all
2327 * rows except those between slices. */
2329 b_deblock &= !h->param.b_full_recon;
2330 b_hpel &= !(b_start && min_y > 0);
2331 b_measure_quality = 0;
2333 /* Final pass: do the rows between slices in sequence. */
2336 b_measure_quality = 0;
2340 if( mb_y & SLICE_MBAFF )
2342 if( min_y < h->i_threadslice_start )
2346 for( int y = min_y; y < mb_y; y += (1 << SLICE_MBAFF) )
2347 x264_frame_deblock_row( h, y );
2349 /* FIXME: Prediction requires different borders for interlaced/progressive mc,
2350 * but the actual image data is equivalent. For now, maintain this
2351 * consistency by copying deblocked pixels between planes. */
2352 if( PARAM_INTERLACED && (!h->param.b_sliced_threads || pass == 1) )
2353 for( int p = 0; p < h->fdec->i_plane; p++ )
2354 for( int i = minpix_y>>(CHROMA_V_SHIFT && p); i < maxpix_y>>(CHROMA_V_SHIFT && p); i++ )
2355 memcpy( h->fdec->plane_fld[p] + i*h->fdec->i_stride[p],
2356 h->fdec->plane[p] + i*h->fdec->i_stride[p],
2357 h->mb.i_mb_width*16*sizeof(pixel) );
2359 if( h->fdec->b_kept_as_ref && (!h->param.b_sliced_threads || pass == 1) )
2360 x264_frame_expand_border( h, h->fdec, min_y );
2363 int end = mb_y == h->mb.i_mb_height;
2364 /* Can't do hpel until the previous slice is done encoding. */
2365 if( h->param.analyse.i_subpel_refine )
2367 x264_frame_filter( h, h->fdec, min_y, end );
2368 x264_frame_expand_border_filtered( h, h->fdec, min_y, end );
2372 if( SLICE_MBAFF && pass == 0 )
2373 for( int i = 0; i < 3; i++ )
2375 XCHG( pixel *, h->intra_border_backup[0][i], h->intra_border_backup[3][i] );
2376 XCHG( pixel *, h->intra_border_backup[1][i], h->intra_border_backup[4][i] );
2379 if( h->i_thread_frames > 1 && h->fdec->b_kept_as_ref )
2380 x264_frame_cond_broadcast( h->fdec, mb_y*16 + (b_end ? 10000 : -(X264_THREAD_HEIGHT << SLICE_MBAFF)) );
2382 if( b_measure_quality )
2384 maxpix_y = X264_MIN( maxpix_y, h->param.i_height );
2385 if( h->param.analyse.b_psnr )
2387 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
2388 h->stat.frame.i_ssd[p] += x264_pixel_ssd_wxh( &h->pixf,
2389 h->fdec->plane[p] + minpix_y * h->fdec->i_stride[p], h->fdec->i_stride[p],
2390 h->fenc->plane[p] + minpix_y * h->fenc->i_stride[p], h->fenc->i_stride[p],
2391 h->param.i_width, maxpix_y-minpix_y );
2394 uint64_t ssd_u, ssd_v;
2395 int v_shift = CHROMA_V_SHIFT;
2396 x264_pixel_ssd_nv12( &h->pixf,
2397 h->fdec->plane[1] + (minpix_y>>v_shift) * h->fdec->i_stride[1], h->fdec->i_stride[1],
2398 h->fenc->plane[1] + (minpix_y>>v_shift) * h->fenc->i_stride[1], h->fenc->i_stride[1],
2399 h->param.i_width>>1, (maxpix_y-minpix_y)>>v_shift, &ssd_u, &ssd_v );
2400 h->stat.frame.i_ssd[1] += ssd_u;
2401 h->stat.frame.i_ssd[2] += ssd_v;
2405 if( h->param.analyse.b_ssim )
2409 /* offset by 2 pixels to avoid alignment of ssim blocks with dct blocks,
2410 * and overlap by 4 */
2411 minpix_y += b_start ? 2 : -6;
2412 h->stat.frame.f_ssim +=
2413 x264_pixel_ssim_wxh( &h->pixf,
2414 h->fdec->plane[0] + 2+minpix_y*h->fdec->i_stride[0], h->fdec->i_stride[0],
2415 h->fenc->plane[0] + 2+minpix_y*h->fenc->i_stride[0], h->fenc->i_stride[0],
2416 h->param.i_width-2, maxpix_y-minpix_y, h->scratch_buffer, &ssim_cnt );
2417 h->stat.frame.i_ssim_cnt += ssim_cnt;
2422 static inline int x264_reference_update( x264_t *h )
2424 if( !h->fdec->b_kept_as_ref )
2426 if( h->i_thread_frames > 1 )
2428 x264_frame_push_unused( h, h->fdec );
2429 h->fdec = x264_frame_pop_unused( h, 1 );
2436 /* apply mmco from previous frame. */
2437 for( int i = 0; i < h->sh.i_mmco_command_count; i++ )
2438 for( int j = 0; h->frames.reference[j]; j++ )
2439 if( h->frames.reference[j]->i_poc == h->sh.mmco[i].i_poc )
2440 x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[j] ) );
2442 /* move frame in the buffer */
2443 x264_frame_push( h->frames.reference, h->fdec );
2444 if( h->frames.reference[h->sps->i_num_ref_frames] )
2445 x264_frame_push_unused( h, x264_frame_shift( h->frames.reference ) );
2446 h->fdec = x264_frame_pop_unused( h, 1 );
2452 static inline void x264_reference_reset( x264_t *h )
2454 while( h->frames.reference[0] )
2455 x264_frame_push_unused( h, x264_frame_pop( h->frames.reference ) );
2460 static inline void x264_reference_hierarchy_reset( x264_t *h )
2463 int b_hasdelayframe = 0;
2465 /* look for delay frames -- chain must only contain frames that are disposable */
2466 for( int i = 0; h->frames.current[i] && IS_DISPOSABLE( h->frames.current[i]->i_type ); i++ )
2467 b_hasdelayframe |= h->frames.current[i]->i_coded
2468 != h->frames.current[i]->i_frame + h->sps->vui.i_num_reorder_frames;
2470 /* This function must handle b-pyramid and clear frames for open-gop */
2471 if( h->param.i_bframe_pyramid != X264_B_PYRAMID_STRICT && !b_hasdelayframe && h->frames.i_poc_last_open_gop == -1 )
2474 /* Remove last BREF. There will never be old BREFs in the
2475 * dpb during a BREF decode when pyramid == STRICT */
2476 for( ref = 0; h->frames.reference[ref]; ref++ )
2478 if( ( h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT
2479 && h->frames.reference[ref]->i_type == X264_TYPE_BREF )
2480 || ( h->frames.reference[ref]->i_poc < h->frames.i_poc_last_open_gop
2481 && h->sh.i_type != SLICE_TYPE_B ) )
2483 int diff = h->i_frame_num - h->frames.reference[ref]->i_frame_num;
2484 h->sh.mmco[h->sh.i_mmco_command_count].i_difference_of_pic_nums = diff;
2485 h->sh.mmco[h->sh.i_mmco_command_count++].i_poc = h->frames.reference[ref]->i_poc;
2486 x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[ref] ) );
2487 h->b_ref_reorder[0] = 1;
2492 /* Prepare room in the dpb for the delayed display time of the later b-frame's */
2493 if( h->param.i_bframe_pyramid )
2494 h->sh.i_mmco_remove_from_end = X264_MAX( ref + 2 - h->frames.i_max_dpb, 0 );
2497 static inline void x264_slice_init( x264_t *h, int i_nal_type, int i_global_qp )
2499 /* ------------------------ Create slice header ----------------------- */
2500 if( i_nal_type == NAL_SLICE_IDR )
2502 x264_slice_header_init( h, &h->sh, h->sps, h->pps, h->i_idr_pic_id, h->i_frame_num, i_global_qp );
2505 if( h->param.i_avcintra_class )
2507 switch( h->i_idr_pic_id )
2510 h->i_idr_pic_id = 3;
2513 h->i_idr_pic_id = 4;
2517 h->i_idr_pic_id = 5;
2522 h->i_idr_pic_id ^= 1;
2526 x264_slice_header_init( h, &h->sh, h->sps, h->pps, -1, h->i_frame_num, i_global_qp );
2528 h->sh.i_num_ref_idx_l0_active = h->i_ref[0] <= 0 ? 1 : h->i_ref[0];
2529 h->sh.i_num_ref_idx_l1_active = h->i_ref[1] <= 0 ? 1 : h->i_ref[1];
2530 if( h->sh.i_num_ref_idx_l0_active != h->pps->i_num_ref_idx_l0_default_active ||
2531 (h->sh.i_type == SLICE_TYPE_B && h->sh.i_num_ref_idx_l1_active != h->pps->i_num_ref_idx_l1_default_active) )
2533 h->sh.b_num_ref_idx_override = 1;
2537 if( h->fenc->i_type == X264_TYPE_BREF && h->param.b_bluray_compat && h->sh.i_mmco_command_count )
2540 h->sh_backup = h->sh;
2543 h->fdec->i_frame_num = h->sh.i_frame_num;
2545 if( h->sps->i_poc_type == 0 )
2547 h->sh.i_poc = h->fdec->i_poc;
2548 if( PARAM_INTERLACED )
2550 h->sh.i_delta_poc_bottom = h->param.b_tff ? 1 : -1;
2551 h->sh.i_poc += h->sh.i_delta_poc_bottom == -1;
2554 h->sh.i_delta_poc_bottom = 0;
2555 h->fdec->i_delta_poc[0] = h->sh.i_delta_poc_bottom == -1;
2556 h->fdec->i_delta_poc[1] = h->sh.i_delta_poc_bottom == 1;
2560 /* Nothing to do ? */
2563 x264_macroblock_slice_init( h );
2569 uint8_t cabac_prevbyte;
2572 x264_frame_stat_t stat;
2575 int field_decoding_flag;
2578 static ALWAYS_INLINE void x264_bitstream_backup( x264_t *h, x264_bs_bak_t *bak, int i_skip, int full )
2582 bak->stat = h->stat.frame;
2583 bak->last_qp = h->mb.i_last_qp;
2584 bak->last_dqp = h->mb.i_last_dqp;
2585 bak->field_decoding_flag = h->mb.field_decoding_flag;
2589 bak->stat.i_mv_bits = h->stat.frame.i_mv_bits;
2590 bak->stat.i_tex_bits = h->stat.frame.i_tex_bits;
2592 /* In the per-MB backup, we don't need the contexts because flushing the CABAC
2593 * encoder has no context dependency and in this case, a slice is ended (and
2594 * thus the content of all contexts are thrown away). */
2595 if( h->param.b_cabac )
2598 memcpy( &bak->cabac, &h->cabac, sizeof(x264_cabac_t) );
2600 memcpy( &bak->cabac, &h->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
2601 /* x264's CABAC writer modifies the previous byte during carry, so it has to be
2603 bak->cabac_prevbyte = h->cabac.p[-1];
2607 bak->bs = h->out.bs;
2612 static ALWAYS_INLINE void x264_bitstream_restore( x264_t *h, x264_bs_bak_t *bak, int *skip, int full )
2616 h->stat.frame = bak->stat;
2617 h->mb.i_last_qp = bak->last_qp;
2618 h->mb.i_last_dqp = bak->last_dqp;
2619 h->mb.field_decoding_flag = bak->field_decoding_flag;
2623 h->stat.frame.i_mv_bits = bak->stat.i_mv_bits;
2624 h->stat.frame.i_tex_bits = bak->stat.i_tex_bits;
2626 if( h->param.b_cabac )
2629 memcpy( &h->cabac, &bak->cabac, sizeof(x264_cabac_t) );
2631 memcpy( &h->cabac, &bak->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
2632 h->cabac.p[-1] = bak->cabac_prevbyte;
2636 h->out.bs = bak->bs;
2641 static intptr_t x264_slice_write( x264_t *h )
2644 int mb_xy, i_mb_x, i_mb_y;
2645 /* NALUs other than the first use a 3-byte startcode.
2646 * Add one extra byte for the rbsp, and one more for the final CABAC putbyte.
2647 * Then add an extra 5 bytes just in case, to account for random NAL escapes and
2648 * other inaccuracies. */
2649 int overhead_guess = (NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal)) + 1 + h->param.b_cabac + 5;
2650 int slice_max_size = h->param.i_slice_max_size > 0 ? (h->param.i_slice_max_size-overhead_guess)*8 : 0;
2651 int back_up_bitstream_cavlc = !h->param.b_cabac && h->sps->i_profile_idc < PROFILE_HIGH;
2652 int back_up_bitstream = slice_max_size || back_up_bitstream_cavlc;
2653 int starting_bits = bs_pos(&h->out.bs);
2654 int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
2655 int b_hpel = h->fdec->b_kept_as_ref;
2656 int orig_last_mb = h->sh.i_last_mb;
2657 int thread_last_mb = h->i_threadslice_end * h->mb.i_mb_width - 1;
2658 uint8_t *last_emu_check;
2659 #define BS_BAK_SLICE_MAX_SIZE 0
2660 #define BS_BAK_CAVLC_OVERFLOW 1
2661 #define BS_BAK_SLICE_MIN_MBS 2
2662 #define BS_BAK_ROW_VBV 3
2663 x264_bs_bak_t bs_bak[4];
2664 b_deblock &= b_hpel || h->param.b_full_recon || h->param.psz_dump_yuv;
2665 bs_realign( &h->out.bs );
2668 x264_nal_start( h, h->i_nal_type, h->i_nal_ref_idc );
2669 h->out.nal[h->out.i_nal].i_first_mb = h->sh.i_first_mb;
2672 x264_macroblock_thread_init( h );
2674 /* Set the QP equal to the first QP in the slice for more accurate CABAC initialization. */
2675 h->mb.i_mb_xy = h->sh.i_first_mb;
2676 h->sh.i_qp = x264_ratecontrol_mb_qp( h );
2677 h->sh.i_qp = SPEC_QP( h->sh.i_qp );
2678 h->sh.i_qp_delta = h->sh.i_qp - h->pps->i_pic_init_qp;
2680 x264_slice_header_write( &h->out.bs, &h->sh, h->i_nal_ref_idc );
2681 if( h->param.b_cabac )
2683 /* alignment needed */
2684 bs_align_1( &h->out.bs );
2687 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 );
2688 x264_cabac_encode_init ( &h->cabac, h->out.bs.p, h->out.bs.p_end );
2689 last_emu_check = h->cabac.p;
2692 last_emu_check = h->out.bs.p;
2693 h->mb.i_last_qp = h->sh.i_qp;
2694 h->mb.i_last_dqp = 0;
2695 h->mb.field_decoding_flag = 0;
2697 i_mb_y = h->sh.i_first_mb / h->mb.i_mb_width;
2698 i_mb_x = h->sh.i_first_mb % h->mb.i_mb_width;
2703 mb_xy = i_mb_x + i_mb_y * h->mb.i_mb_width;
2704 int mb_spos = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
2708 if( x264_bitstream_check_buffer( h ) )
2710 if( !(i_mb_y & SLICE_MBAFF) && h->param.rc.i_vbv_buffer_size )
2711 x264_bitstream_backup( h, &bs_bak[BS_BAK_ROW_VBV], i_skip, 1 );
2712 if( !h->mb.b_reencode_mb )
2713 x264_fdec_filter_row( h, i_mb_y, 0 );
2716 if( back_up_bitstream )
2718 if( back_up_bitstream_cavlc )
2719 x264_bitstream_backup( h, &bs_bak[BS_BAK_CAVLC_OVERFLOW], i_skip, 0 );
2720 if( slice_max_size && !(i_mb_y & SLICE_MBAFF) )
2722 x264_bitstream_backup( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], i_skip, 0 );
2723 if( (thread_last_mb+1-mb_xy) == h->param.i_slice_min_mbs )
2724 x264_bitstream_backup( h, &bs_bak[BS_BAK_SLICE_MIN_MBS], i_skip, 0 );
2728 if( PARAM_INTERLACED )
2730 if( h->mb.b_adaptive_mbaff )
2734 /* FIXME: VSAD is fast but fairly poor at choosing the best interlace type. */
2735 h->mb.b_interlaced = x264_field_vsad( h, i_mb_x, i_mb_y );
2736 memcpy( &h->zigzagf, MB_INTERLACED ? &h->zigzagf_interlaced : &h->zigzagf_progressive, sizeof(h->zigzagf) );
2737 if( !MB_INTERLACED && (i_mb_y+2) == h->mb.i_mb_height )
2738 x264_expand_border_mbpair( h, i_mb_x, i_mb_y );
2741 h->mb.field[mb_xy] = MB_INTERLACED;
2746 x264_macroblock_cache_load_interlaced( h, i_mb_x, i_mb_y );
2748 x264_macroblock_cache_load_progressive( h, i_mb_x, i_mb_y );
2750 x264_macroblock_analyse( h );
2752 /* encode this macroblock -> be careful it can change the mb type to P_SKIP if needed */
2754 x264_macroblock_encode( h );
2756 if( h->param.b_cabac )
2758 if( mb_xy > h->sh.i_first_mb && !(SLICE_MBAFF && (i_mb_y&1)) )
2759 x264_cabac_encode_terminal( &h->cabac );
2761 if( IS_SKIP( h->mb.i_type ) )
2762 x264_cabac_mb_skip( h, 1 );
2765 if( h->sh.i_type != SLICE_TYPE_I )
2766 x264_cabac_mb_skip( h, 0 );
2767 x264_macroblock_write_cabac( h, &h->cabac );
2772 if( IS_SKIP( h->mb.i_type ) )
2776 if( h->sh.i_type != SLICE_TYPE_I )
2778 bs_write_ue( &h->out.bs, i_skip ); /* skip run */
2781 x264_macroblock_write_cavlc( h );
2782 /* If there was a CAVLC level code overflow, try again at a higher QP. */
2783 if( h->mb.b_overflow )
2785 h->mb.i_chroma_qp = h->chroma_qp_table[++h->mb.i_qp];
2786 h->mb.i_skip_intra = 0;
2787 h->mb.b_skip_mc = 0;
2788 h->mb.b_overflow = 0;
2789 x264_bitstream_restore( h, &bs_bak[BS_BAK_CAVLC_OVERFLOW], &i_skip, 0 );
2795 int total_bits = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
2796 int mb_size = total_bits - mb_spos;
2798 if( slice_max_size && (!SLICE_MBAFF || (i_mb_y&1)) )
2800 /* Count the skip run, just in case. */
2801 if( !h->param.b_cabac )
2802 total_bits += bs_size_ue_big( i_skip );
2803 /* Check for escape bytes. */
2804 uint8_t *end = h->param.b_cabac ? h->cabac.p : h->out.bs.p;
2805 for( ; last_emu_check < end - 2; last_emu_check++ )
2806 if( last_emu_check[0] == 0 && last_emu_check[1] == 0 && last_emu_check[2] <= 3 )
2808 slice_max_size -= 8;
2811 /* We'll just re-encode this last macroblock if we go over the max slice size. */
2812 if( total_bits - starting_bits > slice_max_size && !h->mb.b_reencode_mb )
2814 if( !x264_frame_new_slice( h, h->fdec ) )
2816 /* Handle the most obnoxious slice-min-mbs edge case: we need to end the slice
2817 * because it's gone over the maximum size, but doing so would violate slice-min-mbs.
2818 * If possible, roll back to the last checkpoint and try again.
2819 * We could try raising QP, but that would break in the case where a slice spans multiple
2820 * rows, which the re-encoding infrastructure can't currently handle. */
2821 if( mb_xy <= thread_last_mb && (thread_last_mb+1-mb_xy) < h->param.i_slice_min_mbs )
2823 if( thread_last_mb-h->param.i_slice_min_mbs < h->sh.i_first_mb+h->param.i_slice_min_mbs )
2825 x264_log( h, X264_LOG_WARNING, "slice-max-size violated (frame %d, cause: slice-min-mbs)\n", h->i_frame );
2829 x264_bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MIN_MBS], &i_skip, 0 );
2830 h->mb.b_reencode_mb = 1;
2831 h->sh.i_last_mb = thread_last_mb-h->param.i_slice_min_mbs;
2834 if( mb_xy-SLICE_MBAFF*h->mb.i_mb_stride != h->sh.i_first_mb )
2836 x264_bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], &i_skip, 0 );
2837 h->mb.b_reencode_mb = 1;
2840 // set to bottom of previous mbpair
2842 h->sh.i_last_mb = mb_xy-1+h->mb.i_mb_stride*(!(i_mb_y&1));
2844 h->sh.i_last_mb = (i_mb_y-2+!(i_mb_y&1))*h->mb.i_mb_stride + h->mb.i_mb_width - 1;
2847 h->sh.i_last_mb = mb_xy-1;
2851 h->sh.i_last_mb = mb_xy;
2858 h->mb.b_reencode_mb = 0;
2861 x264_macroblock_cache_save( h );
2863 if( x264_ratecontrol_mb( h, mb_size ) < 0 )
2865 x264_bitstream_restore( h, &bs_bak[BS_BAK_ROW_VBV], &i_skip, 1 );
2866 h->mb.b_reencode_mb = 1;
2868 i_mb_y = i_mb_y - SLICE_MBAFF;
2869 h->mb.i_mb_prev_xy = i_mb_y * h->mb.i_mb_stride - 1;
2870 h->sh.i_last_mb = orig_last_mb;
2874 /* accumulate mb stats */
2875 h->stat.frame.i_mb_count[h->mb.i_type]++;
2877 int b_intra = IS_INTRA( h->mb.i_type );
2878 int b_skip = IS_SKIP( h->mb.i_type );
2879 if( h->param.i_log_level >= X264_LOG_INFO || h->param.rc.b_stat_write )
2881 if( !b_intra && !b_skip && !IS_DIRECT( h->mb.i_type ) )
2883 if( h->mb.i_partition != D_8x8 )
2884 h->stat.frame.i_mb_partition[h->mb.i_partition] += 4;
2886 for( int i = 0; i < 4; i++ )
2887 h->stat.frame.i_mb_partition[h->mb.i_sub_partition[i]] ++;
2888 if( h->param.i_frame_reference > 1 )
2889 for( int i_list = 0; i_list <= (h->sh.i_type == SLICE_TYPE_B); i_list++ )
2890 for( int i = 0; i < 4; i++ )
2892 int i_ref = h->mb.cache.ref[i_list][ x264_scan8[4*i] ];
2894 h->stat.frame.i_mb_count_ref[i_list][i_ref] ++;
2899 if( h->param.i_log_level >= X264_LOG_INFO )
2901 if( h->mb.i_cbp_luma | h->mb.i_cbp_chroma )
2905 for( int i = 0; i < 4; i++ )
2906 if( h->mb.i_cbp_luma & (1 << i) )
2907 for( int p = 0; p < 3; p++ )
2910 int nnz8x8 = M16( &h->mb.cache.non_zero_count[x264_scan8[s8]+0] )
2911 | M16( &h->mb.cache.non_zero_count[x264_scan8[s8]+8] );
2912 h->stat.frame.i_mb_cbp[!b_intra + p*2] += !!nnz8x8;
2917 int cbpsum = (h->mb.i_cbp_luma&1) + ((h->mb.i_cbp_luma>>1)&1)
2918 + ((h->mb.i_cbp_luma>>2)&1) + (h->mb.i_cbp_luma>>3);
2919 h->stat.frame.i_mb_cbp[!b_intra + 0] += cbpsum;
2920 h->stat.frame.i_mb_cbp[!b_intra + 2] += !!h->mb.i_cbp_chroma;
2921 h->stat.frame.i_mb_cbp[!b_intra + 4] += h->mb.i_cbp_chroma >> 1;
2924 if( h->mb.i_cbp_luma && !b_intra )
2926 h->stat.frame.i_mb_count_8x8dct[0] ++;
2927 h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
2929 if( b_intra && h->mb.i_type != I_PCM )
2931 if( h->mb.i_type == I_16x16 )
2932 h->stat.frame.i_mb_pred_mode[0][h->mb.i_intra16x16_pred_mode]++;
2933 else if( h->mb.i_type == I_8x8 )
2934 for( int i = 0; i < 16; i += 4 )
2935 h->stat.frame.i_mb_pred_mode[1][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
2936 else //if( h->mb.i_type == I_4x4 )
2937 for( int i = 0; i < 16; i++ )
2938 h->stat.frame.i_mb_pred_mode[2][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
2939 h->stat.frame.i_mb_pred_mode[3][x264_mb_chroma_pred_mode_fix[h->mb.i_chroma_pred_mode]]++;
2941 h->stat.frame.i_mb_field[b_intra?0:b_skip?2:1] += MB_INTERLACED;
2944 /* calculate deblock strength values (actual deblocking is done per-row along with hpel) */
2946 x264_macroblock_deblock_strength( h );
2948 if( mb_xy == h->sh.i_last_mb )
2953 i_mb_x += i_mb_y & 1;
2954 i_mb_y ^= i_mb_x < h->mb.i_mb_width;
2958 if( i_mb_x == h->mb.i_mb_width )
2964 if( h->sh.i_last_mb < h->sh.i_first_mb )
2967 h->out.nal[h->out.i_nal].i_last_mb = h->sh.i_last_mb;
2969 if( h->param.b_cabac )
2971 x264_cabac_encode_flush( h, &h->cabac );
2972 h->out.bs.p = h->cabac.p;
2977 bs_write_ue( &h->out.bs, i_skip ); /* last skip run */
2978 /* rbsp_slice_trailing_bits */
2979 bs_rbsp_trailing( &h->out.bs );
2980 bs_flush( &h->out.bs );
2982 if( x264_nal_end( h ) )
2985 if( h->sh.i_last_mb == (h->i_threadslice_end * h->mb.i_mb_width - 1) )
2987 h->stat.frame.i_misc_bits = bs_pos( &h->out.bs )
2988 + (h->out.i_nal*NALU_OVERHEAD * 8)
2989 - h->stat.frame.i_tex_bits
2990 - h->stat.frame.i_mv_bits;
2991 x264_fdec_filter_row( h, h->i_threadslice_end, 0 );
2993 if( h->param.b_sliced_threads )
2995 /* Tell the main thread we're done. */
2996 x264_threadslice_cond_broadcast( h, 1 );
2998 for( int mb_y = h->i_threadslice_start; mb_y <= h->i_threadslice_end; mb_y++ )
2999 x264_fdec_filter_row( h, mb_y, 1 );
3000 x264_threadslice_cond_broadcast( h, 2 );
3001 /* Do the first row of hpel, now that the previous slice is done */
3002 if( h->i_thread_idx > 0 )
3004 x264_threadslice_cond_wait( h->thread[h->i_thread_idx-1], 2 );
3005 x264_fdec_filter_row( h, h->i_threadslice_start + (1 << SLICE_MBAFF), 2 );
3009 /* Free mb info after the last thread's done using it */
3010 if( h->fdec->mb_info_free && (!h->param.b_sliced_threads || h->i_thread_idx == (h->param.i_threads-1)) )
3012 h->fdec->mb_info_free( h->fdec->mb_info );
3013 h->fdec->mb_info = NULL;
3014 h->fdec->mb_info_free = NULL;
3021 static void x264_thread_sync_context( x264_t *dst, x264_t *src )
3026 // reference counting
3027 for( x264_frame_t **f = src->frames.reference; *f; f++ )
3028 (*f)->i_reference_count++;
3029 for( x264_frame_t **f = dst->frames.reference; *f; f++ )
3030 x264_frame_push_unused( src, *f );
3031 src->fdec->i_reference_count++;
3032 x264_frame_push_unused( src, dst->fdec );
3034 // copy everything except the per-thread pointers and the constants.
3035 memcpy( &dst->i_frame, &src->i_frame, offsetof(x264_t, mb.base) - offsetof(x264_t, i_frame) );
3036 dst->param = src->param;
3037 dst->stat = src->stat;
3038 dst->pixf = src->pixf;
3039 dst->reconfig = src->reconfig;
3042 static void x264_thread_sync_stat( x264_t *dst, x264_t *src )
3046 memcpy( &dst->stat.i_frame_count, &src->stat.i_frame_count, sizeof(dst->stat) - sizeof(dst->stat.frame) );
3049 static void *x264_slices_write( x264_t *h )
3051 int i_slice_num = 0;
3052 int last_thread_mb = h->sh.i_last_mb;
3055 memset( &h->stat.frame, 0, sizeof(h->stat.frame) );
3056 h->mb.b_reencode_mb = 0;
3057 while( h->sh.i_first_mb + SLICE_MBAFF*h->mb.i_mb_stride <= last_thread_mb )
3059 h->sh.i_last_mb = last_thread_mb;
3060 if( !i_slice_num || !x264_frame_new_slice( h, h->fdec ) )
3062 if( h->param.i_slice_max_mbs )
3066 // convert first to mbaff form, add slice-max-mbs, then convert back to normal form
3067 int last_mbaff = 2*(h->sh.i_first_mb % h->mb.i_mb_width)
3068 + h->mb.i_mb_width*(h->sh.i_first_mb / h->mb.i_mb_width)
3069 + h->param.i_slice_max_mbs - 1;
3070 int last_x = (last_mbaff % (2*h->mb.i_mb_width))/2;
3071 int last_y = (last_mbaff / (2*h->mb.i_mb_width))*2 + 1;
3072 h->sh.i_last_mb = last_x + h->mb.i_mb_stride*last_y;
3076 h->sh.i_last_mb = h->sh.i_first_mb + h->param.i_slice_max_mbs - 1;
3077 if( h->sh.i_last_mb < last_thread_mb && last_thread_mb - h->sh.i_last_mb < h->param.i_slice_min_mbs )
3078 h->sh.i_last_mb = last_thread_mb - h->param.i_slice_min_mbs;
3082 else if( h->param.i_slice_count && !h->param.b_sliced_threads )
3084 int height = h->mb.i_mb_height >> PARAM_INTERLACED;
3085 int width = h->mb.i_mb_width << PARAM_INTERLACED;
3087 h->sh.i_last_mb = (height * i_slice_num + h->param.i_slice_count/2) / h->param.i_slice_count * width - 1;
3090 h->sh.i_last_mb = X264_MIN( h->sh.i_last_mb, last_thread_mb );
3091 if( x264_stack_align( x264_slice_write, h ) )
3093 h->sh.i_first_mb = h->sh.i_last_mb + 1;
3094 // if i_first_mb is not the last mb in a row then go to the next mb in MBAFF order
3095 if( SLICE_MBAFF && h->sh.i_first_mb % h->mb.i_mb_width )
3096 h->sh.i_first_mb -= h->mb.i_mb_stride;
3102 /* Tell other threads we're done, so they wouldn't wait for it */
3103 if( h->param.b_sliced_threads )
3104 x264_threadslice_cond_broadcast( h, 2 );
3108 static int x264_threaded_slices_write( x264_t *h )
3110 /* set first/last mb and sync contexts */
3111 for( int i = 0; i < h->param.i_threads; i++ )
3113 x264_t *t = h->thread[i];
3116 t->param = h->param;
3117 memcpy( &t->i_frame, &h->i_frame, offsetof(x264_t, rc) - offsetof(x264_t, i_frame) );
3119 int height = h->mb.i_mb_height >> PARAM_INTERLACED;
3120 t->i_threadslice_start = ((height * i + h->param.i_slice_count/2) / h->param.i_threads) << PARAM_INTERLACED;
3121 t->i_threadslice_end = ((height * (i+1) + h->param.i_slice_count/2) / h->param.i_threads) << PARAM_INTERLACED;
3122 t->sh.i_first_mb = t->i_threadslice_start * h->mb.i_mb_width;
3123 t->sh.i_last_mb = t->i_threadslice_end * h->mb.i_mb_width - 1;
3126 x264_stack_align( x264_analyse_weight_frame, h, h->mb.i_mb_height*16 + 16 );
3128 x264_threads_distribute_ratecontrol( h );
3131 for( int i = 0; i < h->param.i_threads; i++ )
3133 h->thread[i]->i_thread_idx = i;
3134 h->thread[i]->b_thread_active = 1;
3135 x264_threadslice_cond_broadcast( h->thread[i], 0 );
3138 for( int i = 0; i < h->param.i_threads; i++ )
3139 x264_threadpool_run( h->threadpool, (void*)x264_slices_write, h->thread[i] );
3141 for( int i = 0; i < h->param.i_threads; i++ )
3142 x264_threadslice_cond_wait( h->thread[i], 1 );
3144 x264_threads_merge_ratecontrol( h );
3146 for( int i = 1; i < h->param.i_threads; i++ )
3148 x264_t *t = h->thread[i];
3149 for( int j = 0; j < t->out.i_nal; j++ )
3151 h->out.nal[h->out.i_nal] = t->out.nal[j];
3153 x264_nal_check_buffer( h );
3155 /* All entries in stat.frame are ints except for ssd/ssim. */
3156 for( int j = 0; j < (offsetof(x264_t,stat.frame.i_ssd) - offsetof(x264_t,stat.frame.i_mv_bits)) / sizeof(int); j++ )
3157 ((int*)&h->stat.frame)[j] += ((int*)&t->stat.frame)[j];
3158 for( int j = 0; j < 3; j++ )
3159 h->stat.frame.i_ssd[j] += t->stat.frame.i_ssd[j];
3160 h->stat.frame.f_ssim += t->stat.frame.f_ssim;
3161 h->stat.frame.i_ssim_cnt += t->stat.frame.i_ssim_cnt;
3167 void x264_encoder_intra_refresh( x264_t *h )
3169 h = h->thread[h->i_thread_phase];
3170 h->b_queued_intra_refresh = 1;
3173 int x264_encoder_invalidate_reference( x264_t *h, int64_t pts )
3175 if( h->param.i_bframe )
3177 x264_log( h, X264_LOG_ERROR, "x264_encoder_invalidate_reference is not supported with B-frames enabled\n" );
3180 if( h->param.b_intra_refresh )
3182 x264_log( h, X264_LOG_ERROR, "x264_encoder_invalidate_reference is not supported with intra refresh enabled\n" );
3185 h = h->thread[h->i_thread_phase];
3186 if( pts >= h->i_last_idr_pts )
3188 for( int i = 0; h->frames.reference[i]; i++ )
3189 if( pts <= h->frames.reference[i]->i_pts )
3190 h->frames.reference[i]->b_corrupt = 1;
3191 if( pts <= h->fdec->i_pts )
3192 h->fdec->b_corrupt = 1;
3197 /****************************************************************************
3198 * x264_encoder_encode:
3199 * XXX: i_poc : is the poc of the current given picture
3200 * i_frame : is the number of the frame being coded
3201 * ex: type frame poc
3209 ****************************************************************************/
3210 int x264_encoder_encode( x264_t *h,
3211 x264_nal_t **pp_nal, int *pi_nal,
3212 x264_picture_t *pic_in,
3213 x264_picture_t *pic_out )
3215 x264_t *thread_current, *thread_prev, *thread_oldest;
3216 int i_nal_type, i_nal_ref_idc, i_global_qp;
3217 int overhead = NALU_OVERHEAD;
3220 if( h->opencl.b_fatal_error )
3224 if( h->i_thread_frames > 1 )
3226 thread_prev = h->thread[ h->i_thread_phase ];
3227 h->i_thread_phase = (h->i_thread_phase + 1) % h->i_thread_frames;
3228 thread_current = h->thread[ h->i_thread_phase ];
3229 thread_oldest = h->thread[ (h->i_thread_phase + 1) % h->i_thread_frames ];
3230 x264_thread_sync_context( thread_current, thread_prev );
3231 x264_thread_sync_ratecontrol( thread_current, thread_prev, thread_oldest );
3239 h->i_cpb_delay_pir_offset = h->i_cpb_delay_pir_offset_next;
3245 /* ------------------- Setup new frame from picture -------------------- */
3246 if( pic_in != NULL )
3248 if( h->lookahead->b_exit_thread )
3250 x264_log( h, X264_LOG_ERROR, "lookahead thread is already stopped\n" );
3254 /* 1: Copy the picture to a frame and move it to a buffer */
3255 x264_frame_t *fenc = x264_frame_pop_unused( h, 0 );
3259 if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 )
3262 if( h->param.i_width != 16 * h->mb.i_mb_width ||
3263 h->param.i_height != 16 * h->mb.i_mb_height )
3264 x264_frame_expand_border_mod16( h, fenc );
3266 fenc->i_frame = h->frames.i_input++;
3268 if( fenc->i_frame == 0 )
3269 h->frames.i_first_pts = fenc->i_pts;
3270 if( h->frames.i_bframe_delay && fenc->i_frame == h->frames.i_bframe_delay )
3271 h->frames.i_bframe_delay_time = fenc->i_pts - h->frames.i_first_pts;
3273 if( h->param.b_vfr_input && fenc->i_pts <= h->frames.i_largest_pts )
3274 x264_log( h, X264_LOG_WARNING, "non-strictly-monotonic PTS\n" );
3276 h->frames.i_second_largest_pts = h->frames.i_largest_pts;
3277 h->frames.i_largest_pts = fenc->i_pts;
3279 if( (fenc->i_pic_struct < PIC_STRUCT_AUTO) || (fenc->i_pic_struct > PIC_STRUCT_TRIPLE) )
3280 fenc->i_pic_struct = PIC_STRUCT_AUTO;
3282 if( fenc->i_pic_struct == PIC_STRUCT_AUTO )
3285 int b_interlaced = fenc->param ? fenc->param->b_interlaced : h->param.b_interlaced;
3287 int b_interlaced = 0;
3291 int b_tff = fenc->param ? fenc->param->b_tff : h->param.b_tff;
3292 fenc->i_pic_struct = b_tff ? PIC_STRUCT_TOP_BOTTOM : PIC_STRUCT_BOTTOM_TOP;
3295 fenc->i_pic_struct = PIC_STRUCT_PROGRESSIVE;
3298 if( h->param.rc.b_mb_tree && h->param.rc.b_stat_read )
3300 if( x264_macroblock_tree_read( h, fenc, pic_in->prop.quant_offsets ) )
3304 x264_stack_align( x264_adaptive_quant_frame, h, fenc, pic_in->prop.quant_offsets );
3306 if( pic_in->prop.quant_offsets_free )
3307 pic_in->prop.quant_offsets_free( pic_in->prop.quant_offsets );
3309 if( h->frames.b_have_lowres )
3310 x264_frame_init_lowres( h, fenc );
3312 /* 2: Place the frame into the queue for its slice type decision */
3313 x264_lookahead_put_frame( h, fenc );
3315 if( h->frames.i_input <= h->frames.i_delay + 1 - h->i_thread_frames )
3317 /* Nothing yet to encode, waiting for filling of buffers */
3318 pic_out->i_type = X264_TYPE_AUTO;
3324 /* signal kills for lookahead thread */
3325 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
3326 h->lookahead->b_exit_thread = 1;
3327 x264_pthread_cond_broadcast( &h->lookahead->ifbuf.cv_fill );
3328 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
3332 /* 3: The picture is analyzed in the lookahead */
3333 if( !h->frames.current[0] )
3334 x264_lookahead_get_frames( h );
3336 if( !h->frames.current[0] && x264_lookahead_is_empty( h ) )
3337 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
3339 /* ------------------- Get frame to be encoded ------------------------- */
3340 /* 4: get picture to encode */
3341 h->fenc = x264_frame_shift( h->frames.current );
3343 /* If applicable, wait for previous frame reconstruction to finish */
3344 if( h->param.b_sliced_threads )
3345 if( x264_threadpool_wait_all( h ) < 0 )
3348 if( h->i_frame == h->i_thread_frames - 1 )
3349 h->i_reordered_pts_delay = h->fenc->i_reordered_pts;
3352 x264_encoder_reconfig_apply( h, &h->reconfig_h->param );
3355 if( h->fenc->param )
3357 x264_encoder_reconfig_apply( h, h->fenc->param );
3358 if( h->fenc->param->param_free )
3360 h->fenc->param->param_free( h->fenc->param );
3361 h->fenc->param = NULL;
3365 // ok to call this before encoding any frames, since the initial values of fdec have b_kept_as_ref=0
3366 if( x264_reference_update( h ) )
3368 h->fdec->i_lines_completed = -1;
3370 if( !IS_X264_TYPE_I( h->fenc->i_type ) )
3372 int valid_refs_left = 0;
3373 for( int i = 0; h->frames.reference[i]; i++ )
3374 if( !h->frames.reference[i]->b_corrupt )
3376 /* No valid reference frames left: force an IDR. */
3377 if( !valid_refs_left )
3379 h->fenc->b_keyframe = 1;
3380 h->fenc->i_type = X264_TYPE_IDR;
3384 if( h->fenc->b_keyframe )
3386 h->frames.i_last_keyframe = h->fenc->i_frame;
3387 if( h->fenc->i_type == X264_TYPE_IDR )
3390 h->frames.i_last_idr = h->fenc->i_frame;
3393 h->sh.i_mmco_command_count =
3394 h->sh.i_mmco_remove_from_end = 0;
3395 h->b_ref_reorder[0] =
3396 h->b_ref_reorder[1] = 0;
3398 h->fenc->i_poc = 2 * ( h->fenc->i_frame - X264_MAX( h->frames.i_last_idr, 0 ) );
3400 /* ------------------- Setup frame context ----------------------------- */
3401 /* 5: Init data dependent of frame type */
3402 if( h->fenc->i_type == X264_TYPE_IDR )
3404 /* reset ref pictures */
3405 i_nal_type = NAL_SLICE_IDR;
3406 i_nal_ref_idc = NAL_PRIORITY_HIGHEST;
3407 h->sh.i_type = SLICE_TYPE_I;
3408 x264_reference_reset( h );
3409 h->frames.i_poc_last_open_gop = -1;
3411 else if( h->fenc->i_type == X264_TYPE_I )
3413 i_nal_type = NAL_SLICE;
3414 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
3415 h->sh.i_type = SLICE_TYPE_I;
3416 x264_reference_hierarchy_reset( h );
3417 if( h->param.b_open_gop )
3418 h->frames.i_poc_last_open_gop = h->fenc->b_keyframe ? h->fenc->i_poc : -1;
3420 else if( h->fenc->i_type == X264_TYPE_P )
3422 i_nal_type = NAL_SLICE;
3423 i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
3424 h->sh.i_type = SLICE_TYPE_P;
3425 x264_reference_hierarchy_reset( h );
3426 h->frames.i_poc_last_open_gop = -1;
3428 else if( h->fenc->i_type == X264_TYPE_BREF )
3430 i_nal_type = NAL_SLICE;
3431 i_nal_ref_idc = h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT ? NAL_PRIORITY_LOW : NAL_PRIORITY_HIGH;
3432 h->sh.i_type = SLICE_TYPE_B;
3433 x264_reference_hierarchy_reset( h );
3437 i_nal_type = NAL_SLICE;
3438 i_nal_ref_idc = NAL_PRIORITY_DISPOSABLE;
3439 h->sh.i_type = SLICE_TYPE_B;
3442 h->fdec->i_type = h->fenc->i_type;
3443 h->fdec->i_frame = h->fenc->i_frame;
3444 h->fenc->b_kept_as_ref =
3445 h->fdec->b_kept_as_ref = i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE && h->param.i_keyint_max > 1;
3447 h->fdec->mb_info = h->fenc->mb_info;
3448 h->fdec->mb_info_free = h->fenc->mb_info_free;
3449 h->fenc->mb_info = NULL;
3450 h->fenc->mb_info_free = NULL;
3452 h->fdec->i_pts = h->fenc->i_pts;
3453 if( h->frames.i_bframe_delay )
3455 int64_t *prev_reordered_pts = thread_current->frames.i_prev_reordered_pts;
3456 h->fdec->i_dts = h->i_frame > h->frames.i_bframe_delay
3457 ? prev_reordered_pts[ (h->i_frame - h->frames.i_bframe_delay) % h->frames.i_bframe_delay ]
3458 : h->fenc->i_reordered_pts - h->frames.i_bframe_delay_time;
3459 prev_reordered_pts[ h->i_frame % h->frames.i_bframe_delay ] = h->fenc->i_reordered_pts;
3462 h->fdec->i_dts = h->fenc->i_reordered_pts;
3463 if( h->fenc->i_type == X264_TYPE_IDR )
3464 h->i_last_idr_pts = h->fdec->i_pts;
3466 /* ------------------- Init ----------------------------- */
3467 /* build ref list 0/1 */
3468 x264_reference_build_list( h, h->fdec->i_poc );
3470 /* ---------------------- Write the bitstream -------------------------- */
3471 /* Init bitstream context */
3472 if( h->param.b_sliced_threads )
3474 for( int i = 0; i < h->param.i_threads; i++ )
3476 bs_init( &h->thread[i]->out.bs, h->thread[i]->out.p_bitstream, h->thread[i]->out.i_bitstream );
3477 h->thread[i]->out.i_nal = 0;
3482 bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
3486 if( h->param.b_aud )
3490 if( h->sh.i_type == SLICE_TYPE_I )
3492 else if( h->sh.i_type == SLICE_TYPE_P )
3494 else if( h->sh.i_type == SLICE_TYPE_B )
3499 x264_nal_start( h, NAL_AUD, NAL_PRIORITY_DISPOSABLE );
3500 bs_write( &h->out.bs, 3, pic_type );
3501 bs_rbsp_trailing( &h->out.bs );
3502 if( x264_nal_end( h ) )
3504 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3507 h->i_nal_type = i_nal_type;
3508 h->i_nal_ref_idc = i_nal_ref_idc;
3510 if( h->param.b_intra_refresh )
3512 if( IS_X264_TYPE_I( h->fenc->i_type ) )
3514 h->fdec->i_frames_since_pir = 0;
3515 h->b_queued_intra_refresh = 0;
3516 /* PIR is currently only supported with ref == 1, so any intra frame effectively refreshes
3517 * the whole frame and counts as an intra refresh. */
3518 h->fdec->f_pir_position = h->mb.i_mb_width;
3520 else if( h->fenc->i_type == X264_TYPE_P )
3522 int pocdiff = (h->fdec->i_poc - h->fref[0][0]->i_poc)/2;
3523 float increment = X264_MAX( ((float)h->mb.i_mb_width-1) / h->param.i_keyint_max, 1 );
3524 h->fdec->f_pir_position = h->fref[0][0]->f_pir_position;
3525 h->fdec->i_frames_since_pir = h->fref[0][0]->i_frames_since_pir + pocdiff;
3526 if( h->fdec->i_frames_since_pir >= h->param.i_keyint_max ||
3527 (h->b_queued_intra_refresh && h->fdec->f_pir_position + 0.5 >= h->mb.i_mb_width) )
3529 h->fdec->f_pir_position = 0;
3530 h->fdec->i_frames_since_pir = 0;
3531 h->b_queued_intra_refresh = 0;
3532 h->fenc->b_keyframe = 1;
3534 h->fdec->i_pir_start_col = h->fdec->f_pir_position+0.5;
3535 h->fdec->f_pir_position += increment * pocdiff;
3536 h->fdec->i_pir_end_col = h->fdec->f_pir_position+0.5;
3537 /* If our intra refresh has reached the right side of the frame, we're done. */
3538 if( h->fdec->i_pir_end_col >= h->mb.i_mb_width - 1 )
3540 h->fdec->f_pir_position = h->mb.i_mb_width;
3541 h->fdec->i_pir_end_col = h->mb.i_mb_width - 1;
3546 if( h->fenc->b_keyframe )
3548 /* Write SPS and PPS */
3549 if( h->param.b_repeat_headers )
3551 /* generate sequence parameters */
3552 x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
3553 x264_sps_write( &h->out.bs, h->sps );
3554 if( x264_nal_end( h ) )
3556 /* Pad AUD/SPS to 256 bytes like Panasonic */
3557 if( h->param.i_avcintra_class )
3558 h->out.nal[h->out.i_nal-1].i_padding = 256 - bs_pos( &h->out.bs ) / 8 - 2*NALU_OVERHEAD;
3559 overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + NALU_OVERHEAD;
3561 /* generate picture parameters */
3562 x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
3563 x264_pps_write( &h->out.bs, h->sps, h->pps );
3564 if( x264_nal_end( h ) )
3566 if( h->param.i_avcintra_class )
3567 h->out.nal[h->out.i_nal-1].i_padding = 256 - h->out.nal[h->out.i_nal-1].i_payload - NALU_OVERHEAD;
3568 overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + NALU_OVERHEAD;
3571 /* when frame threading is used, buffering period sei is written in x264_encoder_frame_end */
3572 if( h->i_thread_frames == 1 && h->sps->vui.b_nal_hrd_parameters_present )
3574 x264_hrd_fullness( h );
3575 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3576 x264_sei_buffering_period_write( h, &h->out.bs );
3577 if( x264_nal_end( h ) )
3579 overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3583 /* write extra sei */
3584 for( int i = 0; i < h->fenc->extra_sei.num_payloads; i++ )
3586 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3587 x264_sei_write( &h->out.bs, h->fenc->extra_sei.payloads[i].payload, h->fenc->extra_sei.payloads[i].payload_size,
3588 h->fenc->extra_sei.payloads[i].payload_type );
3589 if( x264_nal_end( h ) )
3591 overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3592 if( h->fenc->extra_sei.sei_free )
3594 h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads[i].payload );
3595 h->fenc->extra_sei.payloads[i].payload = NULL;
3599 if( h->fenc->extra_sei.sei_free )
3601 h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads );
3602 h->fenc->extra_sei.payloads = NULL;
3603 h->fenc->extra_sei.sei_free = NULL;
3606 if( h->fenc->b_keyframe )
3608 /* Avid's decoder strictly wants two SEIs for AVC-Intra so we can't insert the x264 SEI */
3609 if( h->param.b_repeat_headers && h->fenc->i_frame == 0 && !h->param.i_avcintra_class )
3611 /* identify ourself */
3612 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3613 if( x264_sei_version_write( h, &h->out.bs ) )
3615 if( x264_nal_end( h ) )
3617 overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3620 if( h->fenc->i_type != X264_TYPE_IDR )
3622 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;
3623 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3624 x264_sei_recovery_point_write( h, &h->out.bs, time_to_recovery );
3625 if( x264_nal_end( h ) )
3627 overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3631 if( h->param.i_frame_packing >= 0 && (h->fenc->b_keyframe || h->param.i_frame_packing == 5) )
3633 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3634 x264_sei_frame_packing_write( h, &h->out.bs );
3635 if( x264_nal_end( h ) )
3637 overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3640 /* generate sei pic timing */
3641 if( h->sps->vui.b_pic_struct_present || h->sps->vui.b_nal_hrd_parameters_present )
3643 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3644 x264_sei_pic_timing_write( h, &h->out.bs );
3645 if( x264_nal_end( h ) )
3647 overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3650 /* As required by Blu-ray. */
3651 if( !IS_X264_TYPE_B( h->fenc->i_type ) && h->b_sh_backup )
3654 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3655 x264_sei_dec_ref_pic_marking_write( h, &h->out.bs );
3656 if( x264_nal_end( h ) )
3658 overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3661 if( h->fenc->b_keyframe && h->param.b_intra_refresh )
3662 h->i_cpb_delay_pir_offset_next = h->fenc->i_cpb_delay;
3664 /* Filler space: 10 or 18 SEIs' worth of space, depending on resolution */
3665 if( h->param.i_avcintra_class )
3667 /* Write an empty filler NAL to mimic the AUD in the P2 format*/
3668 x264_nal_start( h, NAL_FILLER, NAL_PRIORITY_DISPOSABLE );
3669 x264_filler_write( h, &h->out.bs, 0 );
3670 if( x264_nal_end( h ) )
3672 overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3674 /* All lengths are magic lengths that decoders expect to see */
3676 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3677 if( x264_sei_avcintra_umid_write( h, &h->out.bs ) < 0 )
3679 if( x264_nal_end( h ) )
3681 overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3685 if( h->param.i_height == 1080 )
3687 unpadded_len = 5780;
3692 unpadded_len = 2900;
3696 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3697 if( x264_sei_avcintra_vanc_write( h, &h->out.bs, unpadded_len ) < 0 )
3699 if( x264_nal_end( h ) )
3702 h->out.nal[h->out.i_nal-1].i_padding = total_len - h->out.nal[h->out.i_nal-1].i_payload - SEI_OVERHEAD;
3703 overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + SEI_OVERHEAD;
3706 /* Init the rate control */
3707 /* FIXME: Include slice header bit cost. */
3708 x264_ratecontrol_start( h, h->fenc->i_qpplus1, overhead*8 );
3709 i_global_qp = x264_ratecontrol_qp( h );
3711 pic_out->i_qpplus1 =
3712 h->fdec->i_qpplus1 = i_global_qp + 1;
3714 if( h->param.rc.b_stat_read && h->sh.i_type != SLICE_TYPE_I )
3716 x264_reference_build_list_optimal( h );
3717 x264_reference_check_reorder( h );
3721 h->fdec->i_poc_l0ref0 = h->fref[0][0]->i_poc;
3723 /* ------------------------ Create slice header ----------------------- */
3724 x264_slice_init( h, i_nal_type, i_global_qp );
3726 /*------------------------- Weights -------------------------------------*/
3727 if( h->sh.i_type == SLICE_TYPE_B )
3728 x264_macroblock_bipred_init( h );
3730 x264_weighted_pred_init( h );
3732 if( i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE )
3736 h->i_threadslice_start = 0;
3737 h->i_threadslice_end = h->mb.i_mb_height;
3738 if( h->i_thread_frames > 1 )
3740 x264_threadpool_run( h->threadpool, (void*)x264_slices_write, h );
3741 h->b_thread_active = 1;
3743 else if( h->param.b_sliced_threads )
3745 if( x264_threaded_slices_write( h ) )
3749 if( (intptr_t)x264_slices_write( h ) )
3752 return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
3755 static int x264_encoder_frame_end( x264_t *h, x264_t *thread_current,
3756 x264_nal_t **pp_nal, int *pi_nal,
3757 x264_picture_t *pic_out )
3759 char psz_message[80];
3761 if( !h->param.b_sliced_threads && h->b_thread_active )
3763 h->b_thread_active = 0;
3764 if( (intptr_t)x264_threadpool_wait( h->threadpool, h ) )
3769 pic_out->i_type = X264_TYPE_AUTO;
3775 /* generate buffering period sei and insert it into place */
3776 if( h->i_thread_frames > 1 && h->fenc->b_keyframe && h->sps->vui.b_nal_hrd_parameters_present )
3778 x264_hrd_fullness( h );
3779 x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3780 x264_sei_buffering_period_write( h, &h->out.bs );
3781 if( x264_nal_end( h ) )
3783 /* buffering period sei must follow AUD, SPS and PPS and precede all other SEIs */
3785 while( h->out.nal[idx].i_type == NAL_AUD ||
3786 h->out.nal[idx].i_type == NAL_SPS ||
3787 h->out.nal[idx].i_type == NAL_PPS )
3789 x264_nal_t nal_tmp = h->out.nal[h->out.i_nal-1];
3790 memmove( &h->out.nal[idx+1], &h->out.nal[idx], (h->out.i_nal-idx-1)*sizeof(x264_nal_t) );
3791 h->out.nal[idx] = nal_tmp;
3794 int frame_size = x264_encoder_encapsulate_nals( h, 0 );
3795 if( frame_size < 0 )
3798 /* Set output picture properties */
3799 pic_out->i_type = h->fenc->i_type;
3801 pic_out->b_keyframe = h->fenc->b_keyframe;
3802 pic_out->i_pic_struct = h->fenc->i_pic_struct;
3804 pic_out->i_pts = h->fdec->i_pts;
3805 pic_out->i_dts = h->fdec->i_dts;
3807 if( pic_out->i_pts < pic_out->i_dts )
3808 x264_log( h, X264_LOG_WARNING, "invalid DTS: PTS is less than DTS\n" );
3810 pic_out->opaque = h->fenc->opaque;
3812 pic_out->img.i_csp = h->fdec->i_csp;
3814 pic_out->img.i_csp |= X264_CSP_HIGH_DEPTH;
3816 pic_out->img.i_plane = h->fdec->i_plane;
3817 for( int i = 0; i < pic_out->img.i_plane; i++ )
3819 pic_out->img.i_stride[i] = h->fdec->i_stride[i] * sizeof(pixel);
3820 pic_out->img.plane[i] = (uint8_t*)h->fdec->plane[i];
3823 x264_frame_push_unused( thread_current, h->fenc );
3825 /* ---------------------- Update encoder state ------------------------- */
3829 if( x264_ratecontrol_end( h, frame_size * 8, &filler ) < 0 )
3832 pic_out->hrd_timing = h->fenc->hrd_timing;
3833 pic_out->prop.f_crf_avg = h->fdec->f_crf_avg;
3835 /* Filler in AVC-Intra mode is written as zero bytes to the last slice
3836 * We don't know the size of the last slice until encapsulation so we add filler to the encapsulated NAL */
3837 if( h->param.i_avcintra_class )
3839 x264_t *h0 = h->thread[0];
3840 int ret = x264_check_encapsulated_buffer( h, h0, h->out.i_nal, frame_size, frame_size + filler );
3843 memset( h->out.nal[0].p_payload + frame_size, 0, filler );
3844 h->out.nal[h->out.i_nal-1].i_payload += filler;
3845 h->out.nal[h->out.i_nal-1].i_padding = filler;
3846 frame_size += filler;
3853 overhead = (FILLER_OVERHEAD - h->param.b_annexb);
3854 if( h->param.i_slice_max_size && filler > h->param.i_slice_max_size )
3856 int next_size = filler - h->param.i_slice_max_size;
3857 int overflow = X264_MAX( overhead - next_size, 0 );
3858 f = h->param.i_slice_max_size - overhead - overflow;
3861 f = X264_MAX( 0, filler - overhead );
3863 if( x264_bitstream_check_buffer_filler( h, f ) )
3865 x264_nal_start( h, NAL_FILLER, NAL_PRIORITY_DISPOSABLE );
3866 x264_filler_write( h, &h->out.bs, f );
3867 if( x264_nal_end( h ) )
3869 int total_size = x264_encoder_encapsulate_nals( h, h->out.i_nal-1 );
3870 if( total_size < 0 )
3872 frame_size += total_size;
3873 filler -= total_size;
3877 /* End bitstream, set output */
3878 *pi_nal = h->out.i_nal;
3879 *pp_nal = h->out.nal;
3883 x264_noise_reduction_update( h );
3885 /* ---------------------- Compute/Print statistics --------------------- */
3886 x264_thread_sync_stat( h, h->thread[0] );
3889 h->stat.i_frame_count[h->sh.i_type]++;
3890 h->stat.i_frame_size[h->sh.i_type] += frame_size;
3891 h->stat.f_frame_qp[h->sh.i_type] += h->fdec->f_qp_avg_aq;
3893 for( int i = 0; i < X264_MBTYPE_MAX; i++ )
3894 h->stat.i_mb_count[h->sh.i_type][i] += h->stat.frame.i_mb_count[i];
3895 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
3896 h->stat.i_mb_partition[h->sh.i_type][i] += h->stat.frame.i_mb_partition[i];
3897 for( int i = 0; i < 2; i++ )
3898 h->stat.i_mb_count_8x8dct[i] += h->stat.frame.i_mb_count_8x8dct[i];
3899 for( int i = 0; i < 6; i++ )
3900 h->stat.i_mb_cbp[i] += h->stat.frame.i_mb_cbp[i];
3901 for( int i = 0; i < 4; i++ )
3902 for( int j = 0; j < 13; j++ )
3903 h->stat.i_mb_pred_mode[i][j] += h->stat.frame.i_mb_pred_mode[i][j];
3904 if( h->sh.i_type != SLICE_TYPE_I )
3905 for( int i_list = 0; i_list < 2; i_list++ )
3906 for( int i = 0; i < X264_REF_MAX*2; i++ )
3907 h->stat.i_mb_count_ref[h->sh.i_type][i_list][i] += h->stat.frame.i_mb_count_ref[i_list][i];
3908 for( int i = 0; i < 3; i++ )
3909 h->stat.i_mb_field[i] += h->stat.frame.i_mb_field[i];
3910 if( h->sh.i_type == SLICE_TYPE_P && h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE )
3912 h->stat.i_wpred[0] += !!h->sh.weight[0][0].weightfn;
3913 h->stat.i_wpred[1] += !!h->sh.weight[0][1].weightfn || !!h->sh.weight[0][2].weightfn;
3915 if( h->sh.i_type == SLICE_TYPE_B )
3917 h->stat.i_direct_frames[ h->sh.b_direct_spatial_mv_pred ] ++;
3918 if( h->mb.b_direct_auto_write )
3920 //FIXME somewhat arbitrary time constants
3921 if( h->stat.i_direct_score[0] + h->stat.i_direct_score[1] > h->mb.i_mb_count )
3922 for( int i = 0; i < 2; i++ )
3923 h->stat.i_direct_score[i] = h->stat.i_direct_score[i] * 9/10;
3924 for( int i = 0; i < 2; i++ )
3925 h->stat.i_direct_score[i] += h->stat.frame.i_direct_score[i];
3929 h->stat.i_consecutive_bframes[h->fenc->i_bframes]++;
3931 psz_message[0] = '\0';
3932 double dur = h->fenc->f_duration;
3933 h->stat.f_frame_duration[h->sh.i_type] += dur;
3934 if( h->param.analyse.b_psnr )
3938 h->stat.frame.i_ssd[0],
3939 h->stat.frame.i_ssd[1],
3940 h->stat.frame.i_ssd[2],
3942 int luma_size = h->param.i_width * h->param.i_height;
3943 int chroma_size = CHROMA_SIZE( luma_size );
3944 pic_out->prop.f_psnr[0] = x264_psnr( ssd[0], luma_size );
3945 pic_out->prop.f_psnr[1] = x264_psnr( ssd[1], chroma_size );
3946 pic_out->prop.f_psnr[2] = x264_psnr( ssd[2], chroma_size );
3947 pic_out->prop.f_psnr_avg = x264_psnr( ssd[0] + ssd[1] + ssd[2], luma_size + chroma_size*2 );
3949 h->stat.f_ssd_global[h->sh.i_type] += dur * (ssd[0] + ssd[1] + ssd[2]);
3950 h->stat.f_psnr_average[h->sh.i_type] += dur * pic_out->prop.f_psnr_avg;
3951 h->stat.f_psnr_mean_y[h->sh.i_type] += dur * pic_out->prop.f_psnr[0];
3952 h->stat.f_psnr_mean_u[h->sh.i_type] += dur * pic_out->prop.f_psnr[1];
3953 h->stat.f_psnr_mean_v[h->sh.i_type] += dur * pic_out->prop.f_psnr[2];
3955 snprintf( psz_message, 80, " PSNR Y:%5.2f U:%5.2f V:%5.2f", pic_out->prop.f_psnr[0],
3956 pic_out->prop.f_psnr[1],
3957 pic_out->prop.f_psnr[2] );
3960 if( h->param.analyse.b_ssim )
3962 pic_out->prop.f_ssim = h->stat.frame.f_ssim / h->stat.frame.i_ssim_cnt;
3963 h->stat.f_ssim_mean_y[h->sh.i_type] += pic_out->prop.f_ssim * dur;
3964 snprintf( psz_message + strlen(psz_message), 80 - strlen(psz_message),
3965 " SSIM Y:%.5f", pic_out->prop.f_ssim );
3967 psz_message[79] = '\0';
3969 x264_log( h, X264_LOG_DEBUG,
3970 "frame=%4d QP=%.2f NAL=%d Slice:%c Poc:%-3d I:%-4d P:%-4d SKIP:%-4d size=%d bytes%s\n",
3972 h->fdec->f_qp_avg_aq,
3974 h->sh.i_type == SLICE_TYPE_I ? 'I' : (h->sh.i_type == SLICE_TYPE_P ? 'P' : 'B' ),
3976 h->stat.frame.i_mb_count_i,
3977 h->stat.frame.i_mb_count_p,
3978 h->stat.frame.i_mb_count_skip,
3982 // keep stats all in one place
3983 x264_thread_sync_stat( h->thread[0], h );
3984 // for the use of the next frame
3985 x264_thread_sync_stat( thread_current, h );
3987 #ifdef DEBUG_MB_TYPE
3989 static const char mb_chars[] = { 'i', 'i', 'I', 'C', 'P', '8', 'S',
3990 'D', '<', 'X', 'B', 'X', '>', 'B', 'B', 'B', 'B', '8', 'S' };
3991 for( int mb_xy = 0; mb_xy < h->mb.i_mb_width * h->mb.i_mb_height; mb_xy++ )
3993 if( h->mb.type[mb_xy] < X264_MBTYPE_MAX && h->mb.type[mb_xy] >= 0 )
3994 fprintf( stderr, "%c ", mb_chars[ h->mb.type[mb_xy] ] );
3996 fprintf( stderr, "? " );
3998 if( (mb_xy+1) % h->mb.i_mb_width == 0 )
3999 fprintf( stderr, "\n" );
4004 /* Remove duplicates, must be done near the end as breaks h->fref0 array
4005 * by freeing some of its pointers. */
4006 for( int i = 0; i < h->i_ref[0]; i++ )
4007 if( h->fref[0][i] && h->fref[0][i]->b_duplicate )
4009 x264_frame_push_blank_unused( h, h->fref[0][i] );
4013 if( h->param.psz_dump_yuv )
4014 x264_frame_dump( h );
4020 static void x264_print_intra( int64_t *i_mb_count, double i_count, int b_print_pcm, char *intra )
4022 intra += sprintf( intra, "I16..4%s: %4.1f%% %4.1f%% %4.1f%%",
4023 b_print_pcm ? "..PCM" : "",
4024 i_mb_count[I_16x16]/ i_count,
4025 i_mb_count[I_8x8] / i_count,
4026 i_mb_count[I_4x4] / i_count );
4028 sprintf( intra, " %4.1f%%", i_mb_count[I_PCM] / i_count );
4031 /****************************************************************************
4032 * x264_encoder_close:
4033 ****************************************************************************/
4034 void x264_encoder_close ( x264_t *h )
4036 int64_t i_yuv_size = FRAME_SIZE( h->param.i_width * h->param.i_height );
4037 int64_t i_mb_count_size[2][7] = {{0}};
4039 int b_print_pcm = h->stat.i_mb_count[SLICE_TYPE_I][I_PCM]
4040 || h->stat.i_mb_count[SLICE_TYPE_P][I_PCM]
4041 || h->stat.i_mb_count[SLICE_TYPE_B][I_PCM];
4043 x264_lookahead_delete( h );
4046 x264_opencl_lookahead_delete( h );
4047 x264_opencl_function_t *ocl = h->opencl.ocl;
4050 if( h->param.b_sliced_threads )
4051 x264_threadpool_wait_all( h );
4052 if( h->param.i_threads > 1 )
4053 x264_threadpool_delete( h->threadpool );
4054 if( h->param.i_lookahead_threads > 1 )
4055 x264_threadpool_delete( h->lookaheadpool );
4056 if( h->i_thread_frames > 1 )
4058 for( int i = 0; i < h->i_thread_frames; i++ )
4059 if( h->thread[i]->b_thread_active )
4061 assert( h->thread[i]->fenc->i_reference_count == 1 );
4062 x264_frame_delete( h->thread[i]->fenc );
4065 x264_t *thread_prev = h->thread[h->i_thread_phase];
4066 x264_thread_sync_ratecontrol( h, thread_prev, h );
4067 x264_thread_sync_ratecontrol( thread_prev, thread_prev, h );
4068 h->i_frame = thread_prev->i_frame + 1 - h->i_thread_frames;
4072 /* Slices used and PSNR */
4073 for( int i = 0; i < 3; i++ )
4075 static const uint8_t slice_order[] = { SLICE_TYPE_I, SLICE_TYPE_P, SLICE_TYPE_B };
4076 int i_slice = slice_order[i];
4078 if( h->stat.i_frame_count[i_slice] > 0 )
4080 int i_count = h->stat.i_frame_count[i_slice];
4081 double dur = h->stat.f_frame_duration[i_slice];
4082 if( h->param.analyse.b_psnr )
4084 x264_log( h, X264_LOG_INFO,
4085 "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",
4086 slice_type_to_char[i_slice],
4088 h->stat.f_frame_qp[i_slice] / i_count,
4089 (double)h->stat.i_frame_size[i_slice] / i_count,
4090 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,
4091 h->stat.f_psnr_average[i_slice] / dur,
4092 x264_psnr( h->stat.f_ssd_global[i_slice], dur * i_yuv_size ) );
4096 x264_log( h, X264_LOG_INFO,
4097 "frame %c:%-5d Avg QP:%5.2f size:%6.0f\n",
4098 slice_type_to_char[i_slice],
4100 h->stat.f_frame_qp[i_slice] / i_count,
4101 (double)h->stat.i_frame_size[i_slice] / i_count );
4105 if( h->param.i_bframe && h->stat.i_frame_count[SLICE_TYPE_B] )
4109 // weight by number of frames (including the I/P-frames) that are in a sequence of N B-frames
4110 for( int i = 0; i <= h->param.i_bframe; i++ )
4111 den += (i+1) * h->stat.i_consecutive_bframes[i];
4112 for( int i = 0; i <= h->param.i_bframe; i++ )
4113 p += sprintf( p, " %4.1f%%", 100. * (i+1) * h->stat.i_consecutive_bframes[i] / den );
4114 x264_log( h, X264_LOG_INFO, "consecutive B-frames:%s\n", buf );
4117 for( int i_type = 0; i_type < 2; i_type++ )
4118 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
4120 if( i == D_DIRECT_8x8 ) continue; /* direct is counted as its own type */
4121 i_mb_count_size[i_type][x264_mb_partition_pixel_table[i]] += h->stat.i_mb_partition[i_type][i];
4125 if( h->stat.i_frame_count[SLICE_TYPE_I] > 0 )
4127 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_I];
4128 double i_count = (double)h->stat.i_frame_count[SLICE_TYPE_I] * h->mb.i_mb_count / 100.0;
4129 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
4130 x264_log( h, X264_LOG_INFO, "mb I %s\n", buf );
4132 if( h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
4134 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_P];
4135 double i_count = (double)h->stat.i_frame_count[SLICE_TYPE_P] * h->mb.i_mb_count / 100.0;
4136 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_P];
4137 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
4138 x264_log( h, X264_LOG_INFO,
4139 "mb P %s P16..4: %4.1f%% %4.1f%% %4.1f%% %4.1f%% %4.1f%% skip:%4.1f%%\n",
4141 i_mb_size[PIXEL_16x16] / (i_count*4),
4142 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
4143 i_mb_size[PIXEL_8x8] / (i_count*4),
4144 (i_mb_size[PIXEL_8x4] + i_mb_size[PIXEL_4x8]) / (i_count*4),
4145 i_mb_size[PIXEL_4x4] / (i_count*4),
4146 i_mb_count[P_SKIP] / i_count );
4148 if( h->stat.i_frame_count[SLICE_TYPE_B] > 0 )
4150 int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_B];
4151 double i_count = (double)h->stat.i_frame_count[SLICE_TYPE_B] * h->mb.i_mb_count / 100.0;
4152 double i_mb_list_count;
4153 int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_B];
4154 int64_t list_count[3] = {0}; /* 0 == L0, 1 == L1, 2 == BI */
4155 x264_print_intra( i_mb_count, i_count, b_print_pcm, buf );
4156 for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
4157 for( int j = 0; j < 2; j++ )
4159 int l0 = x264_mb_type_list_table[i][0][j];
4160 int l1 = x264_mb_type_list_table[i][1][j];
4162 list_count[l1+l0*l1] += h->stat.i_mb_count[SLICE_TYPE_B][i] * 2;
4164 list_count[0] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L0_8x8];
4165 list_count[1] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L1_8x8];
4166 list_count[2] += h->stat.i_mb_partition[SLICE_TYPE_B][D_BI_8x8];
4167 i_mb_count[B_DIRECT] += (h->stat.i_mb_partition[SLICE_TYPE_B][D_DIRECT_8x8]+2)/4;
4168 i_mb_list_count = (list_count[0] + list_count[1] + list_count[2]) / 100.0;
4169 sprintf( buf + strlen(buf), " B16..8: %4.1f%% %4.1f%% %4.1f%% direct:%4.1f%% skip:%4.1f%%",
4170 i_mb_size[PIXEL_16x16] / (i_count*4),
4171 (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
4172 i_mb_size[PIXEL_8x8] / (i_count*4),
4173 i_mb_count[B_DIRECT] / i_count,
4174 i_mb_count[B_SKIP] / i_count );
4175 if( i_mb_list_count != 0 )
4176 sprintf( buf + strlen(buf), " L0:%4.1f%% L1:%4.1f%% BI:%4.1f%%",
4177 list_count[0] / i_mb_list_count,
4178 list_count[1] / i_mb_list_count,
4179 list_count[2] / i_mb_list_count );
4180 x264_log( h, X264_LOG_INFO, "mb B %s\n", buf );
4183 x264_ratecontrol_summary( h );
4185 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 )
4187 #define SUM3(p) (p[SLICE_TYPE_I] + p[SLICE_TYPE_P] + p[SLICE_TYPE_B])
4188 #define SUM3b(p,o) (p[SLICE_TYPE_I][o] + p[SLICE_TYPE_P][o] + p[SLICE_TYPE_B][o])
4189 int64_t i_i8x8 = SUM3b( h->stat.i_mb_count, I_8x8 );
4190 int64_t i_intra = i_i8x8 + SUM3b( h->stat.i_mb_count, I_4x4 )
4191 + SUM3b( h->stat.i_mb_count, I_16x16 );
4192 int64_t i_all_intra = i_intra + SUM3b( h->stat.i_mb_count, I_PCM);
4193 int64_t i_skip = SUM3b( h->stat.i_mb_count, P_SKIP )
4194 + SUM3b( h->stat.i_mb_count, B_SKIP );
4195 const int i_count = h->stat.i_frame_count[SLICE_TYPE_I] +
4196 h->stat.i_frame_count[SLICE_TYPE_P] +
4197 h->stat.i_frame_count[SLICE_TYPE_B];
4198 int64_t i_mb_count = (int64_t)i_count * h->mb.i_mb_count;
4199 int64_t i_inter = i_mb_count - i_skip - i_intra;
4200 const double duration = h->stat.f_frame_duration[SLICE_TYPE_I] +
4201 h->stat.f_frame_duration[SLICE_TYPE_P] +
4202 h->stat.f_frame_duration[SLICE_TYPE_B];
4203 float f_bitrate = SUM3(h->stat.i_frame_size) / duration / 125;
4205 if( PARAM_INTERLACED )
4207 char *fieldstats = buf;
4210 fieldstats += sprintf( fieldstats, " inter:%.1f%%", h->stat.i_mb_field[1] * 100.0 / i_inter );
4212 fieldstats += sprintf( fieldstats, " skip:%.1f%%", h->stat.i_mb_field[2] * 100.0 / i_skip );
4213 x264_log( h, X264_LOG_INFO, "field mbs: intra: %.1f%%%s\n",
4214 h->stat.i_mb_field[0] * 100.0 / i_intra, buf );
4217 if( h->pps->b_transform_8x8_mode )
4220 if( h->stat.i_mb_count_8x8dct[0] )
4221 sprintf( buf, " inter:%.1f%%", 100. * h->stat.i_mb_count_8x8dct[1] / h->stat.i_mb_count_8x8dct[0] );
4222 x264_log( h, X264_LOG_INFO, "8x8 transform intra:%.1f%%%s\n", 100. * i_i8x8 / i_intra, buf );
4225 if( (h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO ||
4226 (h->stat.i_direct_frames[0] && h->stat.i_direct_frames[1]))
4227 && h->stat.i_frame_count[SLICE_TYPE_B] )
4229 x264_log( h, X264_LOG_INFO, "direct mvs spatial:%.1f%% temporal:%.1f%%\n",
4230 h->stat.i_direct_frames[1] * 100. / h->stat.i_frame_count[SLICE_TYPE_B],
4231 h->stat.i_direct_frames[0] * 100. / h->stat.i_frame_count[SLICE_TYPE_B] );
4235 int csize = CHROMA444 ? 4 : 1;
4236 if( i_mb_count != i_all_intra )
4237 sprintf( buf, " inter: %.1f%% %.1f%% %.1f%%",
4238 h->stat.i_mb_cbp[1] * 100.0 / ((i_mb_count - i_all_intra)*4),
4239 h->stat.i_mb_cbp[3] * 100.0 / ((i_mb_count - i_all_intra)*csize),
4240 h->stat.i_mb_cbp[5] * 100.0 / ((i_mb_count - i_all_intra)*csize) );
4241 x264_log( h, X264_LOG_INFO, "coded y,%s,%s intra: %.1f%% %.1f%% %.1f%%%s\n",
4242 CHROMA444?"u":"uvDC", CHROMA444?"v":"uvAC",
4243 h->stat.i_mb_cbp[0] * 100.0 / (i_all_intra*4),
4244 h->stat.i_mb_cbp[2] * 100.0 / (i_all_intra*csize),
4245 h->stat.i_mb_cbp[4] * 100.0 / (i_all_intra*csize), buf );
4247 int64_t fixed_pred_modes[4][9] = {{0}};
4248 int64_t sum_pred_modes[4] = {0};
4249 for( int i = 0; i <= I_PRED_16x16_DC_128; i++ )
4251 fixed_pred_modes[0][x264_mb_pred_mode16x16_fix[i]] += h->stat.i_mb_pred_mode[0][i];
4252 sum_pred_modes[0] += h->stat.i_mb_pred_mode[0][i];
4254 if( sum_pred_modes[0] )
4255 x264_log( h, X264_LOG_INFO, "i16 v,h,dc,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
4256 fixed_pred_modes[0][0] * 100.0 / sum_pred_modes[0],
4257 fixed_pred_modes[0][1] * 100.0 / sum_pred_modes[0],
4258 fixed_pred_modes[0][2] * 100.0 / sum_pred_modes[0],
4259 fixed_pred_modes[0][3] * 100.0 / sum_pred_modes[0] );
4260 for( int i = 1; i <= 2; i++ )
4262 for( int j = 0; j <= I_PRED_8x8_DC_128; j++ )
4264 fixed_pred_modes[i][x264_mb_pred_mode4x4_fix(j)] += h->stat.i_mb_pred_mode[i][j];
4265 sum_pred_modes[i] += h->stat.i_mb_pred_mode[i][j];
4267 if( sum_pred_modes[i] )
4268 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,
4269 fixed_pred_modes[i][0] * 100.0 / sum_pred_modes[i],
4270 fixed_pred_modes[i][1] * 100.0 / sum_pred_modes[i],
4271 fixed_pred_modes[i][2] * 100.0 / sum_pred_modes[i],
4272 fixed_pred_modes[i][3] * 100.0 / sum_pred_modes[i],
4273 fixed_pred_modes[i][4] * 100.0 / sum_pred_modes[i],
4274 fixed_pred_modes[i][5] * 100.0 / sum_pred_modes[i],
4275 fixed_pred_modes[i][6] * 100.0 / sum_pred_modes[i],
4276 fixed_pred_modes[i][7] * 100.0 / sum_pred_modes[i],
4277 fixed_pred_modes[i][8] * 100.0 / sum_pred_modes[i] );
4279 for( int i = 0; i <= I_PRED_CHROMA_DC_128; i++ )
4281 fixed_pred_modes[3][x264_mb_chroma_pred_mode_fix[i]] += h->stat.i_mb_pred_mode[3][i];
4282 sum_pred_modes[3] += h->stat.i_mb_pred_mode[3][i];
4284 if( sum_pred_modes[3] && !CHROMA444 )
4285 x264_log( h, X264_LOG_INFO, "i8c dc,h,v,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
4286 fixed_pred_modes[3][0] * 100.0 / sum_pred_modes[3],
4287 fixed_pred_modes[3][1] * 100.0 / sum_pred_modes[3],
4288 fixed_pred_modes[3][2] * 100.0 / sum_pred_modes[3],
4289 fixed_pred_modes[3][3] * 100.0 / sum_pred_modes[3] );
4291 if( h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE && h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
4292 x264_log( h, X264_LOG_INFO, "Weighted P-Frames: Y:%.1f%% UV:%.1f%%\n",
4293 h->stat.i_wpred[0] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P],
4294 h->stat.i_wpred[1] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P] );
4296 for( int i_list = 0; i_list < 2; i_list++ )
4297 for( int i_slice = 0; i_slice < 2; i_slice++ )
4302 for( int i = 0; i < X264_REF_MAX*2; i++ )
4303 if( h->stat.i_mb_count_ref[i_slice][i_list][i] )
4305 i_den += h->stat.i_mb_count_ref[i_slice][i_list][i];
4310 for( int i = 0; i <= i_max; i++ )
4311 p += sprintf( p, " %4.1f%%", 100. * h->stat.i_mb_count_ref[i_slice][i_list][i] / i_den );
4312 x264_log( h, X264_LOG_INFO, "ref %c L%d:%s\n", "PB"[i_slice], i_list, buf );
4315 if( h->param.analyse.b_ssim )
4317 float ssim = SUM3( h->stat.f_ssim_mean_y ) / duration;
4318 x264_log( h, X264_LOG_INFO, "SSIM Mean Y:%.7f (%6.3fdb)\n", ssim, x264_ssim( ssim ) );
4320 if( h->param.analyse.b_psnr )
4322 x264_log( h, X264_LOG_INFO,
4323 "PSNR Mean Y:%6.3f U:%6.3f V:%6.3f Avg:%6.3f Global:%6.3f kb/s:%.2f\n",
4324 SUM3( h->stat.f_psnr_mean_y ) / duration,
4325 SUM3( h->stat.f_psnr_mean_u ) / duration,
4326 SUM3( h->stat.f_psnr_mean_v ) / duration,
4327 SUM3( h->stat.f_psnr_average ) / duration,
4328 x264_psnr( SUM3( h->stat.f_ssd_global ), duration * i_yuv_size ),
4332 x264_log( h, X264_LOG_INFO, "kb/s:%.2f\n", f_bitrate );
4336 x264_ratecontrol_delete( h );
4339 if( h->param.rc.psz_stat_out )
4340 free( h->param.rc.psz_stat_out );
4341 if( h->param.rc.psz_stat_in )
4342 free( h->param.rc.psz_stat_in );
4344 x264_cqm_delete( h );
4345 x264_free( h->nal_buffer );
4346 x264_free( h->reconfig_h );
4347 x264_analyse_free_costs( h );
4349 if( h->i_thread_frames > 1 )
4350 h = h->thread[h->i_thread_phase];
4353 x264_frame_delete_list( h->frames.unused[0] );
4354 x264_frame_delete_list( h->frames.unused[1] );
4355 x264_frame_delete_list( h->frames.current );
4356 x264_frame_delete_list( h->frames.blank_unused );
4360 for( int i = 0; i < h->i_thread_frames; i++ )
4361 if( h->thread[i]->b_thread_active )
4362 for( int j = 0; j < h->thread[i]->i_ref[0]; j++ )
4363 if( h->thread[i]->fref[0][j] && h->thread[i]->fref[0][j]->b_duplicate )
4364 x264_frame_delete( h->thread[i]->fref[0][j] );
4366 if( h->param.i_lookahead_threads > 1 )
4367 for( int i = 0; i < h->param.i_lookahead_threads; i++ )
4368 x264_free( h->lookahead_thread[i] );
4370 for( int i = h->param.i_threads - 1; i >= 0; i-- )
4372 x264_frame_t **frame;
4374 if( !h->param.b_sliced_threads || i == 0 )
4376 for( frame = h->thread[i]->frames.reference; *frame; frame++ )
4378 assert( (*frame)->i_reference_count > 0 );
4379 (*frame)->i_reference_count--;
4380 if( (*frame)->i_reference_count == 0 )
4381 x264_frame_delete( *frame );
4383 frame = &h->thread[i]->fdec;
4386 assert( (*frame)->i_reference_count > 0 );
4387 (*frame)->i_reference_count--;
4388 if( (*frame)->i_reference_count == 0 )
4389 x264_frame_delete( *frame );
4391 x264_macroblock_cache_free( h->thread[i] );
4393 x264_macroblock_thread_free( h->thread[i], 0 );
4394 x264_free( h->thread[i]->out.p_bitstream );
4395 x264_free( h->thread[i]->out.nal );
4396 x264_pthread_mutex_destroy( &h->thread[i]->mutex );
4397 x264_pthread_cond_destroy( &h->thread[i]->cv );
4398 x264_free( h->thread[i] );
4401 x264_opencl_close_library( ocl );
4405 int x264_encoder_delayed_frames( x264_t *h )
4407 int delayed_frames = 0;
4408 if( h->i_thread_frames > 1 )
4410 for( int i = 0; i < h->i_thread_frames; i++ )
4411 delayed_frames += h->thread[i]->b_thread_active;
4412 h = h->thread[h->i_thread_phase];
4414 for( int i = 0; h->frames.current[i]; i++ )
4416 x264_pthread_mutex_lock( &h->lookahead->ofbuf.mutex );
4417 x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
4418 x264_pthread_mutex_lock( &h->lookahead->next.mutex );
4419 delayed_frames += h->lookahead->ifbuf.i_size + h->lookahead->next.i_size + h->lookahead->ofbuf.i_size;
4420 x264_pthread_mutex_unlock( &h->lookahead->next.mutex );
4421 x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
4422 x264_pthread_mutex_unlock( &h->lookahead->ofbuf.mutex );
4423 return delayed_frames;
4426 int x264_encoder_maximum_delayed_frames( x264_t *h )
4428 return h->frames.i_delay;