1 /*****************************************************************************
2  * encoder.c: top-level encoder functions
3  *****************************************************************************
4  * Copyright (C) 2003-2021 x264 project
5  *
6  * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7  *          Loren Merritt <lorenm@u.washington.edu>
8  *          Fiona Glaser <fiona@x264.com>
9  *
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.
14  *
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.
19  *
20  * You should have received a copy of the GNU General Public License
21  * along with this program; if not, write to the Free Software
22  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
23  *
24  * This program is also available under a commercial proprietary license.
25  * For more information, contact us at licensing@x264.com.
26  *****************************************************************************/
27 
28 #include "common/common.h"
29 
30 #include "set.h"
31 #include "analyse.h"
32 #include "ratecontrol.h"
33 #include "macroblock.h"
34 #include "me.h"
35 #if HAVE_INTEL_DISPATCHER
36 #include "extras/intel_dispatcher.h"
37 #endif
38 
39 //#define DEBUG_MB_TYPE
40 
41 #define bs_write_ue bs_write_ue_big
42 
43 // forward declaration needed for template usage
44 void x264_nal_encode( x264_t *h, uint8_t *dst, x264_nal_t *nal );
45 void x264_macroblock_cache_load_progressive( x264_t *h, int i_mb_x, int i_mb_y );
46 
47 static int encoder_frame_end( x264_t *h, x264_t *thread_current,
48                               x264_nal_t **pp_nal, int *pi_nal,
49                               x264_picture_t *pic_out );
50 
51 /****************************************************************************
52  *
53  ******************************* x264 libs **********************************
54  *
55  ****************************************************************************/
calc_psnr(double sqe,double size)56 static double calc_psnr( double sqe, double size )
57 {
58     double mse = sqe / (PIXEL_MAX*PIXEL_MAX * size);
59     if( mse <= 0.0000000001 ) /* Max 100dB */
60         return 100;
61 
62     return -10.0 * log10( mse );
63 }
64 
calc_ssim_db(double ssim)65 static double calc_ssim_db( double ssim )
66 {
67     double inv_ssim = 1 - ssim;
68     if( inv_ssim <= 0.0000000001 ) /* Max 100dB */
69         return 100;
70 
71     return -10.0 * log10( inv_ssim );
72 }
73 
threadpool_wait_all(x264_t * h)74 static int threadpool_wait_all( x264_t *h )
75 {
76     for( int i = 0; i < h->param.i_threads; i++ )
77         if( h->thread[i]->b_thread_active )
78         {
79             h->thread[i]->b_thread_active = 0;
80             if( (intptr_t)x264_threadpool_wait( h->threadpool, h->thread[i] ) < 0 )
81                 return -1;
82         }
83     return 0;
84 }
85 
frame_dump(x264_t * h)86 static void frame_dump( x264_t *h )
87 {
88     FILE *f = x264_fopen( h->param.psz_dump_yuv, "r+b" );
89     if( !f )
90         return;
91 
92     /* Wait for the threads to finish deblocking */
93     if( h->param.b_sliced_threads )
94         threadpool_wait_all( h );
95 
96     /* Write the frame in display order */
97     int frame_size = FRAME_SIZE( h->param.i_height * h->param.i_width * SIZEOF_PIXEL );
98     if( !fseek( f, (int64_t)h->fdec->i_frame * frame_size, SEEK_SET ) )
99     {
100         for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
101             for( int y = 0; y < h->param.i_height; y++ )
102                 fwrite( &h->fdec->plane[p][y*h->fdec->i_stride[p]], SIZEOF_PIXEL, h->param.i_width, f );
103         if( CHROMA_FORMAT == CHROMA_420 || CHROMA_FORMAT == CHROMA_422 )
104         {
105             int cw = h->param.i_width>>1;
106             int ch = h->param.i_height>>CHROMA_V_SHIFT;
107             pixel *planeu = x264_malloc( 2 * (cw*ch*SIZEOF_PIXEL + 32) );
108             if( planeu )
109             {
110                 pixel *planev = planeu + cw*ch + 32/SIZEOF_PIXEL;
111                 h->mc.plane_copy_deinterleave( planeu, cw, planev, cw, h->fdec->plane[1], h->fdec->i_stride[1], cw, ch );
112                 fwrite( planeu, 1, cw*ch*SIZEOF_PIXEL, f );
113                 fwrite( planev, 1, cw*ch*SIZEOF_PIXEL, f );
114                 x264_free( planeu );
115             }
116         }
117     }
118     fclose( f );
119 }
120 
121 /* Fill "default" values */
slice_header_init(x264_t * h,x264_slice_header_t * sh,x264_sps_t * sps,x264_pps_t * pps,int i_idr_pic_id,int i_frame,int i_qp)122 static void slice_header_init( x264_t *h, x264_slice_header_t *sh,
123                                x264_sps_t *sps, x264_pps_t *pps,
124                                int i_idr_pic_id, int i_frame, int i_qp )
125 {
126     x264_param_t *param = &h->param;
127 
128     /* First we fill all fields */
129     sh->sps = sps;
130     sh->pps = pps;
131 
132     sh->i_first_mb  = 0;
133     sh->i_last_mb   = h->mb.i_mb_count - 1;
134     sh->i_pps_id    = pps->i_id;
135 
136     sh->i_frame_num = i_frame;
137 
138     sh->b_mbaff = PARAM_INTERLACED;
139     sh->b_field_pic = 0;    /* no field support for now */
140     sh->b_bottom_field = 0; /* not yet used */
141 
142     sh->i_idr_pic_id = i_idr_pic_id;
143 
144     /* poc stuff, fixed later */
145     sh->i_poc = 0;
146     sh->i_delta_poc_bottom = 0;
147     sh->i_delta_poc[0] = 0;
148     sh->i_delta_poc[1] = 0;
149 
150     sh->i_redundant_pic_cnt = 0;
151 
152     h->mb.b_direct_auto_write = h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO
153                                 && h->param.i_bframe
154                                 && ( h->param.rc.b_stat_write || !h->param.rc.b_stat_read );
155 
156     if( !h->mb.b_direct_auto_read && sh->i_type == SLICE_TYPE_B )
157     {
158         if( h->fref[1][0]->i_poc_l0ref0 == h->fref[0][0]->i_poc )
159         {
160             if( h->mb.b_direct_auto_write )
161                 sh->b_direct_spatial_mv_pred = ( h->stat.i_direct_score[1] > h->stat.i_direct_score[0] );
162             else
163                 sh->b_direct_spatial_mv_pred = ( param->analyse.i_direct_mv_pred == X264_DIRECT_PRED_SPATIAL );
164         }
165         else
166         {
167             h->mb.b_direct_auto_write = 0;
168             sh->b_direct_spatial_mv_pred = 1;
169         }
170     }
171     /* else b_direct_spatial_mv_pred was read from the 2pass statsfile */
172 
173     sh->b_num_ref_idx_override = 0;
174     sh->i_num_ref_idx_l0_active = 1;
175     sh->i_num_ref_idx_l1_active = 1;
176 
177     sh->b_ref_pic_list_reordering[0] = h->b_ref_reorder[0];
178     sh->b_ref_pic_list_reordering[1] = h->b_ref_reorder[1];
179 
180     /* If the ref list isn't in the default order, construct reordering header */
181     for( int list = 0; list < 2; list++ )
182     {
183         if( sh->b_ref_pic_list_reordering[list] )
184         {
185             int pred_frame_num = i_frame;
186             for( int i = 0; i < h->i_ref[list]; i++ )
187             {
188                 int diff = h->fref[list][i]->i_frame_num - pred_frame_num;
189                 sh->ref_pic_list_order[list][i].idc = ( diff > 0 );
190                 sh->ref_pic_list_order[list][i].arg = (abs(diff) - 1) & ((1 << sps->i_log2_max_frame_num) - 1);
191                 pred_frame_num = h->fref[list][i]->i_frame_num;
192             }
193         }
194     }
195 
196     sh->i_cabac_init_idc = param->i_cabac_init_idc;
197 
198     sh->i_qp = SPEC_QP(i_qp);
199     sh->i_qp_delta = sh->i_qp - pps->i_pic_init_qp;
200     sh->b_sp_for_swidth = 0;
201     sh->i_qs_delta = 0;
202 
203     int deblock_thresh = i_qp + 2 * X264_MIN(param->i_deblocking_filter_alphac0, param->i_deblocking_filter_beta);
204     /* If effective qp <= 15, deblocking would have no effect anyway */
205     if( param->b_deblocking_filter && (h->mb.b_variable_qp || 15 < deblock_thresh ) )
206         sh->i_disable_deblocking_filter_idc = param->b_sliced_threads ? 2 : 0;
207     else
208         sh->i_disable_deblocking_filter_idc = 1;
209     sh->i_alpha_c0_offset = param->i_deblocking_filter_alphac0 << 1;
210     sh->i_beta_offset = param->i_deblocking_filter_beta << 1;
211 }
212 
slice_header_write(bs_t * s,x264_slice_header_t * sh,int i_nal_ref_idc)213 static void slice_header_write( bs_t *s, x264_slice_header_t *sh, int i_nal_ref_idc )
214 {
215     if( sh->b_mbaff )
216     {
217         int first_x = sh->i_first_mb % sh->sps->i_mb_width;
218         int first_y = sh->i_first_mb / sh->sps->i_mb_width;
219         assert( (first_y&1) == 0 );
220         bs_write_ue( s, (2*first_x + sh->sps->i_mb_width*(first_y&~1) + (first_y&1)) >> 1 );
221     }
222     else
223         bs_write_ue( s, sh->i_first_mb );
224 
225     bs_write_ue( s, sh->i_type + 5 );   /* same type things */
226     bs_write_ue( s, sh->i_pps_id );
227     bs_write( s, sh->sps->i_log2_max_frame_num, sh->i_frame_num & ((1<<sh->sps->i_log2_max_frame_num)-1) );
228 
229     if( !sh->sps->b_frame_mbs_only )
230     {
231         bs_write1( s, sh->b_field_pic );
232         if( sh->b_field_pic )
233             bs_write1( s, sh->b_bottom_field );
234     }
235 
236     if( sh->i_idr_pic_id >= 0 ) /* NAL IDR */
237         bs_write_ue( s, sh->i_idr_pic_id );
238 
239     if( sh->sps->i_poc_type == 0 )
240     {
241         bs_write( s, sh->sps->i_log2_max_poc_lsb, sh->i_poc & ((1<<sh->sps->i_log2_max_poc_lsb)-1) );
242         if( sh->pps->b_pic_order && !sh->b_field_pic )
243             bs_write_se( s, sh->i_delta_poc_bottom );
244     }
245 
246     if( sh->pps->b_redundant_pic_cnt )
247         bs_write_ue( s, sh->i_redundant_pic_cnt );
248 
249     if( sh->i_type == SLICE_TYPE_B )
250         bs_write1( s, sh->b_direct_spatial_mv_pred );
251 
252     if( sh->i_type == SLICE_TYPE_P || sh->i_type == SLICE_TYPE_B )
253     {
254         bs_write1( s, sh->b_num_ref_idx_override );
255         if( sh->b_num_ref_idx_override )
256         {
257             bs_write_ue( s, sh->i_num_ref_idx_l0_active - 1 );
258             if( sh->i_type == SLICE_TYPE_B )
259                 bs_write_ue( s, sh->i_num_ref_idx_l1_active - 1 );
260         }
261     }
262 
263     /* ref pic list reordering */
264     if( sh->i_type != SLICE_TYPE_I )
265     {
266         bs_write1( s, sh->b_ref_pic_list_reordering[0] );
267         if( sh->b_ref_pic_list_reordering[0] )
268         {
269             for( int i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
270             {
271                 bs_write_ue( s, sh->ref_pic_list_order[0][i].idc );
272                 bs_write_ue( s, sh->ref_pic_list_order[0][i].arg );
273             }
274             bs_write_ue( s, 3 );
275         }
276     }
277     if( sh->i_type == SLICE_TYPE_B )
278     {
279         bs_write1( s, sh->b_ref_pic_list_reordering[1] );
280         if( sh->b_ref_pic_list_reordering[1] )
281         {
282             for( int i = 0; i < sh->i_num_ref_idx_l1_active; i++ )
283             {
284                 bs_write_ue( s, sh->ref_pic_list_order[1][i].idc );
285                 bs_write_ue( s, sh->ref_pic_list_order[1][i].arg );
286             }
287             bs_write_ue( s, 3 );
288         }
289     }
290 
291     sh->b_weighted_pred = 0;
292     if( sh->pps->b_weighted_pred && sh->i_type == SLICE_TYPE_P )
293     {
294         sh->b_weighted_pred = sh->weight[0][0].weightfn || sh->weight[0][1].weightfn || sh->weight[0][2].weightfn;
295         /* pred_weight_table() */
296         bs_write_ue( s, sh->weight[0][0].i_denom ); /* luma_log2_weight_denom */
297         if( sh->sps->i_chroma_format_idc )
298             bs_write_ue( s, sh->weight[0][1].i_denom ); /* chroma_log2_weight_denom */
299         for( int i = 0; i < sh->i_num_ref_idx_l0_active; i++ )
300         {
301             int luma_weight_l0_flag = !!sh->weight[i][0].weightfn;
302             bs_write1( s, luma_weight_l0_flag );
303             if( luma_weight_l0_flag )
304             {
305                 bs_write_se( s, sh->weight[i][0].i_scale );
306                 bs_write_se( s, sh->weight[i][0].i_offset );
307             }
308             if( sh->sps->i_chroma_format_idc )
309             {
310                 int chroma_weight_l0_flag = sh->weight[i][1].weightfn || sh->weight[i][2].weightfn;
311                 bs_write1( s, chroma_weight_l0_flag );
312                 if( chroma_weight_l0_flag )
313                 {
314                     for( int j = 1; j < 3; j++ )
315                     {
316                         bs_write_se( s, sh->weight[i][j].i_scale );
317                         bs_write_se( s, sh->weight[i][j].i_offset );
318                     }
319                 }
320             }
321         }
322     }
323     else if( sh->pps->b_weighted_bipred == 1 && sh->i_type == SLICE_TYPE_B )
324     {
325       /* TODO */
326     }
327 
328     if( i_nal_ref_idc != 0 )
329     {
330         if( sh->i_idr_pic_id >= 0 )
331         {
332             bs_write1( s, 0 );  /* no output of prior pics flag */
333             bs_write1( s, 0 );  /* long term reference flag */
334         }
335         else
336         {
337             bs_write1( s, sh->i_mmco_command_count > 0 ); /* adaptive_ref_pic_marking_mode_flag */
338             if( sh->i_mmco_command_count > 0 )
339             {
340                 for( int i = 0; i < sh->i_mmco_command_count; i++ )
341                 {
342                     bs_write_ue( s, 1 ); /* mark short term ref as unused */
343                     bs_write_ue( s, sh->mmco[i].i_difference_of_pic_nums - 1 );
344                 }
345                 bs_write_ue( s, 0 ); /* end command list */
346             }
347         }
348     }
349 
350     if( sh->pps->b_cabac && sh->i_type != SLICE_TYPE_I )
351         bs_write_ue( s, sh->i_cabac_init_idc );
352 
353     bs_write_se( s, sh->i_qp_delta );      /* slice qp delta */
354 
355     if( sh->pps->b_deblocking_filter_control )
356     {
357         bs_write_ue( s, sh->i_disable_deblocking_filter_idc );
358         if( sh->i_disable_deblocking_filter_idc != 1 )
359         {
360             bs_write_se( s, sh->i_alpha_c0_offset >> 1 );
361             bs_write_se( s, sh->i_beta_offset >> 1 );
362         }
363     }
364 }
365 
366 /* If we are within a reasonable distance of the end of the memory allocated for the bitstream, */
367 /* reallocate, adding an arbitrary amount of space. */
bitstream_check_buffer_internal(x264_t * h,int size,int b_cabac,int i_nal)368 static int bitstream_check_buffer_internal( x264_t *h, int size, int b_cabac, int i_nal )
369 {
370     if( (b_cabac && (h->cabac.p_end - h->cabac.p < size)) ||
371         (h->out.bs.p_end - h->out.bs.p < size) )
372     {
373         if( size > INT_MAX - h->out.i_bitstream )
374             return -1;
375         int buf_size = h->out.i_bitstream + size;
376         uint8_t *buf = x264_malloc( buf_size );
377         if( !buf )
378             return -1;
379         int aligned_size = h->out.i_bitstream & ~15;
380         h->mc.memcpy_aligned( buf, h->out.p_bitstream, aligned_size );
381         memcpy( buf + aligned_size, h->out.p_bitstream + aligned_size, h->out.i_bitstream - aligned_size );
382 
383         intptr_t delta = buf - h->out.p_bitstream;
384 
385         h->out.bs.p_start += delta;
386         h->out.bs.p += delta;
387         h->out.bs.p_end = buf + buf_size;
388 
389         h->cabac.p_start += delta;
390         h->cabac.p += delta;
391         h->cabac.p_end = buf + buf_size;
392 
393         for( int i = 0; i <= i_nal; i++ )
394             h->out.nal[i].p_payload += delta;
395 
396         x264_free( h->out.p_bitstream );
397         h->out.p_bitstream = buf;
398         h->out.i_bitstream = buf_size;
399     }
400     return 0;
401 }
402 
bitstream_check_buffer(x264_t * h)403 static int bitstream_check_buffer( x264_t *h )
404 {
405     int max_row_size = (2500 << SLICE_MBAFF) * h->mb.i_mb_width;
406     return bitstream_check_buffer_internal( h, max_row_size, h->param.b_cabac, h->out.i_nal );
407 }
408 
bitstream_check_buffer_filler(x264_t * h,int filler)409 static int bitstream_check_buffer_filler( x264_t *h, int filler )
410 {
411     filler += 32; // add padding for safety
412     return bitstream_check_buffer_internal( h, filler, 0, -1 );
413 }
414 
415 #if HAVE_THREAD
encoder_thread_init(x264_t * h)416 static void encoder_thread_init( x264_t *h )
417 {
418     if( h->param.i_sync_lookahead )
419         x264_lower_thread_priority( 10 );
420 }
421 #endif
422 
423 /****************************************************************************
424  *
425  ****************************************************************************
426  ****************************** External API*********************************
427  ****************************************************************************
428  *
429  ****************************************************************************/
430 
validate_parameters(x264_t * h,int b_open)431 static int validate_parameters( x264_t *h, int b_open )
432 {
433     if( !h->param.pf_log )
434     {
435         x264_log_internal( X264_LOG_ERROR, "pf_log not set! did you forget to call x264_param_default?\n" );
436         return -1;
437     }
438 
439 #if HAVE_MMX
440     if( b_open )
441     {
442         uint32_t cpuflags = x264_cpu_detect();
443         int fail = 0;
444 #ifdef __SSE__
445         if( !(cpuflags & X264_CPU_SSE) )
446         {
447             x264_log( h, X264_LOG_ERROR, "your cpu does not support SSE1, but x264 was compiled with asm\n");
448             fail = 1;
449         }
450 #else
451         if( !(cpuflags & X264_CPU_MMX2) )
452         {
453             x264_log( h, X264_LOG_ERROR, "your cpu does not support MMXEXT, but x264 was compiled with asm\n");
454             fail = 1;
455         }
456 #endif
457         if( fail )
458         {
459             x264_log( h, X264_LOG_ERROR, "to run x264, recompile without asm (configure --disable-asm)\n");
460             return -1;
461         }
462     }
463 #endif
464 
465 #if HAVE_INTERLACED
466     h->param.b_interlaced = !!PARAM_INTERLACED;
467 #else
468     if( h->param.b_interlaced )
469     {
470         x264_log( h, X264_LOG_ERROR, "not compiled with interlaced support\n" );
471         return -1;
472     }
473 #endif
474 
475 #define MAX_RESOLUTION 16384
476     if( h->param.i_width <= 0 || h->param.i_height <= 0 ||
477         h->param.i_width > MAX_RESOLUTION || h->param.i_height > MAX_RESOLUTION )
478     {
479         x264_log( h, X264_LOG_ERROR, "invalid width x height (%dx%d)\n",
480                   h->param.i_width, h->param.i_height );
481         return -1;
482     }
483 
484     int i_csp = h->param.i_csp & X264_CSP_MASK;
485 #if X264_CHROMA_FORMAT
486     if( CHROMA_FORMAT != CHROMA_400 && i_csp == X264_CSP_I400 )
487     {
488         x264_log( h, X264_LOG_ERROR, "not compiled with 4:0:0 support\n" );
489         return -1;
490     }
491     else if( CHROMA_FORMAT != CHROMA_420 && i_csp >= X264_CSP_I420 && i_csp < X264_CSP_I422 )
492     {
493         x264_log( h, X264_LOG_ERROR, "not compiled with 4:2:0 support\n" );
494         return -1;
495     }
496     else if( CHROMA_FORMAT != CHROMA_422 && i_csp >= X264_CSP_I422 && i_csp < X264_CSP_I444 )
497     {
498         x264_log( h, X264_LOG_ERROR, "not compiled with 4:2:2 support\n" );
499         return -1;
500     }
501     else if( CHROMA_FORMAT != CHROMA_444 && i_csp >= X264_CSP_I444 && i_csp <= X264_CSP_RGB )
502     {
503         x264_log( h, X264_LOG_ERROR, "not compiled with 4:4:4 support\n" );
504         return -1;
505     }
506 #endif
507     if( i_csp <= X264_CSP_NONE || i_csp >= X264_CSP_MAX )
508     {
509         x264_log( h, X264_LOG_ERROR, "invalid CSP (only I400/I420/YV12/NV12/NV21/I422/YV16/NV16/YUYV/UYVY/"
510                                      "I444/YV24/BGR/BGRA/RGB supported)\n" );
511         return -1;
512     }
513 
514     int w_mod = 1;
515     int h_mod = 1 << (PARAM_INTERLACED || h->param.b_fake_interlaced);
516     if( i_csp == X264_CSP_I400 )
517     {
518         h->param.analyse.i_chroma_qp_offset = 0;
519         h->param.analyse.b_chroma_me = 0;
520         h->param.vui.i_colmatrix = 2; /* undefined */
521     }
522     else if( i_csp < X264_CSP_I444 )
523     {
524         w_mod = 2;
525         if( i_csp < X264_CSP_I422 )
526             h_mod *= 2;
527     }
528 
529     if( h->param.i_width % w_mod )
530     {
531         x264_log( h, X264_LOG_ERROR, "width not divisible by %d (%dx%d)\n",
532                   w_mod, h->param.i_width, h->param.i_height );
533         return -1;
534     }
535     if( h->param.i_height % h_mod )
536     {
537         x264_log( h, X264_LOG_ERROR, "height not divisible by %d (%dx%d)\n",
538                   h_mod, h->param.i_width, h->param.i_height );
539         return -1;
540     }
541 
542     if( h->param.crop_rect.i_left   < 0 || h->param.crop_rect.i_left   >= h->param.i_width ||
543         h->param.crop_rect.i_right  < 0 || h->param.crop_rect.i_right  >= h->param.i_width ||
544         h->param.crop_rect.i_top    < 0 || h->param.crop_rect.i_top    >= h->param.i_height ||
545         h->param.crop_rect.i_bottom < 0 || h->param.crop_rect.i_bottom >= h->param.i_height ||
546         h->param.crop_rect.i_left + h->param.crop_rect.i_right  >= h->param.i_width ||
547         h->param.crop_rect.i_top  + h->param.crop_rect.i_bottom >= h->param.i_height )
548     {
549         x264_log( h, X264_LOG_ERROR, "invalid crop-rect %d,%d,%d,%d\n", h->param.crop_rect.i_left,
550                   h->param.crop_rect.i_top, h->param.crop_rect.i_right,  h->param.crop_rect.i_bottom );
551         return -1;
552     }
553     if( h->param.crop_rect.i_left % w_mod || h->param.crop_rect.i_right  % w_mod ||
554         h->param.crop_rect.i_top  % h_mod || h->param.crop_rect.i_bottom % h_mod )
555     {
556         x264_log( h, X264_LOG_ERROR, "crop-rect %d,%d,%d,%d not divisible by %dx%d\n", h->param.crop_rect.i_left,
557                   h->param.crop_rect.i_top, h->param.crop_rect.i_right,  h->param.crop_rect.i_bottom, w_mod, h_mod );
558         return -1;
559     }
560 
561     if( h->param.vui.i_sar_width <= 0 || h->param.vui.i_sar_height <= 0 )
562     {
563         h->param.vui.i_sar_width = 0;
564         h->param.vui.i_sar_height = 0;
565     }
566 
567     if( h->param.i_threads == X264_THREADS_AUTO )
568     {
569         h->param.i_threads = x264_cpu_num_processors() * (h->param.b_sliced_threads?2:3)/2;
570         /* Avoid too many threads as they don't improve performance and
571          * complicate VBV. Capped at an arbitrary 2 rows per thread. */
572         int max_threads = X264_MAX( 1, (h->param.i_height+15)/16 / 2 );
573         h->param.i_threads = X264_MIN( h->param.i_threads, max_threads );
574     }
575     int max_sliced_threads = X264_MAX( 1, (h->param.i_height+15)/16 / 4 );
576     if( h->param.i_threads > 1 )
577     {
578 #if !HAVE_THREAD
579         x264_log( h, X264_LOG_WARNING, "not compiled with thread support!\n");
580         h->param.i_threads = 1;
581 #endif
582         /* Avoid absurdly small thread slices as they can reduce performance
583          * and VBV compliance.  Capped at an arbitrary 4 rows per thread. */
584         if( h->param.b_sliced_threads )
585             h->param.i_threads = X264_MIN( h->param.i_threads, max_sliced_threads );
586     }
587     h->param.i_threads = x264_clip3( h->param.i_threads, 1, X264_THREAD_MAX );
588     if( h->param.i_threads == 1 )
589     {
590         h->param.b_sliced_threads = 0;
591         h->param.i_lookahead_threads = 1;
592     }
593     h->i_thread_frames = h->param.b_sliced_threads ? 1 : h->param.i_threads;
594     if( h->i_thread_frames > 1 )
595         h->param.nalu_process = NULL;
596 
597     if( h->param.b_opencl )
598     {
599 #if !HAVE_OPENCL
600         x264_log( h, X264_LOG_WARNING, "OpenCL: not compiled with OpenCL support, disabling\n" );
601         h->param.b_opencl = 0;
602 #elif BIT_DEPTH > 8
603         x264_log( h, X264_LOG_WARNING, "OpenCL lookahead does not support high bit depth, disabling opencl\n" );
604         h->param.b_opencl = 0;
605 #else
606         if( h->param.i_width < 32 || h->param.i_height < 32 )
607         {
608             x264_log( h, X264_LOG_WARNING, "OpenCL: frame size is too small, disabling opencl\n" );
609             h->param.b_opencl = 0;
610         }
611 #endif
612         if( h->param.opencl_device_id && h->param.i_opencl_device )
613         {
614             x264_log( h, X264_LOG_WARNING, "OpenCL: device id and device skip count configured; dropping skip\n" );
615             h->param.i_opencl_device = 0;
616         }
617     }
618 
619     h->param.i_keyint_max = x264_clip3( h->param.i_keyint_max, 1, X264_KEYINT_MAX_INFINITE );
620     if( h->param.i_keyint_max == 1 )
621     {
622         h->param.b_intra_refresh = 0;
623         h->param.analyse.i_weighted_pred = 0;
624         h->param.i_frame_reference = 1;
625         h->param.i_dpb_size = 1;
626     }
627 
628     if( h->param.i_frame_packing < -1 || h->param.i_frame_packing > 7 )
629     {
630         x264_log( h, X264_LOG_WARNING, "ignoring unknown frame packing value\n" );
631         h->param.i_frame_packing = -1;
632     }
633     if( h->param.i_frame_packing == 7 &&
634         ((h->param.i_width - h->param.crop_rect.i_left - h->param.crop_rect.i_right)  % 3 ||
635          (h->param.i_height - h->param.crop_rect.i_top - h->param.crop_rect.i_bottom) % 3) )
636     {
637         x264_log( h, X264_LOG_ERROR, "cropped resolution %dx%d not compatible with tile format frame packing\n",
638                   h->param.i_width - h->param.crop_rect.i_left - h->param.crop_rect.i_right,
639                   h->param.i_height - h->param.crop_rect.i_top - h->param.crop_rect.i_bottom );
640         return -1;
641     }
642 
643     if( h->param.mastering_display.b_mastering_display )
644     {
645         if( h->param.mastering_display.i_green_x > UINT16_MAX || h->param.mastering_display.i_green_x < 0 ||
646             h->param.mastering_display.i_green_y > UINT16_MAX || h->param.mastering_display.i_green_y < 0 ||
647             h->param.mastering_display.i_blue_x > UINT16_MAX || h->param.mastering_display.i_blue_x < 0 ||
648             h->param.mastering_display.i_blue_y > UINT16_MAX || h->param.mastering_display.i_blue_y < 0 ||
649             h->param.mastering_display.i_red_x > UINT16_MAX || h->param.mastering_display.i_red_x < 0 ||
650             h->param.mastering_display.i_red_y > UINT16_MAX || h->param.mastering_display.i_red_y < 0 ||
651             h->param.mastering_display.i_white_x > UINT16_MAX || h->param.mastering_display.i_white_x < 0 ||
652             h->param.mastering_display.i_white_y > UINT16_MAX || h->param.mastering_display.i_white_y < 0 )
653         {
654             x264_log( h, X264_LOG_ERROR, "mastering display xy coordinates out of range [0,%u]\n", UINT16_MAX );
655             return -1;
656         }
657         if( h->param.mastering_display.i_display_max > UINT32_MAX || h->param.mastering_display.i_display_max < 0 ||
658             h->param.mastering_display.i_display_min > UINT32_MAX || h->param.mastering_display.i_display_min < 0 )
659         {
660             x264_log( h, X264_LOG_ERROR, "mastering display brightness out of range [0,%u]\n", UINT32_MAX );
661             return -1;
662         }
663         if( h->param.mastering_display.i_display_min == 50000 && h->param.mastering_display.i_display_max == 50000 )
664         {
665             x264_log( h, X264_LOG_ERROR, "mastering display min and max brightness cannot both be 50000\n" );
666             return -1;
667         }
668     }
669 
670     if( h->param.content_light_level.b_cll &&
671         (h->param.content_light_level.i_max_cll > UINT16_MAX || h->param.content_light_level.i_max_cll < 0 ||
672          h->param.content_light_level.i_max_fall > UINT16_MAX || h->param.content_light_level.i_max_fall < 0) )
673     {
674         x264_log( h, X264_LOG_ERROR, "content light levels out of range [0,%u]\n", UINT16_MAX );
675         return -1;
676     }
677 
678     /* Detect default ffmpeg settings and terminate with an error. */
679     if( b_open )
680     {
681         int score = 0;
682         score += h->param.analyse.i_me_range == 0;
683         score += h->param.rc.i_qp_step == 3;
684         score += h->param.i_keyint_max == 12;
685         score += h->param.rc.i_qp_min == 2;
686         score += h->param.rc.i_qp_max == 31;
687         score += h->param.rc.f_qcompress == 0.5;
688         score += fabs(h->param.rc.f_ip_factor - 1.25) < 0.01;
689         score += fabs(h->param.rc.f_pb_factor - 1.25) < 0.01;
690         score += h->param.analyse.inter == 0 && h->param.analyse.i_subpel_refine == 8;
691         if( score >= 5 )
692         {
693             x264_log( h, X264_LOG_ERROR, "broken ffmpeg default settings detected\n" );
694             x264_log( h, X264_LOG_ERROR, "use an encoding preset (e.g. -vpre medium)\n" );
695             x264_log( h, X264_LOG_ERROR, "preset usage: -vpre <speed> -vpre <profile>\n" );
696             x264_log( h, X264_LOG_ERROR, "speed presets are listed in x264 --help\n" );
697             x264_log( h, X264_LOG_ERROR, "profile is optional; x264 defaults to high\n" );
698             return -1;
699         }
700     }
701 
702     if( h->param.rc.i_rc_method < 0 || h->param.rc.i_rc_method > 2 )
703     {
704         x264_log( h, X264_LOG_ERROR, "no ratecontrol method specified\n" );
705         return -1;
706     }
707 
708     if( PARAM_INTERLACED )
709         h->param.b_pic_struct = 1;
710 
711     if( h->param.i_avcintra_class )
712     {
713         if( BIT_DEPTH != 10 )
714         {
715             x264_log( h, X264_LOG_ERROR, "%2d-bit AVC-Intra is not widely compatible\n", BIT_DEPTH );
716             x264_log( h, X264_LOG_ERROR, "10-bit x264 is required to encode AVC-Intra\n" );
717             return -1;
718         }
719 
720         int type = h->param.i_avcintra_class == 200 ? 2 :
721                    h->param.i_avcintra_class == 100 ? 1 :
722                    h->param.i_avcintra_class == 50 ? 0 : -1;
723         if( type < 0 )
724         {
725             x264_log( h, X264_LOG_ERROR, "Invalid AVC-Intra class\n" );
726             return -1;
727         }
728 
729         /* [50/100/200][res][fps] */
730         static const struct
731         {
732             uint16_t fps_num;
733             uint16_t fps_den;
734             uint8_t interlaced;
735             uint16_t frame_size;
736             const uint8_t *cqm_4ic;
737             const uint8_t *cqm_8iy;
738         } avcintra_lut[3][2][7] =
739         {
740             {{{ 60000, 1001, 0,  912, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
741               {    50,    1, 0, 1100, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
742               { 30000, 1001, 0,  912, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
743               {    25,    1, 0, 1100, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
744               { 24000, 1001, 0,  912, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy }},
745              {{ 30000, 1001, 1, 1820, x264_cqm_avci50_4ic, x264_cqm_avci50_1080i_8iy },
746               {    25,    1, 1, 2196, x264_cqm_avci50_4ic, x264_cqm_avci50_1080i_8iy },
747               { 60000, 1001, 0, 1820, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
748               { 30000, 1001, 0, 1820, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
749               {    50,    1, 0, 2196, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
750               {    25,    1, 0, 2196, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy },
751               { 24000, 1001, 0, 1820, x264_cqm_avci50_4ic, x264_cqm_avci50_p_8iy }}},
752             {{{ 60000, 1001, 0, 1848, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
753               {    50,    1, 0, 2224, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
754               { 30000, 1001, 0, 1848, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
755               {    25,    1, 0, 2224, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
756               { 24000, 1001, 0, 1848, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy }},
757              {{ 30000, 1001, 1, 3692, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080i_8iy },
758               {    25,    1, 1, 4444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080i_8iy },
759               { 60000, 1001, 0, 3692, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
760               { 30000, 1001, 0, 3692, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
761               {    50,    1, 0, 4444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
762               {    25,    1, 0, 4444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
763               { 24000, 1001, 0, 3692, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy }}},
764             {{{ 60000, 1001, 0, 3724, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy },
765               {    50,    1, 0, 4472, x264_cqm_avci100_720p_4ic, x264_cqm_avci100_720p_8iy }},
766              {{ 30000, 1001, 1, 7444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080i_8iy },
767               {    25,    1, 1, 8940, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080i_8iy },
768               { 60000, 1001, 0, 7444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
769               { 30000, 1001, 0, 7444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
770               {    50,    1, 0, 8940, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
771               {    25,    1, 0, 8940, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy },
772               { 24000, 1001, 0, 7444, x264_cqm_avci100_1080_4ic, x264_cqm_avci100_1080p_8iy }}}
773         };
774 
775         int res = -1;
776         if( i_csp >= X264_CSP_I420 && i_csp < X264_CSP_I422 && !type )
777         {
778             if(      h->param.i_width == 1440 && h->param.i_height == 1080 ) res =  1;
779             else if( h->param.i_width ==  960 && h->param.i_height ==  720 ) res =  0;
780         }
781         else if( i_csp >= X264_CSP_I422 && i_csp < X264_CSP_I444 && type )
782         {
783             if(      h->param.i_width == 1920 && h->param.i_height == 1080 ) res =  1;
784             else if( h->param.i_width == 1280 && h->param.i_height ==  720 ) res =  0;
785         }
786         else
787         {
788             x264_log( h, X264_LOG_ERROR, "Invalid colorspace for AVC-Intra %d\n", h->param.i_avcintra_class );
789             return -1;
790         }
791 
792         if( res < 0 )
793         {
794             x264_log( h, X264_LOG_ERROR, "Resolution %dx%d invalid for AVC-Intra %d\n",
795                       h->param.i_width, h->param.i_height, h->param.i_avcintra_class );
796             return -1;
797         }
798 
799         if( h->param.nalu_process )
800         {
801             x264_log( h, X264_LOG_ERROR, "nalu_process is not supported in AVC-Intra mode\n" );
802             return -1;
803         }
804 
805         if( !h->param.b_repeat_headers )
806         {
807             x264_log( h, X264_LOG_ERROR, "Separate headers not supported in AVC-Intra mode\n" );
808             return -1;
809         }
810 
811         int i;
812         uint32_t fps_num = h->param.i_fps_num, fps_den = h->param.i_fps_den;
813         x264_reduce_fraction( &fps_num, &fps_den );
814         for( i = 0; i < 7; i++ )
815         {
816             if( avcintra_lut[type][res][i].fps_num == fps_num &&
817                 avcintra_lut[type][res][i].fps_den == fps_den &&
818                 avcintra_lut[type][res][i].interlaced == PARAM_INTERLACED )
819             {
820                 break;
821             }
822         }
823         if( i == 7 )
824         {
825             x264_log( h, X264_LOG_ERROR, "FPS %d/%d%c not compatible with AVC-Intra\n",
826                       h->param.i_fps_num, h->param.i_fps_den, PARAM_INTERLACED ? 'i' : 'p' );
827             return -1;
828         }
829 
830         h->param.i_keyint_max = 1;
831         h->param.b_intra_refresh = 0;
832         h->param.analyse.i_weighted_pred = 0;
833         h->param.i_frame_reference = 1;
834         h->param.i_dpb_size = 1;
835 
836         h->param.b_bluray_compat = 0;
837         h->param.b_vfr_input = 0;
838         h->param.b_aud = 1;
839         h->param.vui.i_chroma_loc = 0;
840         h->param.i_nal_hrd = X264_NAL_HRD_NONE;
841         h->param.b_deblocking_filter = 0;
842         h->param.b_stitchable = 1;
843         h->param.b_pic_struct = 0;
844         h->param.analyse.b_transform_8x8 = 1;
845         h->param.analyse.intra = X264_ANALYSE_I8x8;
846         h->param.analyse.i_chroma_qp_offset = res && type ? 3 : 4;
847         h->param.b_cabac = !type;
848         h->param.rc.i_vbv_buffer_size = avcintra_lut[type][res][i].frame_size;
849         h->param.rc.i_vbv_max_bitrate =
850         h->param.rc.i_bitrate = h->param.rc.i_vbv_buffer_size * fps_num / fps_den;
851         h->param.rc.i_rc_method = X264_RC_ABR;
852         h->param.rc.f_vbv_buffer_init = 1.0;
853         h->param.rc.b_filler = 1;
854         h->param.i_cqm_preset = X264_CQM_CUSTOM;
855         memcpy( h->param.cqm_4iy, x264_cqm_jvt4i, sizeof(h->param.cqm_4iy) );
856         memcpy( h->param.cqm_4ic, avcintra_lut[type][res][i].cqm_4ic, sizeof(h->param.cqm_4ic) );
857         memcpy( h->param.cqm_8iy, avcintra_lut[type][res][i].cqm_8iy, sizeof(h->param.cqm_8iy) );
858 
859         /* Sony XAVC flavor much more simple */
860         if( h->param.i_avcintra_flavor == X264_AVCINTRA_FLAVOR_SONY )
861         {
862             h->param.i_slice_count = 8;
863             if( h->param.b_sliced_threads )
864                 h->param.i_threads = h->param.i_slice_count;
865             /* Sony XAVC unlike AVC-Intra doesn't seem to have a QP floor */
866         }
867         else
868         {
869             /* Need exactly 10 slices of equal MB count... why?  $deity knows... */
870             h->param.i_slice_max_mbs = ((h->param.i_width + 15) / 16) * ((h->param.i_height + 15) / 16) / 10;
871             h->param.i_slice_max_size = 0;
872             /* The slice structure only allows a maximum of 2 threads for 1080i/p
873              * and 1 or 5 threads for 720p */
874             if( h->param.b_sliced_threads )
875             {
876                 if( res )
877                     h->param.i_threads = X264_MIN( 2, h->param.i_threads );
878                 else
879                 {
880                     h->param.i_threads = X264_MIN( 5, h->param.i_threads );
881                     if( h->param.i_threads < 5 )
882                         h->param.i_threads = 1;
883                 }
884             }
885 
886             /* Official encoder doesn't appear to go under 13
887              * and Avid cannot handle negative QPs */
888             h->param.rc.i_qp_min = X264_MAX( h->param.rc.i_qp_min, QP_BD_OFFSET + 1 );
889         }
890 
891         if( type )
892             h->param.vui.i_sar_width = h->param.vui.i_sar_height = 1;
893         else
894         {
895             h->param.vui.i_sar_width  = 4;
896             h->param.vui.i_sar_height = 3;
897         }
898     }
899 
900     h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, -QP_BD_OFFSET, 51 );
901     h->param.rc.f_rf_constant_max = x264_clip3f( h->param.rc.f_rf_constant_max, -QP_BD_OFFSET, 51 );
902     h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, -1, QP_MAX );
903     h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 11 );
904     h->param.rc.f_ip_factor = x264_clip3f( h->param.rc.f_ip_factor, 0.01, 10.0 );
905     h->param.rc.f_pb_factor = x264_clip3f( h->param.rc.f_pb_factor, 0.01, 10.0 );
906     if( h->param.rc.i_rc_method == X264_RC_CRF )
907     {
908         h->param.rc.i_qp_constant = h->param.rc.f_rf_constant + QP_BD_OFFSET;
909         h->param.rc.i_bitrate = 0;
910     }
911     if( b_open && (h->param.rc.i_rc_method == X264_RC_CQP || h->param.rc.i_rc_method == X264_RC_CRF)
912         && h->param.rc.i_qp_constant == 0 )
913     {
914         h->mb.b_lossless = 1;
915         h->param.i_cqm_preset = X264_CQM_FLAT;
916         h->param.psz_cqm_file = NULL;
917         h->param.rc.i_rc_method = X264_RC_CQP;
918         h->param.rc.f_ip_factor = 1;
919         h->param.rc.f_pb_factor = 1;
920         h->param.analyse.b_psnr = 0;
921         h->param.analyse.b_ssim = 0;
922         h->param.analyse.i_chroma_qp_offset = 0;
923         h->param.analyse.i_trellis = 0;
924         h->param.analyse.b_fast_pskip = 0;
925         h->param.analyse.i_noise_reduction = 0;
926         h->param.analyse.b_psy = 0;
927         h->param.i_bframe = 0;
928         /* 8x8dct is not useful without RD in CAVLC lossless */
929         if( !h->param.b_cabac && h->param.analyse.i_subpel_refine < 6 )
930             h->param.analyse.b_transform_8x8 = 0;
931     }
932     if( h->param.rc.i_rc_method == X264_RC_CQP )
933     {
934         float qp_p = h->param.rc.i_qp_constant;
935         float qp_i = qp_p - 6*log2f( h->param.rc.f_ip_factor );
936         float qp_b = qp_p + 6*log2f( h->param.rc.f_pb_factor );
937         if( qp_p < 0 )
938         {
939             x264_log( h, X264_LOG_ERROR, "qp not specified\n" );
940             return -1;
941         }
942 
943         h->param.rc.i_qp_min = x264_clip3( (int)(X264_MIN3( qp_p, qp_i, qp_b )), 0, QP_MAX );
944         h->param.rc.i_qp_max = x264_clip3( (int)(X264_MAX3( qp_p, qp_i, qp_b ) + .999), 0, QP_MAX );
945         h->param.rc.i_aq_mode = 0;
946         h->param.rc.b_mb_tree = 0;
947         h->param.rc.i_bitrate = 0;
948     }
949     h->param.rc.i_qp_max = x264_clip3( h->param.rc.i_qp_max, 0, QP_MAX );
950     h->param.rc.i_qp_min = x264_clip3( h->param.rc.i_qp_min, 0, h->param.rc.i_qp_max );
951     h->param.rc.i_qp_step = x264_clip3( h->param.rc.i_qp_step, 2, QP_MAX );
952     h->param.rc.i_bitrate = x264_clip3( h->param.rc.i_bitrate, 0, 2000000 );
953     if( h->param.rc.i_rc_method == X264_RC_ABR && !h->param.rc.i_bitrate )
954     {
955         x264_log( h, X264_LOG_ERROR, "bitrate not specified\n" );
956         return -1;
957     }
958     h->param.rc.i_vbv_buffer_size = x264_clip3( h->param.rc.i_vbv_buffer_size, 0, 2000000 );
959     h->param.rc.i_vbv_max_bitrate = x264_clip3( h->param.rc.i_vbv_max_bitrate, 0, 2000000 );
960     h->param.rc.f_vbv_buffer_init = x264_clip3f( h->param.rc.f_vbv_buffer_init, 0, 2000000 );
961     if( h->param.rc.i_vbv_buffer_size )
962     {
963         if( h->param.rc.i_rc_method == X264_RC_CQP )
964         {
965             x264_log( h, X264_LOG_WARNING, "VBV is incompatible with constant QP, ignored.\n" );
966             h->param.rc.i_vbv_max_bitrate = 0;
967             h->param.rc.i_vbv_buffer_size = 0;
968         }
969         else if( h->param.rc.i_vbv_max_bitrate == 0 )
970         {
971             if( h->param.rc.i_rc_method == X264_RC_ABR )
972             {
973                 x264_log( h, X264_LOG_WARNING, "VBV maxrate unspecified, assuming CBR\n" );
974                 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
975             }
976             else
977             {
978                 x264_log( h, X264_LOG_WARNING, "VBV bufsize set but maxrate unspecified, ignored\n" );
979                 h->param.rc.i_vbv_buffer_size = 0;
980             }
981         }
982         else if( h->param.rc.i_vbv_max_bitrate < h->param.rc.i_bitrate &&
983                  h->param.rc.i_rc_method == X264_RC_ABR )
984         {
985             x264_log( h, X264_LOG_WARNING, "max bitrate less than average bitrate, assuming CBR\n" );
986             h->param.rc.i_bitrate = h->param.rc.i_vbv_max_bitrate;
987         }
988     }
989     else if( h->param.rc.i_vbv_max_bitrate )
990     {
991         x264_log( h, X264_LOG_WARNING, "VBV maxrate specified, but no bufsize, ignored\n" );
992         h->param.rc.i_vbv_max_bitrate = 0;
993     }
994 
995     h->param.i_slice_max_size = X264_MAX( h->param.i_slice_max_size, 0 );
996     h->param.i_slice_max_mbs = X264_MAX( h->param.i_slice_max_mbs, 0 );
997     h->param.i_slice_min_mbs = X264_MAX( h->param.i_slice_min_mbs, 0 );
998     if( h->param.i_slice_max_mbs )
999         h->param.i_slice_min_mbs = X264_MIN( h->param.i_slice_min_mbs, h->param.i_slice_max_mbs/2 );
1000     else if( !h->param.i_slice_max_size )
1001         h->param.i_slice_min_mbs = 0;
1002     if( PARAM_INTERLACED && h->param.i_slice_min_mbs )
1003     {
1004         x264_log( h, X264_LOG_WARNING, "interlace + slice-min-mbs is not implemented\n" );
1005         h->param.i_slice_min_mbs = 0;
1006     }
1007     int mb_width = (h->param.i_width+15)/16;
1008     if( h->param.i_slice_min_mbs > mb_width )
1009     {
1010         x264_log( h, X264_LOG_WARNING, "slice-min-mbs > row mb size (%d) not implemented\n", mb_width );
1011         h->param.i_slice_min_mbs = mb_width;
1012     }
1013 
1014     int max_slices = (h->param.i_height+((16<<PARAM_INTERLACED)-1))/(16<<PARAM_INTERLACED);
1015     if( h->param.b_sliced_threads )
1016         h->param.i_slice_count = x264_clip3( h->param.i_threads, 0, max_slices );
1017     else
1018     {
1019         h->param.i_slice_count = x264_clip3( h->param.i_slice_count, 0, max_slices );
1020         if( h->param.i_slice_max_mbs || h->param.i_slice_max_size )
1021             h->param.i_slice_count = 0;
1022     }
1023     if( h->param.i_slice_count_max > 0 )
1024         h->param.i_slice_count_max = X264_MAX( h->param.i_slice_count, h->param.i_slice_count_max );
1025 
1026     if( h->param.b_bluray_compat )
1027     {
1028         h->param.i_bframe_pyramid = X264_MIN( X264_B_PYRAMID_STRICT, h->param.i_bframe_pyramid );
1029         h->param.i_bframe = X264_MIN( h->param.i_bframe, 3 );
1030         h->param.b_aud = 1;
1031         h->param.i_nal_hrd = X264_MAX( h->param.i_nal_hrd, X264_NAL_HRD_VBR );
1032         h->param.i_slice_max_size = 0;
1033         h->param.i_slice_max_mbs = 0;
1034         h->param.b_intra_refresh = 0;
1035         h->param.i_frame_reference = X264_MIN( h->param.i_frame_reference, 6 );
1036         h->param.i_dpb_size = X264_MIN( h->param.i_dpb_size, 6 );
1037         /* Don't use I-frames, because Blu-ray treats them the same as IDR. */
1038         h->param.i_keyint_min = 1;
1039         /* Due to the proliferation of broken players that don't handle dupes properly. */
1040         h->param.analyse.i_weighted_pred = X264_MIN( h->param.analyse.i_weighted_pred, X264_WEIGHTP_SIMPLE );
1041         if( h->param.b_fake_interlaced )
1042             h->param.b_pic_struct = 1;
1043     }
1044 
1045     h->param.i_frame_reference = x264_clip3( h->param.i_frame_reference, 1, X264_REF_MAX );
1046     h->param.i_dpb_size = x264_clip3( h->param.i_dpb_size, 1, X264_REF_MAX );
1047     if( h->param.i_scenecut_threshold < 0 )
1048         h->param.i_scenecut_threshold = 0;
1049     h->param.analyse.i_direct_mv_pred = x264_clip3( h->param.analyse.i_direct_mv_pred, X264_DIRECT_PRED_NONE, X264_DIRECT_PRED_AUTO );
1050     if( !h->param.analyse.i_subpel_refine && h->param.analyse.i_direct_mv_pred > X264_DIRECT_PRED_SPATIAL )
1051     {
1052         x264_log( h, X264_LOG_WARNING, "subme=0 + direct=temporal is not supported\n" );
1053         h->param.analyse.i_direct_mv_pred = X264_DIRECT_PRED_SPATIAL;
1054     }
1055     h->param.i_bframe = x264_clip3( h->param.i_bframe, 0, X264_MIN( X264_BFRAME_MAX, h->param.i_keyint_max-1 ) );
1056     h->param.i_bframe_bias = x264_clip3( h->param.i_bframe_bias, -90, 100 );
1057     if( h->param.i_bframe <= 1 )
1058         h->param.i_bframe_pyramid = X264_B_PYRAMID_NONE;
1059     h->param.i_bframe_pyramid = x264_clip3( h->param.i_bframe_pyramid, X264_B_PYRAMID_NONE, X264_B_PYRAMID_NORMAL );
1060     h->param.i_bframe_adaptive = x264_clip3( h->param.i_bframe_adaptive, X264_B_ADAPT_NONE, X264_B_ADAPT_TRELLIS );
1061     if( !h->param.i_bframe )
1062     {
1063         h->param.i_bframe_adaptive = X264_B_ADAPT_NONE;
1064         h->param.analyse.i_direct_mv_pred = 0;
1065         h->param.analyse.b_weighted_bipred = 0;
1066         h->param.b_open_gop = 0;
1067     }
1068     if( h->param.b_intra_refresh && h->param.i_bframe_pyramid == X264_B_PYRAMID_NORMAL )
1069     {
1070         x264_log( h, X264_LOG_WARNING, "b-pyramid normal + intra-refresh is not supported\n" );
1071         h->param.i_bframe_pyramid = X264_B_PYRAMID_STRICT;
1072     }
1073     if( h->param.b_intra_refresh && (h->param.i_frame_reference > 1 || h->param.i_dpb_size > 1) )
1074     {
1075         x264_log( h, X264_LOG_WARNING, "ref > 1 + intra-refresh is not supported\n" );
1076         h->param.i_frame_reference = 1;
1077         h->param.i_dpb_size = 1;
1078     }
1079     if( h->param.b_intra_refresh && h->param.b_open_gop )
1080     {
1081         x264_log( h, X264_LOG_WARNING, "intra-refresh is not compatible with open-gop\n" );
1082         h->param.b_open_gop = 0;
1083     }
1084     if( !h->param.i_fps_num || !h->param.i_fps_den )
1085     {
1086         h->param.i_fps_num = 25;
1087         h->param.i_fps_den = 1;
1088     }
1089     float fps = (float)h->param.i_fps_num / h->param.i_fps_den;
1090     if( h->param.i_keyint_min == X264_KEYINT_MIN_AUTO )
1091         h->param.i_keyint_min = X264_MIN( h->param.i_keyint_max / 10, (int)fps );
1092     h->param.i_keyint_min = x264_clip3( h->param.i_keyint_min, 1, h->param.i_keyint_max/2+1 );
1093     h->param.rc.i_lookahead = x264_clip3( h->param.rc.i_lookahead, 0, X264_LOOKAHEAD_MAX );
1094     {
1095         int maxrate = X264_MAX( h->param.rc.i_vbv_max_bitrate, h->param.rc.i_bitrate );
1096         float bufsize = maxrate ? (float)h->param.rc.i_vbv_buffer_size / maxrate : 0;
1097         h->param.rc.i_lookahead = X264_MIN( h->param.rc.i_lookahead, X264_MAX( h->param.i_keyint_max, bufsize*fps ) );
1098     }
1099 
1100     if( !h->param.i_timebase_num || !h->param.i_timebase_den || !(h->param.b_vfr_input || h->param.b_pulldown) )
1101     {
1102         h->param.i_timebase_num = h->param.i_fps_den;
1103         h->param.i_timebase_den = h->param.i_fps_num;
1104     }
1105 
1106     h->param.rc.f_qcompress = x264_clip3f( h->param.rc.f_qcompress, 0.0, 1.0 );
1107     if( h->param.i_keyint_max == 1 || h->param.rc.f_qcompress == 1 )
1108         h->param.rc.b_mb_tree = 0;
1109     if( (!h->param.b_intra_refresh && h->param.i_keyint_max != X264_KEYINT_MAX_INFINITE) &&
1110         !h->param.rc.i_lookahead && h->param.rc.b_mb_tree )
1111     {
1112         x264_log( h, X264_LOG_WARNING, "lookaheadless mb-tree requires intra refresh or infinite keyint\n" );
1113         h->param.rc.b_mb_tree = 0;
1114     }
1115     if( b_open && h->param.rc.b_stat_read )
1116         h->param.rc.i_lookahead = 0;
1117 #if HAVE_THREAD
1118     if( h->param.i_sync_lookahead < 0 )
1119         h->param.i_sync_lookahead = h->param.i_bframe + 1;
1120     h->param.i_sync_lookahead = X264_MIN( h->param.i_sync_lookahead, X264_LOOKAHEAD_MAX );
1121     if( h->param.rc.b_stat_read || h->i_thread_frames == 1 )
1122         h->param.i_sync_lookahead = 0;
1123 #else
1124     h->param.i_sync_lookahead = 0;
1125 #endif
1126 
1127     h->param.i_deblocking_filter_alphac0 = x264_clip3( h->param.i_deblocking_filter_alphac0, -6, 6 );
1128     h->param.i_deblocking_filter_beta    = x264_clip3( h->param.i_deblocking_filter_beta, -6, 6 );
1129     h->param.analyse.i_luma_deadzone[0] = x264_clip3( h->param.analyse.i_luma_deadzone[0], 0, 32 );
1130     h->param.analyse.i_luma_deadzone[1] = x264_clip3( h->param.analyse.i_luma_deadzone[1], 0, 32 );
1131 
1132     h->param.i_cabac_init_idc = x264_clip3( h->param.i_cabac_init_idc, 0, 2 );
1133 
1134     if( h->param.i_cqm_preset < X264_CQM_FLAT || h->param.i_cqm_preset > X264_CQM_CUSTOM )
1135         h->param.i_cqm_preset = X264_CQM_FLAT;
1136 
1137     if( h->param.analyse.i_me_method < X264_ME_DIA ||
1138         h->param.analyse.i_me_method > X264_ME_TESA )
1139         h->param.analyse.i_me_method = X264_ME_HEX;
1140     h->param.analyse.i_me_range = x264_clip3( h->param.analyse.i_me_range, 4, 1024 );
1141     if( h->param.analyse.i_me_range > 16 && h->param.analyse.i_me_method <= X264_ME_HEX )
1142         h->param.analyse.i_me_range = 16;
1143     if( h->param.analyse.i_me_method == X264_ME_TESA &&
1144         (h->mb.b_lossless || h->param.analyse.i_subpel_refine <= 1) )
1145         h->param.analyse.i_me_method = X264_ME_ESA;
1146     h->param.analyse.b_mixed_references = h->param.analyse.b_mixed_references && h->param.i_frame_reference > 1;
1147     h->param.analyse.inter &= X264_ANALYSE_PSUB16x16|X264_ANALYSE_PSUB8x8|X264_ANALYSE_BSUB16x16|
1148                               X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
1149     h->param.analyse.intra &= X264_ANALYSE_I4x4|X264_ANALYSE_I8x8;
1150     if( !(h->param.analyse.inter & X264_ANALYSE_PSUB16x16) )
1151         h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
1152     if( !h->param.analyse.b_transform_8x8 )
1153     {
1154         h->param.analyse.inter &= ~X264_ANALYSE_I8x8;
1155         h->param.analyse.intra &= ~X264_ANALYSE_I8x8;
1156     }
1157     h->param.analyse.i_trellis = x264_clip3( h->param.analyse.i_trellis, 0, 2 );
1158     h->param.rc.i_aq_mode = x264_clip3( h->param.rc.i_aq_mode, 0, 3 );
1159     h->param.rc.f_aq_strength = x264_clip3f( h->param.rc.f_aq_strength, 0, 3 );
1160     if( h->param.rc.f_aq_strength == 0 )
1161         h->param.rc.i_aq_mode = 0;
1162 
1163     if( h->param.i_log_level < X264_LOG_INFO )
1164     {
1165         h->param.analyse.b_psnr = 0;
1166         h->param.analyse.b_ssim = 0;
1167     }
1168     /* Warn users trying to measure PSNR/SSIM with psy opts on. */
1169     if( b_open && (h->param.analyse.b_psnr || h->param.analyse.b_ssim) )
1170     {
1171         char *s = NULL;
1172 
1173         if( h->param.analyse.b_psy )
1174         {
1175             s = h->param.analyse.b_psnr ? "psnr" : "ssim";
1176             x264_log( h, X264_LOG_WARNING, "--%s used with psy on: results will be invalid!\n", s );
1177         }
1178         else if( !h->param.rc.i_aq_mode && h->param.analyse.b_ssim )
1179         {
1180             x264_log( h, X264_LOG_WARNING, "--ssim used with AQ off: results will be invalid!\n" );
1181             s = "ssim";
1182         }
1183         else if(  h->param.rc.i_aq_mode && h->param.analyse.b_psnr )
1184         {
1185             x264_log( h, X264_LOG_WARNING, "--psnr used with AQ on: results will be invalid!\n" );
1186             s = "psnr";
1187         }
1188         if( s )
1189             x264_log( h, X264_LOG_WARNING, "--tune %s should be used if attempting to benchmark %s!\n", s, s );
1190     }
1191 
1192     if( !h->param.analyse.b_psy )
1193     {
1194         h->param.analyse.f_psy_rd = 0;
1195         h->param.analyse.f_psy_trellis = 0;
1196     }
1197     h->param.analyse.f_psy_rd = x264_clip3f( h->param.analyse.f_psy_rd, 0, 10 );
1198     h->param.analyse.f_psy_trellis = x264_clip3f( h->param.analyse.f_psy_trellis, 0, 10 );
1199     h->mb.i_psy_rd = h->param.analyse.i_subpel_refine >= 6 ? FIX8( h->param.analyse.f_psy_rd ) : 0;
1200     h->mb.i_psy_trellis = h->param.analyse.i_trellis ? FIX8( h->param.analyse.f_psy_trellis / 4 ) : 0;
1201     h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -32, 32);
1202     /* In 4:4:4 mode, chroma gets twice as much resolution, so we can halve its quality. */
1203     if( b_open && i_csp >= X264_CSP_I444 && i_csp < X264_CSP_BGR && h->param.analyse.b_psy )
1204         h->param.analyse.i_chroma_qp_offset += 6;
1205     /* Psy RDO increases overall quantizers to improve the quality of luma--this indirectly hurts chroma quality */
1206     /* so we lower the chroma QP offset to compensate */
1207     if( b_open && h->mb.i_psy_rd && !h->param.i_avcintra_class )
1208         h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_rd < 0.25 ? 1 : 2;
1209     /* Psy trellis has a similar effect. */
1210     if( b_open && h->mb.i_psy_trellis && !h->param.i_avcintra_class )
1211         h->param.analyse.i_chroma_qp_offset -= h->param.analyse.f_psy_trellis < 0.25 ? 1 : 2;
1212     h->param.analyse.i_chroma_qp_offset = x264_clip3(h->param.analyse.i_chroma_qp_offset, -12, 12);
1213     /* MB-tree requires AQ to be on, even if the strength is zero. */
1214     if( !h->param.rc.i_aq_mode && h->param.rc.b_mb_tree )
1215     {
1216         h->param.rc.i_aq_mode = 1;
1217         h->param.rc.f_aq_strength = 0;
1218     }
1219     h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
1220     if( h->param.analyse.i_subpel_refine >= 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
1221         h->param.analyse.i_subpel_refine = 9;
1222 
1223     if( b_open )
1224     {
1225         const x264_level_t *l = x264_levels;
1226         if( h->param.i_level_idc < 0 )
1227         {
1228             int maxrate_bak = h->param.rc.i_vbv_max_bitrate;
1229             if( h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.i_vbv_buffer_size <= 0 )
1230                 h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate * 2;
1231             x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1232             do h->param.i_level_idc = l->level_idc;
1233                 while( l[1].level_idc && x264_validate_levels( h, 0 ) && l++ );
1234             h->param.rc.i_vbv_max_bitrate = maxrate_bak;
1235         }
1236         else
1237         {
1238             while( l->level_idc && l->level_idc != h->param.i_level_idc )
1239                 l++;
1240             if( l->level_idc == 0 )
1241             {
1242                 x264_log( h, X264_LOG_ERROR, "invalid level_idc: %d\n", h->param.i_level_idc );
1243                 return -1;
1244             }
1245         }
1246         if( h->param.analyse.i_mv_range <= 0 )
1247             h->param.analyse.i_mv_range = l->mv_range >> PARAM_INTERLACED;
1248         else
1249             h->param.analyse.i_mv_range = x264_clip3(h->param.analyse.i_mv_range, 32, 8192 >> PARAM_INTERLACED);
1250     }
1251 
1252     h->param.analyse.i_weighted_pred = x264_clip3( h->param.analyse.i_weighted_pred, X264_WEIGHTP_NONE, X264_WEIGHTP_SMART );
1253 
1254     if( h->param.i_lookahead_threads == X264_THREADS_AUTO )
1255     {
1256         if( h->param.b_sliced_threads )
1257             h->param.i_lookahead_threads = h->param.i_threads;
1258         else
1259         {
1260             /* If we're using much slower lookahead settings than encoding settings, it helps a lot to use
1261              * more lookahead threads.  This typically happens in the first pass of a two-pass encode, so
1262              * try to guess at this sort of case.
1263              *
1264              * Tuned by a little bit of real encoding with the various presets. */
1265             int badapt = h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS;
1266             int subme = X264_MIN( h->param.analyse.i_subpel_refine / 3, 3 ) + (h->param.analyse.i_subpel_refine > 1);
1267             int bframes = X264_MIN( (h->param.i_bframe - 1) / 3, 3 );
1268 
1269             /* [b-adapt 0/1 vs 2][quantized subme][quantized bframes] */
1270             static const uint8_t lookahead_thread_div[2][5][4] =
1271             {{{6,6,6,6}, {3,3,3,3}, {4,4,4,4}, {6,6,6,6}, {12,12,12,12}},
1272              {{3,2,1,1}, {2,1,1,1}, {4,3,2,1}, {6,4,3,2}, {12, 9, 6, 4}}};
1273 
1274             h->param.i_lookahead_threads = h->param.i_threads / lookahead_thread_div[badapt][subme][bframes];
1275             /* Since too many lookahead threads significantly degrades lookahead accuracy, limit auto
1276              * lookahead threads to about 8 macroblock rows high each at worst.  This number is chosen
1277              * pretty much arbitrarily. */
1278             h->param.i_lookahead_threads = X264_MIN( h->param.i_lookahead_threads, h->param.i_height / 128 );
1279         }
1280     }
1281     h->param.i_lookahead_threads = x264_clip3( h->param.i_lookahead_threads, 1, X264_MIN( max_sliced_threads, X264_LOOKAHEAD_THREAD_MAX ) );
1282 
1283     if( PARAM_INTERLACED )
1284     {
1285         if( h->param.analyse.i_me_method >= X264_ME_ESA )
1286         {
1287             x264_log( h, X264_LOG_WARNING, "interlace + me=esa is not implemented\n" );
1288             h->param.analyse.i_me_method = X264_ME_UMH;
1289         }
1290         if( h->param.analyse.i_weighted_pred > 0 )
1291         {
1292             x264_log( h, X264_LOG_WARNING, "interlace + weightp is not implemented\n" );
1293             h->param.analyse.i_weighted_pred = X264_WEIGHTP_NONE;
1294         }
1295     }
1296 
1297     if( !h->param.analyse.i_weighted_pred && h->param.rc.b_mb_tree && h->param.analyse.b_psy )
1298         h->param.analyse.i_weighted_pred = X264_WEIGHTP_FAKE;
1299 
1300     if( h->i_thread_frames > 1 )
1301     {
1302         int r = h->param.analyse.i_mv_range_thread;
1303         int r2;
1304         if( r <= 0 )
1305         {
1306             // half of the available space is reserved and divided evenly among the threads,
1307             // the rest is allocated to whichever thread is far enough ahead to use it.
1308             // reserving more space increases quality for some videos, but costs more time
1309             // in thread synchronization.
1310             int max_range = (h->param.i_height + X264_THREAD_HEIGHT) / h->i_thread_frames - X264_THREAD_HEIGHT;
1311             r = max_range / 2;
1312         }
1313         r = X264_MAX( r, h->param.analyse.i_me_range );
1314         r = X264_MIN( r, h->param.analyse.i_mv_range );
1315         // round up to use the whole mb row
1316         r2 = (r & ~15) + ((-X264_THREAD_HEIGHT) & 15);
1317         if( r2 < r )
1318             r2 += 16;
1319         x264_log( h, X264_LOG_DEBUG, "using mv_range_thread = %d\n", r2 );
1320         h->param.analyse.i_mv_range_thread = r2;
1321     }
1322 
1323     if( h->param.rc.f_rate_tolerance < 0 )
1324         h->param.rc.f_rate_tolerance = 0;
1325     if( h->param.rc.f_qblur < 0 )
1326         h->param.rc.f_qblur = 0;
1327     if( h->param.rc.f_complexity_blur < 0 )
1328         h->param.rc.f_complexity_blur = 0;
1329 
1330     h->param.i_sps_id &= 31;
1331 
1332     h->param.i_nal_hrd = x264_clip3( h->param.i_nal_hrd, X264_NAL_HRD_NONE, X264_NAL_HRD_CBR );
1333 
1334     if( h->param.i_nal_hrd && !h->param.rc.i_vbv_buffer_size )
1335     {
1336         x264_log( h, X264_LOG_WARNING, "NAL HRD parameters require VBV parameters\n" );
1337         h->param.i_nal_hrd = X264_NAL_HRD_NONE;
1338     }
1339 
1340     if( h->param.i_nal_hrd == X264_NAL_HRD_CBR &&
1341        (h->param.rc.i_bitrate != h->param.rc.i_vbv_max_bitrate || !h->param.rc.i_vbv_max_bitrate) )
1342     {
1343         x264_log( h, X264_LOG_WARNING, "CBR HRD requires constant bitrate\n" );
1344         h->param.i_nal_hrd = X264_NAL_HRD_VBR;
1345     }
1346 
1347     if( h->param.i_nal_hrd == X264_NAL_HRD_CBR )
1348         h->param.rc.b_filler = 1;
1349 
1350     /* ensure the booleans are 0 or 1 so they can be used in math */
1351 #define BOOLIFY(x) h->param.x = !!h->param.x
1352     BOOLIFY( b_cabac );
1353     BOOLIFY( b_constrained_intra );
1354     BOOLIFY( b_deblocking_filter );
1355     BOOLIFY( b_deterministic );
1356     BOOLIFY( b_sliced_threads );
1357     BOOLIFY( b_interlaced );
1358     BOOLIFY( b_intra_refresh );
1359     BOOLIFY( b_aud );
1360     BOOLIFY( b_repeat_headers );
1361     BOOLIFY( b_annexb );
1362     BOOLIFY( b_vfr_input );
1363     BOOLIFY( b_pulldown );
1364     BOOLIFY( b_tff );
1365     BOOLIFY( b_pic_struct );
1366     BOOLIFY( b_fake_interlaced );
1367     BOOLIFY( b_open_gop );
1368     BOOLIFY( b_bluray_compat );
1369     BOOLIFY( b_stitchable );
1370     BOOLIFY( b_full_recon );
1371     BOOLIFY( b_opencl );
1372     BOOLIFY( analyse.b_transform_8x8 );
1373     BOOLIFY( analyse.b_weighted_bipred );
1374     BOOLIFY( analyse.b_chroma_me );
1375     BOOLIFY( analyse.b_mixed_references );
1376     BOOLIFY( analyse.b_fast_pskip );
1377     BOOLIFY( analyse.b_dct_decimate );
1378     BOOLIFY( analyse.b_psy );
1379     BOOLIFY( analyse.b_psnr );
1380     BOOLIFY( analyse.b_ssim );
1381     BOOLIFY( rc.b_stat_write );
1382     BOOLIFY( rc.b_stat_read );
1383     BOOLIFY( rc.b_mb_tree );
1384     BOOLIFY( rc.b_filler );
1385 #undef BOOLIFY
1386 
1387     return 0;
1388 }
1389 
mbcmp_init(x264_t * h)1390 static void mbcmp_init( x264_t *h )
1391 {
1392     int satd = !h->mb.b_lossless && h->param.analyse.i_subpel_refine > 1;
1393     memcpy( h->pixf.mbcmp, satd ? h->pixf.satd : h->pixf.sad_aligned, sizeof(h->pixf.mbcmp) );
1394     memcpy( h->pixf.mbcmp_unaligned, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.mbcmp_unaligned) );
1395     h->pixf.intra_mbcmp_x3_16x16 = satd ? h->pixf.intra_satd_x3_16x16 : h->pixf.intra_sad_x3_16x16;
1396     h->pixf.intra_mbcmp_x3_8x16c = satd ? h->pixf.intra_satd_x3_8x16c : h->pixf.intra_sad_x3_8x16c;
1397     h->pixf.intra_mbcmp_x3_8x8c  = satd ? h->pixf.intra_satd_x3_8x8c  : h->pixf.intra_sad_x3_8x8c;
1398     h->pixf.intra_mbcmp_x3_8x8 = satd ? h->pixf.intra_sa8d_x3_8x8 : h->pixf.intra_sad_x3_8x8;
1399     h->pixf.intra_mbcmp_x3_4x4 = satd ? h->pixf.intra_satd_x3_4x4 : h->pixf.intra_sad_x3_4x4;
1400     h->pixf.intra_mbcmp_x9_4x4 = h->param.b_cpu_independent || h->mb.b_lossless ? NULL
1401                                : satd ? h->pixf.intra_satd_x9_4x4 : h->pixf.intra_sad_x9_4x4;
1402     h->pixf.intra_mbcmp_x9_8x8 = h->param.b_cpu_independent || h->mb.b_lossless ? NULL
1403                                : satd ? h->pixf.intra_sa8d_x9_8x8 : h->pixf.intra_sad_x9_8x8;
1404     satd &= h->param.analyse.i_me_method == X264_ME_TESA;
1405     memcpy( h->pixf.fpelcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.fpelcmp) );
1406     memcpy( h->pixf.fpelcmp_x3, satd ? h->pixf.satd_x3 : h->pixf.sad_x3, sizeof(h->pixf.fpelcmp_x3) );
1407     memcpy( h->pixf.fpelcmp_x4, satd ? h->pixf.satd_x4 : h->pixf.sad_x4, sizeof(h->pixf.fpelcmp_x4) );
1408 }
1409 
chroma_dsp_init(x264_t * h)1410 static void chroma_dsp_init( x264_t *h )
1411 {
1412     memcpy( h->luma2chroma_pixel, x264_luma2chroma_pixel[CHROMA_FORMAT], sizeof(h->luma2chroma_pixel) );
1413 
1414     switch( CHROMA_FORMAT )
1415     {
1416         case CHROMA_400:
1417             h->mc.prefetch_fenc = h->mc.prefetch_fenc_400;
1418             break;
1419         case CHROMA_420:
1420             memcpy( h->predict_chroma, h->predict_8x8c, sizeof(h->predict_chroma) );
1421             h->mc.prefetch_fenc = h->mc.prefetch_fenc_420;
1422             h->loopf.deblock_chroma[0] = h->loopf.deblock_h_chroma_420;
1423             h->loopf.deblock_chroma_intra[0] = h->loopf.deblock_h_chroma_420_intra;
1424             h->loopf.deblock_chroma_mbaff = h->loopf.deblock_chroma_420_mbaff;
1425             h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_chroma_420_intra_mbaff;
1426             h->pixf.intra_mbcmp_x3_chroma = h->pixf.intra_mbcmp_x3_8x8c;
1427             h->quantf.coeff_last[DCT_CHROMA_DC] = h->quantf.coeff_last4;
1428             h->quantf.coeff_level_run[DCT_CHROMA_DC] = h->quantf.coeff_level_run4;
1429             break;
1430         case CHROMA_422:
1431             memcpy( h->predict_chroma, h->predict_8x16c, sizeof(h->predict_chroma) );
1432             h->mc.prefetch_fenc = h->mc.prefetch_fenc_422;
1433             h->loopf.deblock_chroma[0] = h->loopf.deblock_h_chroma_422;
1434             h->loopf.deblock_chroma_intra[0] = h->loopf.deblock_h_chroma_422_intra;
1435             h->loopf.deblock_chroma_mbaff = h->loopf.deblock_chroma_422_mbaff;
1436             h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_chroma_422_intra_mbaff;
1437             h->pixf.intra_mbcmp_x3_chroma = h->pixf.intra_mbcmp_x3_8x16c;
1438             h->quantf.coeff_last[DCT_CHROMA_DC] = h->quantf.coeff_last8;
1439             h->quantf.coeff_level_run[DCT_CHROMA_DC] = h->quantf.coeff_level_run8;
1440             break;
1441         case CHROMA_444:
1442             h->mc.prefetch_fenc = h->mc.prefetch_fenc_422; /* FIXME: doesn't cover V plane */
1443             h->loopf.deblock_chroma_mbaff = h->loopf.deblock_luma_mbaff;
1444             h->loopf.deblock_chroma_intra_mbaff = h->loopf.deblock_luma_intra_mbaff;
1445             break;
1446     }
1447 }
1448 
set_aspect_ratio(x264_t * h,x264_param_t * param,int initial)1449 static void set_aspect_ratio( x264_t *h, x264_param_t *param, int initial )
1450 {
1451     /* VUI */
1452     if( param->vui.i_sar_width > 0 && param->vui.i_sar_height > 0 )
1453     {
1454         uint32_t i_w = param->vui.i_sar_width;
1455         uint32_t i_h = param->vui.i_sar_height;
1456         uint32_t old_w = h->param.vui.i_sar_width;
1457         uint32_t old_h = h->param.vui.i_sar_height;
1458 
1459         x264_reduce_fraction( &i_w, &i_h );
1460 
1461         while( i_w > 65535 || i_h > 65535 )
1462         {
1463             i_w /= 2;
1464             i_h /= 2;
1465         }
1466 
1467         x264_reduce_fraction( &i_w, &i_h );
1468 
1469         if( i_w != old_w || i_h != old_h || initial )
1470         {
1471             h->param.vui.i_sar_width = 0;
1472             h->param.vui.i_sar_height = 0;
1473             if( i_w == 0 || i_h == 0 )
1474                 x264_log( h, X264_LOG_WARNING, "cannot create valid sample aspect ratio\n" );
1475             else
1476             {
1477                 x264_log( h, initial?X264_LOG_INFO:X264_LOG_DEBUG, "using SAR=%d/%d\n", i_w, i_h );
1478                 h->param.vui.i_sar_width = i_w;
1479                 h->param.vui.i_sar_height = i_h;
1480             }
1481         }
1482     }
1483 }
1484 
1485 /****************************************************************************
1486  * x264_encoder_open:
1487  ****************************************************************************/
x264_encoder_open(x264_param_t * param,void * api)1488 x264_t *x264_encoder_open( x264_param_t *param, void *api )
1489 {
1490     x264_t *h;
1491     char buf[1000], *p;
1492     int i_slicetype_length;
1493 
1494     CHECKED_MALLOCZERO( h, sizeof(x264_t) );
1495 
1496     /* Create a copy of param */
1497     memcpy( &h->param, param, sizeof(x264_param_t) );
1498     h->param.opaque = NULL;
1499     h->param.param_free = NULL;
1500 
1501     if( h->param.psz_cqm_file )
1502         CHECKED_PARAM_STRDUP( h->param.psz_cqm_file, &h->param, h->param.psz_cqm_file );
1503     if( h->param.psz_dump_yuv )
1504         CHECKED_PARAM_STRDUP( h->param.psz_dump_yuv, &h->param, h->param.psz_dump_yuv );
1505     if( h->param.rc.psz_stat_out )
1506         CHECKED_PARAM_STRDUP( h->param.rc.psz_stat_out, &h->param, h->param.rc.psz_stat_out );
1507     if( h->param.rc.psz_stat_in )
1508         CHECKED_PARAM_STRDUP( h->param.rc.psz_stat_in, &h->param, h->param.rc.psz_stat_in );
1509     if( h->param.rc.psz_zones )
1510         CHECKED_PARAM_STRDUP( h->param.rc.psz_zones, &h->param, h->param.rc.psz_zones );
1511     if( h->param.psz_clbin_file )
1512         CHECKED_PARAM_STRDUP( h->param.psz_clbin_file, &h->param, h->param.psz_clbin_file );
1513 
1514     if( param->param_free )
1515     {
1516         x264_param_cleanup( param );
1517         param->param_free( param );
1518     }
1519 
1520     /* Save pointer to bit depth independent interface */
1521     h->api = api;
1522 
1523 #if HAVE_INTEL_DISPATCHER
1524     x264_intel_dispatcher_override();
1525 #endif
1526 
1527     if( x264_threading_init() )
1528     {
1529         x264_log( h, X264_LOG_ERROR, "unable to initialize threading\n" );
1530         goto fail;
1531     }
1532 
1533     if( validate_parameters( h, 1 ) < 0 )
1534         goto fail;
1535 
1536     if( h->param.psz_cqm_file )
1537         if( x264_cqm_parse_file( h, h->param.psz_cqm_file ) < 0 )
1538             goto fail;
1539 
1540     x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den );
1541     x264_reduce_fraction( &h->param.i_timebase_num, &h->param.i_timebase_den );
1542 
1543     /* Init x264_t */
1544     h->i_frame = -1;
1545     h->i_frame_num = 0;
1546 
1547     if( h->param.i_avcintra_class )
1548         h->i_idr_pic_id = 5;
1549     else
1550         h->i_idr_pic_id = 0;
1551 
1552     if( (uint64_t)h->param.i_timebase_den * 2 > UINT32_MAX )
1553     {
1554         x264_log( h, X264_LOG_ERROR, "Effective timebase denominator %u exceeds H.264 maximum\n", h->param.i_timebase_den );
1555         goto fail;
1556     }
1557 
1558     set_aspect_ratio( h, &h->param, 1 );
1559 
1560     x264_sps_init( h->sps, h->param.i_sps_id, &h->param );
1561     x264_sps_init_scaling_list( h->sps, &h->param );
1562     x264_pps_init( h->pps, h->param.i_sps_id, &h->param, h->sps );
1563 
1564     x264_validate_levels( h, 1 );
1565 
1566     h->chroma_qp_table = i_chroma_qp_table + 12 + h->pps->i_chroma_qp_index_offset;
1567 
1568     if( x264_cqm_init( h ) < 0 )
1569         goto fail;
1570 
1571     h->mb.i_mb_width = h->sps->i_mb_width;
1572     h->mb.i_mb_height = h->sps->i_mb_height;
1573     h->mb.i_mb_count = h->mb.i_mb_width * h->mb.i_mb_height;
1574 
1575     h->mb.chroma_h_shift = CHROMA_FORMAT == CHROMA_420 || CHROMA_FORMAT == CHROMA_422;
1576     h->mb.chroma_v_shift = CHROMA_FORMAT == CHROMA_420;
1577 
1578     /* Adaptive MBAFF and subme 0 are not supported as we require halving motion
1579      * vectors during prediction, resulting in hpel mvs.
1580      * The chosen solution is to make MBAFF non-adaptive in this case. */
1581     h->mb.b_adaptive_mbaff = PARAM_INTERLACED && h->param.analyse.i_subpel_refine;
1582 
1583     /* Init frames. */
1584     if( h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS && !h->param.rc.b_stat_read )
1585         h->frames.i_delay = X264_MAX(h->param.i_bframe,3)*4;
1586     else
1587         h->frames.i_delay = h->param.i_bframe;
1588     if( h->param.rc.b_mb_tree || h->param.rc.i_vbv_buffer_size )
1589         h->frames.i_delay = X264_MAX( h->frames.i_delay, h->param.rc.i_lookahead );
1590     i_slicetype_length = h->frames.i_delay;
1591     h->frames.i_delay += h->i_thread_frames - 1;
1592     h->frames.i_delay += h->param.i_sync_lookahead;
1593     h->frames.i_delay += h->param.b_vfr_input;
1594     h->frames.i_bframe_delay = h->param.i_bframe ? (h->param.i_bframe_pyramid ? 2 : 1) : 0;
1595 
1596     h->frames.i_max_ref0 = h->param.i_frame_reference;
1597     h->frames.i_max_ref1 = X264_MIN( h->sps->vui.i_num_reorder_frames, h->param.i_frame_reference );
1598     h->frames.i_max_dpb  = h->sps->vui.i_max_dec_frame_buffering;
1599     h->frames.b_have_lowres = !h->param.rc.b_stat_read
1600         && ( h->param.rc.i_rc_method == X264_RC_ABR
1601           || h->param.rc.i_rc_method == X264_RC_CRF
1602           || h->param.i_bframe_adaptive
1603           || h->param.i_scenecut_threshold
1604           || h->param.rc.b_mb_tree
1605           || h->param.analyse.i_weighted_pred );
1606     h->frames.b_have_lowres |= h->param.rc.b_stat_read && h->param.rc.i_vbv_buffer_size > 0;
1607     h->frames.b_have_sub8x8_esa = !!(h->param.analyse.inter & X264_ANALYSE_PSUB8x8);
1608 
1609     h->frames.i_last_idr =
1610     h->frames.i_last_keyframe = - h->param.i_keyint_max;
1611     h->frames.i_input    = 0;
1612     h->frames.i_largest_pts = h->frames.i_second_largest_pts = -1;
1613     h->frames.i_poc_last_open_gop = -1;
1614 
1615     CHECKED_MALLOCZERO( h->cost_table, sizeof(*h->cost_table) );
1616     CHECKED_MALLOCZERO( h->frames.unused[0], (h->frames.i_delay + 3) * sizeof(x264_frame_t *) );
1617     /* Allocate room for max refs plus a few extra just in case. */
1618     CHECKED_MALLOCZERO( h->frames.unused[1], (h->i_thread_frames + X264_REF_MAX + 4) * sizeof(x264_frame_t *) );
1619     CHECKED_MALLOCZERO( h->frames.current, (h->param.i_sync_lookahead + h->param.i_bframe
1620                         + h->i_thread_frames + 3) * sizeof(x264_frame_t *) );
1621     if( h->param.analyse.i_weighted_pred > 0 )
1622         CHECKED_MALLOCZERO( h->frames.blank_unused, h->i_thread_frames * 4 * sizeof(x264_frame_t *) );
1623     h->i_ref[0] = h->i_ref[1] = 0;
1624     h->i_cpb_delay = h->i_coded_fields = h->i_disp_fields = 0;
1625     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);
1626     h->i_disp_fields_last_frame = -1;
1627     x264_rdo_init();
1628 
1629     /* init CPU functions */
1630 #if (ARCH_X86 || ARCH_X86_64) && HIGH_BIT_DEPTH
1631     /* FIXME: Only 8-bit has been optimized for AVX-512 so far. The few AVX-512 functions
1632      * enabled in high bit-depth are insignificant and just causes potential issues with
1633      * unnecessary thermal throttling and whatnot, so keep it disabled for now. */
1634     h->param.cpu &= ~X264_CPU_AVX512;
1635 #endif
1636     x264_predict_16x16_init( h->param.cpu, h->predict_16x16 );
1637     x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c );
1638     x264_predict_8x16c_init( h->param.cpu, h->predict_8x16c );
1639     x264_predict_8x8_init( h->param.cpu, h->predict_8x8, &h->predict_8x8_filter );
1640     x264_predict_4x4_init( h->param.cpu, h->predict_4x4 );
1641     x264_pixel_init( h->param.cpu, &h->pixf );
1642     x264_dct_init( h->param.cpu, &h->dctf );
1643     x264_zigzag_init( h->param.cpu, &h->zigzagf_progressive, &h->zigzagf_interlaced );
1644     memcpy( &h->zigzagf, PARAM_INTERLACED ? &h->zigzagf_interlaced : &h->zigzagf_progressive, sizeof(h->zigzagf) );
1645     x264_mc_init( h->param.cpu, &h->mc, h->param.b_cpu_independent );
1646     x264_quant_init( h, h->param.cpu, &h->quantf );
1647     x264_deblock_init( h->param.cpu, &h->loopf, PARAM_INTERLACED );
1648     x264_bitstream_init( h->param.cpu, &h->bsf );
1649     if( h->param.b_cabac )
1650         x264_cabac_init( h );
1651     else
1652         x264_cavlc_init( h );
1653 
1654     mbcmp_init( h );
1655     chroma_dsp_init( h );
1656 
1657     p = buf + sprintf( buf, "using cpu capabilities:" );
1658     for( int i = 0; x264_cpu_names[i].flags; i++ )
1659     {
1660         if( !strcmp(x264_cpu_names[i].name, "SSE")
1661             && h->param.cpu & (X264_CPU_SSE2) )
1662             continue;
1663         if( !strcmp(x264_cpu_names[i].name, "SSE2")
1664             && h->param.cpu & (X264_CPU_SSE2_IS_FAST|X264_CPU_SSE2_IS_SLOW) )
1665             continue;
1666         if( !strcmp(x264_cpu_names[i].name, "SSE3")
1667             && (h->param.cpu & X264_CPU_SSSE3 || !(h->param.cpu & X264_CPU_CACHELINE_64)) )
1668             continue;
1669         if( !strcmp(x264_cpu_names[i].name, "SSE4.1")
1670             && (h->param.cpu & X264_CPU_SSE42) )
1671             continue;
1672         if( !strcmp(x264_cpu_names[i].name, "LZCNT")
1673             && (h->param.cpu & X264_CPU_BMI1) )
1674             continue;
1675         if( !strcmp(x264_cpu_names[i].name, "BMI1")
1676             && (h->param.cpu & X264_CPU_BMI2) )
1677             continue;
1678         if( !strcmp(x264_cpu_names[i].name, "FMA4")
1679             && (h->param.cpu & X264_CPU_FMA3) )
1680             continue;
1681         if( (h->param.cpu & x264_cpu_names[i].flags) == x264_cpu_names[i].flags
1682             && (!i || x264_cpu_names[i].flags != x264_cpu_names[i-1].flags) )
1683             p += sprintf( p, " %s", x264_cpu_names[i].name );
1684     }
1685     if( !h->param.cpu )
1686         p += sprintf( p, " none!" );
1687     x264_log( h, X264_LOG_INFO, "%s\n", buf );
1688 
1689     if( x264_analyse_init_costs( h ) )
1690         goto fail;
1691 
1692     /* Must be volatile or else GCC will optimize it out. */
1693     volatile int temp = 392;
1694     if( x264_clz( temp ) != 23 )
1695     {
1696         x264_log( h, X264_LOG_ERROR, "CLZ test failed: x264 has been miscompiled!\n" );
1697 #if ARCH_X86 || ARCH_X86_64
1698         x264_log( h, X264_LOG_ERROR, "Are you attempting to run an SSE4a/LZCNT-targeted build on a CPU that\n" );
1699         x264_log( h, X264_LOG_ERROR, "doesn't support it?\n" );
1700 #endif
1701         goto fail;
1702     }
1703 
1704     h->out.i_nal = 0;
1705     h->out.i_bitstream = x264_clip3f(
1706         h->param.i_width * h->param.i_height * 4
1707         * ( h->param.rc.i_rc_method == X264_RC_ABR
1708             ? pow( 0.95, h->param.rc.i_qp_min )
1709             : pow( 0.95, h->param.rc.i_qp_constant ) * X264_MAX( 1, h->param.rc.f_ip_factor ) ),
1710         1000000, INT_MAX/3
1711     );
1712 
1713     h->nal_buffer_size = h->out.i_bitstream * 3/2 + 4 + 64; /* +4 for startcode, +64 for nal_escape assembly padding */
1714     CHECKED_MALLOC( h->nal_buffer, h->nal_buffer_size );
1715 
1716     CHECKED_MALLOC( h->reconfig_h, sizeof(x264_t) );
1717 
1718     if( h->param.i_threads > 1 &&
1719         x264_threadpool_init( &h->threadpool, h->param.i_threads, (void*)encoder_thread_init, h ) )
1720         goto fail;
1721     if( h->param.i_lookahead_threads > 1 &&
1722         x264_threadpool_init( &h->lookaheadpool, h->param.i_lookahead_threads, NULL, NULL ) )
1723         goto fail;
1724 
1725 #if HAVE_OPENCL
1726     if( h->param.b_opencl )
1727     {
1728         h->opencl.ocl = x264_opencl_load_library();
1729         if( !h->opencl.ocl )
1730         {
1731             x264_log( h, X264_LOG_WARNING, "failed to load OpenCL\n" );
1732             h->param.b_opencl = 0;
1733         }
1734     }
1735 #endif
1736 
1737     h->thread[0] = h;
1738     for( int i = 1; i < h->param.i_threads + !!h->param.i_sync_lookahead; i++ )
1739         CHECKED_MALLOC( h->thread[i], sizeof(x264_t) );
1740     if( h->param.i_lookahead_threads > 1 )
1741         for( int i = 0; i < h->param.i_lookahead_threads; i++ )
1742         {
1743             CHECKED_MALLOC( h->lookahead_thread[i], sizeof(x264_t) );
1744             *h->lookahead_thread[i] = *h;
1745         }
1746     *h->reconfig_h = *h;
1747 
1748     for( int i = 0; i < h->param.i_threads; i++ )
1749     {
1750         int init_nal_count = h->param.i_slice_count + 3;
1751         int allocate_threadlocal_data = !h->param.b_sliced_threads || !i;
1752         if( i > 0 )
1753             *h->thread[i] = *h;
1754 
1755         if( x264_pthread_mutex_init( &h->thread[i]->mutex, NULL ) )
1756             goto fail;
1757         if( x264_pthread_cond_init( &h->thread[i]->cv, NULL ) )
1758             goto fail;
1759 
1760         if( allocate_threadlocal_data )
1761         {
1762             h->thread[i]->fdec = x264_frame_pop_unused( h, 1 );
1763             if( !h->thread[i]->fdec )
1764                 goto fail;
1765         }
1766         else
1767             h->thread[i]->fdec = h->thread[0]->fdec;
1768 
1769         CHECKED_MALLOC( h->thread[i]->out.p_bitstream, h->out.i_bitstream );
1770         /* Start each thread with room for init_nal_count NAL units; it'll realloc later if needed. */
1771         CHECKED_MALLOC( h->thread[i]->out.nal, init_nal_count*sizeof(x264_nal_t) );
1772         h->thread[i]->out.i_nals_allocated = init_nal_count;
1773 
1774         if( allocate_threadlocal_data && x264_macroblock_cache_allocate( h->thread[i] ) < 0 )
1775             goto fail;
1776     }
1777 
1778 #if HAVE_OPENCL
1779     if( h->param.b_opencl && x264_opencl_lookahead_init( h ) < 0 )
1780         h->param.b_opencl = 0;
1781 #endif
1782 
1783     if( x264_lookahead_init( h, i_slicetype_length ) )
1784         goto fail;
1785 
1786     for( int i = 0; i < h->param.i_threads; i++ )
1787         if( x264_macroblock_thread_allocate( h->thread[i], 0 ) < 0 )
1788             goto fail;
1789 
1790     if( x264_ratecontrol_new( h ) < 0 )
1791         goto fail;
1792 
1793     if( h->param.i_nal_hrd )
1794     {
1795         x264_log( h, X264_LOG_DEBUG, "HRD bitrate: %i bits/sec\n", h->sps->vui.hrd.i_bit_rate_unscaled );
1796         x264_log( h, X264_LOG_DEBUG, "CPB size: %i bits\n", h->sps->vui.hrd.i_cpb_size_unscaled );
1797     }
1798 
1799     if( h->param.psz_dump_yuv )
1800     {
1801         /* create or truncate the reconstructed video file */
1802         FILE *f = x264_fopen( h->param.psz_dump_yuv, "w" );
1803         if( !f )
1804         {
1805             x264_log( h, X264_LOG_ERROR, "dump_yuv: can't write to %s\n", h->param.psz_dump_yuv );
1806             goto fail;
1807         }
1808         else if( !x264_is_regular_file( f ) )
1809         {
1810             x264_log( h, X264_LOG_ERROR, "dump_yuv: incompatible with non-regular file %s\n", h->param.psz_dump_yuv );
1811             fclose( f );
1812             goto fail;
1813         }
1814         fclose( f );
1815     }
1816 
1817     const char *profile = h->sps->i_profile_idc == PROFILE_BASELINE ? "Constrained Baseline" :
1818                           h->sps->i_profile_idc == PROFILE_MAIN ? "Main" :
1819                           h->sps->i_profile_idc == PROFILE_HIGH ? "High" :
1820                           h->sps->i_profile_idc == PROFILE_HIGH10 ?
1821                               (h->sps->b_constraint_set3 ? "High 10 Intra" : "High 10") :
1822                           h->sps->i_profile_idc == PROFILE_HIGH422 ?
1823                               (h->sps->b_constraint_set3 ? "High 4:2:2 Intra" : "High 4:2:2") :
1824                           h->sps->b_constraint_set3 ? "High 4:4:4 Intra" : "High 4:4:4 Predictive";
1825     char level[16];
1826     if( h->sps->i_level_idc == 9 || ( h->sps->i_level_idc == 11 && h->sps->b_constraint_set3 &&
1827         (h->sps->i_profile_idc == PROFILE_BASELINE || h->sps->i_profile_idc == PROFILE_MAIN) ) )
1828         strcpy( level, "1b" );
1829     else
1830         snprintf( level, sizeof(level), "%d.%d", h->sps->i_level_idc / 10, h->sps->i_level_idc % 10 );
1831 
1832     static const char * const subsampling[4] = { "4:0:0", "4:2:0", "4:2:2", "4:4:4" };
1833     x264_log( h, X264_LOG_INFO, "profile %s, level %s, %s, %d-bit\n",
1834               profile, level, subsampling[CHROMA_FORMAT], BIT_DEPTH );
1835 
1836     return h;
1837 fail:
1838     x264_free( h );
1839     return NULL;
1840 }
1841 
1842 /****************************************************************************/
encoder_try_reconfig(x264_t * h,x264_param_t * param,int * rc_reconfig)1843 static int encoder_try_reconfig( x264_t *h, x264_param_t *param, int *rc_reconfig )
1844 {
1845     *rc_reconfig = 0;
1846     set_aspect_ratio( h, param, 0 );
1847 #define COPY(var) h->param.var = param->var
1848     COPY( i_frame_reference ); // but never uses more refs than initially specified
1849     COPY( i_bframe_bias );
1850     if( h->param.i_scenecut_threshold )
1851         COPY( i_scenecut_threshold ); // can't turn it on or off, only vary the threshold
1852     COPY( b_deblocking_filter );
1853     COPY( i_deblocking_filter_alphac0 );
1854     COPY( i_deblocking_filter_beta );
1855     COPY( i_frame_packing );
1856     COPY( mastering_display );
1857     COPY( content_light_level );
1858     COPY( i_alternative_transfer );
1859     COPY( analyse.inter );
1860     COPY( analyse.intra );
1861     COPY( analyse.i_direct_mv_pred );
1862     /* Scratch buffer prevents me_range from being increased for esa/tesa */
1863     if( h->param.analyse.i_me_method < X264_ME_ESA || param->analyse.i_me_range < h->param.analyse.i_me_range )
1864         COPY( analyse.i_me_range );
1865     COPY( analyse.i_noise_reduction );
1866     /* We can't switch out of subme=0 during encoding. */
1867     if( h->param.analyse.i_subpel_refine )
1868         COPY( analyse.i_subpel_refine );
1869     COPY( analyse.i_trellis );
1870     COPY( analyse.b_chroma_me );
1871     COPY( analyse.b_dct_decimate );
1872     COPY( analyse.b_fast_pskip );
1873     COPY( analyse.b_mixed_references );
1874     COPY( analyse.f_psy_rd );
1875     COPY( analyse.f_psy_trellis );
1876     COPY( crop_rect );
1877     // can only twiddle these if they were enabled to begin with:
1878     if( h->param.analyse.i_me_method >= X264_ME_ESA || param->analyse.i_me_method < X264_ME_ESA )
1879         COPY( analyse.i_me_method );
1880     if( h->param.analyse.i_me_method >= X264_ME_ESA && !h->frames.b_have_sub8x8_esa )
1881         h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
1882     if( h->pps->b_transform_8x8_mode )
1883         COPY( analyse.b_transform_8x8 );
1884     if( h->frames.i_max_ref1 > 1 )
1885         COPY( i_bframe_pyramid );
1886     COPY( i_slice_max_size );
1887     COPY( i_slice_max_mbs );
1888     COPY( i_slice_min_mbs );
1889     COPY( i_slice_count );
1890     COPY( i_slice_count_max );
1891     COPY( b_tff );
1892 
1893     /* VBV can't be turned on if it wasn't on to begin with */
1894     if( h->param.rc.i_vbv_max_bitrate > 0 && h->param.rc.i_vbv_buffer_size > 0 &&
1895           param->rc.i_vbv_max_bitrate > 0 &&   param->rc.i_vbv_buffer_size > 0 )
1896     {
1897         *rc_reconfig |= h->param.rc.i_vbv_max_bitrate != param->rc.i_vbv_max_bitrate;
1898         *rc_reconfig |= h->param.rc.i_vbv_buffer_size != param->rc.i_vbv_buffer_size;
1899         *rc_reconfig |= h->param.rc.i_bitrate != param->rc.i_bitrate;
1900         COPY( rc.i_vbv_max_bitrate );
1901         COPY( rc.i_vbv_buffer_size );
1902         COPY( rc.i_bitrate );
1903     }
1904     *rc_reconfig |= h->param.rc.f_rf_constant != param->rc.f_rf_constant;
1905     *rc_reconfig |= h->param.rc.f_rf_constant_max != param->rc.f_rf_constant_max;
1906     COPY( rc.f_rf_constant );
1907     COPY( rc.f_rf_constant_max );
1908 #undef COPY
1909 
1910     return validate_parameters( h, 0 );
1911 }
1912 
x264_encoder_reconfig_apply(x264_t * h,x264_param_t * param)1913 int x264_encoder_reconfig_apply( x264_t *h, x264_param_t *param )
1914 {
1915     int rc_reconfig;
1916     int ret = encoder_try_reconfig( h, param, &rc_reconfig );
1917 
1918     mbcmp_init( h );
1919     if( !ret )
1920         x264_sps_init_reconfigurable( h->sps, &h->param );
1921 
1922     /* Supported reconfiguration options (1-pass only):
1923      * vbv-maxrate
1924      * vbv-bufsize
1925      * crf
1926      * bitrate (CBR only) */
1927     if( !ret && rc_reconfig )
1928         x264_ratecontrol_init_reconfigurable( h, 0 );
1929 
1930     return ret;
1931 }
1932 
1933 /****************************************************************************
1934  * x264_encoder_reconfig:
1935  ****************************************************************************/
x264_encoder_reconfig(x264_t * h,x264_param_t * param)1936 int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
1937 {
1938     h = h->thread[h->thread[0]->i_thread_phase];
1939     x264_param_t param_save = h->reconfig_h->param;
1940     h->reconfig_h->param = h->param;
1941 
1942     int rc_reconfig;
1943     int ret = encoder_try_reconfig( h->reconfig_h, param, &rc_reconfig );
1944     if( !ret )
1945         h->reconfig = 1;
1946     else
1947         h->reconfig_h->param = param_save;
1948 
1949     return ret;
1950 }
1951 
1952 /****************************************************************************
1953  * x264_encoder_parameters:
1954  ****************************************************************************/
x264_encoder_parameters(x264_t * h,x264_param_t * param)1955 void x264_encoder_parameters( x264_t *h, x264_param_t *param )
1956 {
1957     memcpy( param, &h->thread[h->i_thread_phase]->param, sizeof(x264_param_t) );
1958     param->opaque = NULL;
1959 }
1960 
1961 /* internal usage */
nal_start(x264_t * h,int i_type,int i_ref_idc)1962 static void nal_start( x264_t *h, int i_type, int i_ref_idc )
1963 {
1964     x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1965 
1966     nal->i_ref_idc        = i_ref_idc;
1967     nal->i_type           = i_type;
1968     nal->b_long_startcode = 1;
1969 
1970     nal->i_payload= 0;
1971     nal->p_payload= &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1972     nal->i_padding= 0;
1973 }
1974 
1975 /* if number of allocated nals is not enough, re-allocate a larger one. */
nal_check_buffer(x264_t * h)1976 static int nal_check_buffer( x264_t *h )
1977 {
1978     if( h->out.i_nal >= h->out.i_nals_allocated )
1979     {
1980         x264_nal_t *new_out = x264_malloc( sizeof(x264_nal_t) * (h->out.i_nals_allocated*2) );
1981         if( !new_out )
1982             return -1;
1983         memcpy( new_out, h->out.nal, sizeof(x264_nal_t) * (h->out.i_nals_allocated) );
1984         x264_free( h->out.nal );
1985         h->out.nal = new_out;
1986         h->out.i_nals_allocated *= 2;
1987     }
1988     return 0;
1989 }
1990 
nal_end(x264_t * h)1991 static int nal_end( x264_t *h )
1992 {
1993     x264_nal_t *nal = &h->out.nal[h->out.i_nal];
1994     uint8_t *end = &h->out.p_bitstream[bs_pos( &h->out.bs ) / 8];
1995     nal->i_payload = end - nal->p_payload;
1996     /* Assembly implementation of nal_escape reads past the end of the input.
1997      * While undefined padding wouldn't actually affect the output, it makes valgrind unhappy. */
1998     memset( end, 0xff, 64 );
1999     if( h->param.nalu_process )
2000         h->param.nalu_process( (x264_t *)h->api, nal, h->fenc->opaque );
2001     h->out.i_nal++;
2002 
2003     return nal_check_buffer( h );
2004 }
2005 
check_encapsulated_buffer(x264_t * h,x264_t * h0,int start,int64_t previous_nal_size,int64_t necessary_size)2006 static int check_encapsulated_buffer( x264_t *h, x264_t *h0, int start,
2007                                       int64_t previous_nal_size, int64_t necessary_size )
2008 {
2009     if( h0->nal_buffer_size < necessary_size )
2010     {
2011         necessary_size *= 2;
2012         if( necessary_size > INT_MAX )
2013             return -1;
2014         uint8_t *buf = x264_malloc( necessary_size );
2015         if( !buf )
2016             return -1;
2017         if( previous_nal_size )
2018             memcpy( buf, h0->nal_buffer, previous_nal_size );
2019 
2020         intptr_t delta = buf - h0->nal_buffer;
2021         for( int i = 0; i < start; i++ )
2022             h->out.nal[i].p_payload += delta;
2023 
2024         x264_free( h0->nal_buffer );
2025         h0->nal_buffer = buf;
2026         h0->nal_buffer_size = necessary_size;
2027     }
2028 
2029     return 0;
2030 }
2031 
encoder_encapsulate_nals(x264_t * h,int start)2032 static int encoder_encapsulate_nals( x264_t *h, int start )
2033 {
2034     x264_t *h0 = h->thread[0];
2035     int64_t nal_size = 0, previous_nal_size = 0;
2036 
2037     if( h->param.nalu_process )
2038     {
2039         for( int i = start; i < h->out.i_nal; i++ )
2040             nal_size += h->out.nal[i].i_payload;
2041         if( nal_size > INT_MAX )
2042             return -1;
2043         return nal_size;
2044     }
2045 
2046     for( int i = 0; i < start; i++ )
2047         previous_nal_size += h->out.nal[i].i_payload;
2048 
2049     for( int i = start; i < h->out.i_nal; i++ )
2050         nal_size += h->out.nal[i].i_payload;
2051 
2052     /* Worst-case NAL unit escaping: reallocate the buffer if it's too small. */
2053     int64_t necessary_size = previous_nal_size + nal_size * 3/2 + h->out.i_nal * 4 + 4 + 64;
2054     for( int i = start; i < h->out.i_nal; i++ )
2055         necessary_size += h->out.nal[i].i_padding;
2056     if( check_encapsulated_buffer( h, h0, start, previous_nal_size, necessary_size ) )
2057         return -1;
2058 
2059     uint8_t *nal_buffer = h0->nal_buffer + previous_nal_size;
2060 
2061     for( int i = start; i < h->out.i_nal; i++ )
2062     {
2063         h->out.nal[i].b_long_startcode = !i || h->out.nal[i].i_type == NAL_SPS || h->out.nal[i].i_type == NAL_PPS ||
2064                                          h->param.i_avcintra_class;
2065         x264_nal_encode( h, nal_buffer, &h->out.nal[i] );
2066         nal_buffer += h->out.nal[i].i_payload;
2067     }
2068 
2069     x264_emms();
2070 
2071     return nal_buffer - (h0->nal_buffer + previous_nal_size);
2072 }
2073 
2074 /****************************************************************************
2075  * x264_encoder_headers:
2076  ****************************************************************************/
x264_encoder_headers(x264_t * h,x264_nal_t ** pp_nal,int * pi_nal)2077 int x264_encoder_headers( x264_t *h, x264_nal_t **pp_nal, int *pi_nal )
2078 {
2079     int frame_size = 0;
2080     /* init bitstream context */
2081     h->out.i_nal = 0;
2082     bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
2083 
2084     /* Write SEI, SPS and PPS. */
2085 
2086     /* generate sequence parameters */
2087     nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
2088     x264_sps_write( &h->out.bs, h->sps );
2089     if( nal_end( h ) )
2090         return -1;
2091 
2092     /* generate picture parameters */
2093     nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
2094     x264_pps_write( &h->out.bs, h->sps, h->pps );
2095     if( nal_end( h ) )
2096         return -1;
2097 
2098     /* identify ourselves */
2099     nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
2100     if( x264_sei_version_write( h, &h->out.bs ) )
2101         return -1;
2102     if( nal_end( h ) )
2103         return -1;
2104 
2105     frame_size = encoder_encapsulate_nals( h, 0 );
2106     if( frame_size < 0 )
2107         return -1;
2108 
2109     /* now set output*/
2110     *pi_nal = h->out.i_nal;
2111     *pp_nal = &h->out.nal[0];
2112     h->out.i_nal = 0;
2113 
2114     return frame_size;
2115 }
2116 
2117 /* Check to see whether we have chosen a reference list ordering different
2118  * from the standard's default. */
reference_check_reorder(x264_t * h)2119 static inline void reference_check_reorder( x264_t *h )
2120 {
2121     /* The reorder check doesn't check for missing frames, so just
2122      * force a reorder if one of the reference list is corrupt. */
2123     for( int i = 0; h->frames.reference[i]; i++ )
2124         if( h->frames.reference[i]->b_corrupt )
2125         {
2126             h->b_ref_reorder[0] = 1;
2127             return;
2128         }
2129     for( int list = 0; list <= (h->sh.i_type == SLICE_TYPE_B); list++ )
2130         for( int i = 0; i < h->i_ref[list] - 1; i++ )
2131         {
2132             int framenum_diff = h->fref[list][i+1]->i_frame_num - h->fref[list][i]->i_frame_num;
2133             int poc_diff = h->fref[list][i+1]->i_poc - h->fref[list][i]->i_poc;
2134             /* P and B-frames use different default orders. */
2135             if( h->sh.i_type == SLICE_TYPE_P ? framenum_diff > 0 : list == 1 ? poc_diff < 0 : poc_diff > 0 )
2136             {
2137                 h->b_ref_reorder[list] = 1;
2138                 return;
2139             }
2140         }
2141 }
2142 
2143 /* return -1 on failure, else return the index of the new reference frame */
weighted_reference_duplicate(x264_t * h,int i_ref,const x264_weight_t * w)2144 static int weighted_reference_duplicate( x264_t *h, int i_ref, const x264_weight_t *w )
2145 {
2146     int i = h->i_ref[0];
2147     int j = 1;
2148     x264_frame_t *newframe;
2149     if( i <= 1 ) /* empty list, definitely can't duplicate frame */
2150         return -1;
2151 
2152     //Duplication is only used in X264_WEIGHTP_SMART
2153     if( h->param.analyse.i_weighted_pred != X264_WEIGHTP_SMART )
2154         return -1;
2155 
2156     /* Duplication is a hack to compensate for crappy rounding in motion compensation.
2157      * With high bit depth, it's not worth doing, so turn it off except in the case of
2158      * unweighted dupes. */
2159     if( BIT_DEPTH > 8 && w != x264_weight_none )
2160         return -1;
2161 
2162     newframe = x264_frame_pop_blank_unused( h );
2163     if( !newframe )
2164         return -1;
2165 
2166     //FIXME: probably don't need to copy everything
2167     *newframe = *h->fref[0][i_ref];
2168     newframe->i_reference_count = 1;
2169     newframe->orig = h->fref[0][i_ref];
2170     newframe->b_duplicate = 1;
2171     memcpy( h->fenc->weight[j], w, sizeof(h->fenc->weight[i]) );
2172 
2173     /* shift the frames to make space for the dupe. */
2174     h->b_ref_reorder[0] = 1;
2175     if( h->i_ref[0] < X264_REF_MAX )
2176         ++h->i_ref[0];
2177     h->fref[0][X264_REF_MAX-1] = NULL;
2178     x264_frame_unshift( &h->fref[0][j], newframe );
2179 
2180     return j;
2181 }
2182 
weighted_pred_init(x264_t * h)2183 static void weighted_pred_init( x264_t *h )
2184 {
2185     /* for now no analysis and set all weights to nothing */
2186     for( int i_ref = 0; i_ref < h->i_ref[0]; i_ref++ )
2187         h->fenc->weighted[i_ref] = h->fref[0][i_ref]->filtered[0][0];
2188 
2189     // FIXME: This only supports weighting of one reference frame
2190     // and duplicates of that frame.
2191     h->fenc->i_lines_weighted = 0;
2192 
2193     for( int i_ref = 0; i_ref < (h->i_ref[0] << SLICE_MBAFF); i_ref++ )
2194         for( int i = 0; i < 3; i++ )
2195             h->sh.weight[i_ref][i].weightfn = NULL;
2196 
2197 
2198     if( h->sh.i_type != SLICE_TYPE_P || h->param.analyse.i_weighted_pred <= 0 )
2199         return;
2200 
2201     int i_padv = PADV << PARAM_INTERLACED;
2202     int denom = -1;
2203     int weightplane[2] = { 0, 0 };
2204     int buffer_next = 0;
2205     for( int i = 0; i < 3; i++ )
2206     {
2207         for( int j = 0; j < h->i_ref[0]; j++ )
2208         {
2209             if( h->fenc->weight[j][i].weightfn )
2210             {
2211                 h->sh.weight[j][i] = h->fenc->weight[j][i];
2212                 // if weight is useless, don't write it to stream
2213                 if( h->sh.weight[j][i].i_scale == 1<<h->sh.weight[j][i].i_denom && h->sh.weight[j][i].i_offset == 0 )
2214                     h->sh.weight[j][i].weightfn = NULL;
2215                 else
2216                 {
2217                     if( !weightplane[!!i] )
2218                     {
2219                         weightplane[!!i] = 1;
2220                         h->sh.weight[0][!!i].i_denom = denom = h->sh.weight[j][i].i_denom;
2221                         assert( x264_clip3( denom, 0, 7 ) == denom );
2222                     }
2223 
2224                     assert( h->sh.weight[j][i].i_denom == denom );
2225                     if( !i )
2226                     {
2227                         h->fenc->weighted[j] = h->mb.p_weight_buf[buffer_next++] + h->fenc->i_stride[0] * i_padv + PADH_ALIGN;
2228                         //scale full resolution frame
2229                         if( h->param.i_threads == 1 )
2230                         {
2231                             pixel *src = h->fref[0][j]->filtered[0][0] - h->fref[0][j]->i_stride[0]*i_padv - PADH_ALIGN;
2232                             pixel *dst = h->fenc->weighted[j] - h->fenc->i_stride[0]*i_padv - PADH_ALIGN;
2233                             int stride = h->fenc->i_stride[0];
2234                             int width = h->fenc->i_width[0] + PADH2;
2235                             int height = h->fenc->i_lines[0] + i_padv*2;
2236                             x264_weight_scale_plane( h, dst, stride, src, stride, width, height, &h->sh.weight[j][0] );
2237                             h->fenc->i_lines_weighted = height;
2238                         }
2239                     }
2240                 }
2241             }
2242         }
2243     }
2244 
2245     if( weightplane[1] )
2246         for( int i = 0; i < h->i_ref[0]; i++ )
2247         {
2248             if( h->sh.weight[i][1].weightfn && !h->sh.weight[i][2].weightfn )
2249             {
2250                 h->sh.weight[i][2].i_scale = 1 << h->sh.weight[0][1].i_denom;
2251                 h->sh.weight[i][2].i_offset = 0;
2252             }
2253             else if( h->sh.weight[i][2].weightfn && !h->sh.weight[i][1].weightfn )
2254             {
2255                 h->sh.weight[i][1].i_scale = 1 << h->sh.weight[0][1].i_denom;
2256                 h->sh.weight[i][1].i_offset = 0;
2257             }
2258         }
2259 
2260     if( !weightplane[0] )
2261         h->sh.weight[0][0].i_denom = 0;
2262     if( !weightplane[1] )
2263         h->sh.weight[0][1].i_denom = 0;
2264     h->sh.weight[0][2].i_denom = h->sh.weight[0][1].i_denom;
2265 }
2266 
reference_distance(x264_t * h,x264_frame_t * frame)2267 static inline int reference_distance( x264_t *h, x264_frame_t *frame )
2268 {
2269     if( h->param.i_frame_packing == 5 )
2270         return abs((h->fenc->i_frame&~1) - (frame->i_frame&~1)) +
2271                   ((h->fenc->i_frame&1) != (frame->i_frame&1));
2272     else
2273         return abs(h->fenc->i_frame - frame->i_frame);
2274 }
2275 
reference_build_list(x264_t * h,int i_poc)2276 static inline void reference_build_list( x264_t *h, int i_poc )
2277 {
2278     int b_ok;
2279 
2280     /* build ref list 0/1 */
2281     h->mb.pic.i_fref[0] = h->i_ref[0] = 0;
2282     h->mb.pic.i_fref[1] = h->i_ref[1] = 0;
2283     if( h->sh.i_type == SLICE_TYPE_I )
2284         return;
2285 
2286     for( int i = 0; h->frames.reference[i]; i++ )
2287     {
2288         if( h->frames.reference[i]->b_corrupt )
2289             continue;
2290         if( h->frames.reference[i]->i_poc < i_poc )
2291             h->fref[0][h->i_ref[0]++] = h->frames.reference[i];
2292         else if( h->frames.reference[i]->i_poc > i_poc )
2293             h->fref[1][h->i_ref[1]++] = h->frames.reference[i];
2294     }
2295 
2296     if( h->sh.i_mmco_remove_from_end )
2297     {
2298         /* Order ref0 for MMCO remove */
2299         do
2300         {
2301             b_ok = 1;
2302             for( int i = 0; i < h->i_ref[0] - 1; i++ )
2303             {
2304                 if( h->fref[0][i]->i_frame < h->fref[0][i+1]->i_frame )
2305                 {
2306                     XCHG( x264_frame_t*, h->fref[0][i], h->fref[0][i+1] );
2307                     b_ok = 0;
2308                     break;
2309                 }
2310             }
2311         } while( !b_ok );
2312 
2313         for( int i = h->i_ref[0]-1; i >= h->i_ref[0] - h->sh.i_mmco_remove_from_end; i-- )
2314         {
2315             int diff = h->i_frame_num - h->fref[0][i]->i_frame_num;
2316             h->sh.mmco[h->sh.i_mmco_command_count].i_poc = h->fref[0][i]->i_poc;
2317             h->sh.mmco[h->sh.i_mmco_command_count++].i_difference_of_pic_nums = diff;
2318         }
2319     }
2320 
2321     /* Order reference lists by distance from the current frame. */
2322     for( int list = 0; list < 2; list++ )
2323     {
2324         h->fref_nearest[list] = h->fref[list][0];
2325         do
2326         {
2327             b_ok = 1;
2328             for( int i = 0; i < h->i_ref[list] - 1; i++ )
2329             {
2330                 if( list ? h->fref[list][i+1]->i_poc < h->fref_nearest[list]->i_poc
2331                          : h->fref[list][i+1]->i_poc > h->fref_nearest[list]->i_poc )
2332                     h->fref_nearest[list] = h->fref[list][i+1];
2333                 if( reference_distance( h, h->fref[list][i] ) > reference_distance( h, h->fref[list][i+1] ) )
2334                 {
2335                     XCHG( x264_frame_t*, h->fref[list][i], h->fref[list][i+1] );
2336                     b_ok = 0;
2337                     break;
2338                 }
2339             }
2340         } while( !b_ok );
2341     }
2342 
2343     reference_check_reorder( h );
2344 
2345     h->i_ref[1] = X264_MIN( h->i_ref[1], h->frames.i_max_ref1 );
2346     h->i_ref[0] = X264_MIN( h->i_ref[0], h->frames.i_max_ref0 );
2347     h->i_ref[0] = X264_MIN( h->i_ref[0], h->param.i_frame_reference ); // if reconfig() has lowered the limit
2348 
2349     /* For Blu-ray compliance, don't reference frames outside of the minigop. */
2350     if( IS_X264_TYPE_B( h->fenc->i_type ) && h->param.b_bluray_compat )
2351         h->i_ref[0] = X264_MIN( h->i_ref[0], IS_X264_TYPE_B( h->fref[0][0]->i_type ) + 1 );
2352 
2353     /* add duplicates */
2354     if( h->fenc->i_type == X264_TYPE_P )
2355     {
2356         int idx = -1;
2357         if( h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE )
2358         {
2359             x264_weight_t w[3];
2360             w[1].weightfn = w[2].weightfn = NULL;
2361             if( h->param.rc.b_stat_read )
2362                 x264_ratecontrol_set_weights( h, h->fenc );
2363 
2364             if( !h->fenc->weight[0][0].weightfn )
2365             {
2366                 h->fenc->weight[0][0].i_denom = 0;
2367                 SET_WEIGHT( w[0], 1, 1, 0, -1 );
2368                 idx = weighted_reference_duplicate( h, 0, w );
2369             }
2370             else
2371             {
2372                 if( h->fenc->weight[0][0].i_scale == 1<<h->fenc->weight[0][0].i_denom )
2373                 {
2374                     SET_WEIGHT( h->fenc->weight[0][0], 1, 1, 0, h->fenc->weight[0][0].i_offset );
2375                 }
2376                 weighted_reference_duplicate( h, 0, x264_weight_none );
2377                 if( h->fenc->weight[0][0].i_offset > -128 )
2378                 {
2379                     w[0] = h->fenc->weight[0][0];
2380                     w[0].i_offset--;
2381                     h->mc.weight_cache( h, &w[0] );
2382                     idx = weighted_reference_duplicate( h, 0, w );
2383                 }
2384             }
2385         }
2386         h->mb.ref_blind_dupe = idx;
2387     }
2388 
2389     assert( h->i_ref[0] + h->i_ref[1] <= X264_REF_MAX );
2390     h->mb.pic.i_fref[0] = h->i_ref[0];
2391     h->mb.pic.i_fref[1] = h->i_ref[1];
2392 }
2393 
fdec_filter_row(x264_t * h,int mb_y,int pass)2394 static void fdec_filter_row( x264_t *h, int mb_y, int pass )
2395 {
2396     /* mb_y is the mb to be encoded next, not the mb to be filtered here */
2397     int b_hpel = h->fdec->b_kept_as_ref;
2398     int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
2399     int b_end = mb_y == h->i_threadslice_end;
2400     int b_measure_quality = 1;
2401     int min_y = mb_y - (1 << SLICE_MBAFF);
2402     int b_start = min_y == h->i_threadslice_start;
2403     /* Even in interlaced mode, deblocking never modifies more than 4 pixels
2404      * above each MB, as bS=4 doesn't happen for the top of interlaced mbpairs. */
2405     int minpix_y = min_y*16 - 4 * !b_start;
2406     int maxpix_y = mb_y*16 - 4 * !b_end;
2407     b_deblock &= b_hpel || h->param.b_full_recon || h->param.psz_dump_yuv;
2408     if( h->param.b_sliced_threads )
2409     {
2410         switch( pass )
2411         {
2412             /* During encode: only do deblock if asked for */
2413             default:
2414             case 0:
2415                 b_deblock &= h->param.b_full_recon;
2416                 b_hpel = 0;
2417                 break;
2418             /* During post-encode pass: do deblock if not done yet, do hpel for all
2419              * rows except those between slices. */
2420             case 1:
2421                 b_deblock &= !h->param.b_full_recon;
2422                 b_hpel &= !(b_start && min_y > 0);
2423                 b_measure_quality = 0;
2424                 break;
2425             /* Final pass: do the rows between slices in sequence. */
2426             case 2:
2427                 b_deblock = 0;
2428                 b_measure_quality = 0;
2429                 break;
2430         }
2431     }
2432     if( mb_y & SLICE_MBAFF )
2433         return;
2434     if( min_y < h->i_threadslice_start )
2435         return;
2436 
2437     if( b_deblock )
2438         for( int y = min_y; y < mb_y; y += (1 << SLICE_MBAFF) )
2439             x264_frame_deblock_row( h, y );
2440 
2441     /* FIXME: Prediction requires different borders for interlaced/progressive mc,
2442      * but the actual image data is equivalent. For now, maintain this
2443      * consistency by copying deblocked pixels between planes. */
2444     if( PARAM_INTERLACED && (!h->param.b_sliced_threads || pass == 1) )
2445         for( int p = 0; p < h->fdec->i_plane; p++ )
2446             for( int i = minpix_y>>(CHROMA_V_SHIFT && p); i < maxpix_y>>(CHROMA_V_SHIFT && p); i++ )
2447                 memcpy( h->fdec->plane_fld[p] + i*h->fdec->i_stride[p],
2448                         h->fdec->plane[p]     + i*h->fdec->i_stride[p],
2449                         h->mb.i_mb_width*16*SIZEOF_PIXEL );
2450 
2451     if( h->fdec->b_kept_as_ref && (!h->param.b_sliced_threads || pass == 1) )
2452         x264_frame_expand_border( h, h->fdec, min_y );
2453     if( b_hpel )
2454     {
2455         int end = mb_y == h->mb.i_mb_height;
2456         /* Can't do hpel until the previous slice is done encoding. */
2457         if( h->param.analyse.i_subpel_refine )
2458         {
2459             x264_frame_filter( h, h->fdec, min_y, end );
2460             x264_frame_expand_border_filtered( h, h->fdec, min_y, end );
2461         }
2462     }
2463 
2464     if( SLICE_MBAFF && pass == 0 )
2465         for( int i = 0; i < 3; i++ )
2466         {
2467             XCHG( pixel *, h->intra_border_backup[0][i], h->intra_border_backup[3][i] );
2468             XCHG( pixel *, h->intra_border_backup[1][i], h->intra_border_backup[4][i] );
2469         }
2470 
2471     if( h->i_thread_frames > 1 && h->fdec->b_kept_as_ref )
2472         x264_frame_cond_broadcast( h->fdec, mb_y*16 + (b_end ? 10000 : -(X264_THREAD_HEIGHT << SLICE_MBAFF)) );
2473 
2474     if( b_measure_quality )
2475     {
2476         maxpix_y = X264_MIN( maxpix_y, h->param.i_height );
2477         if( h->param.analyse.b_psnr )
2478         {
2479             for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
2480                 h->stat.frame.i_ssd[p] += x264_pixel_ssd_wxh( &h->pixf,
2481                     h->fdec->plane[p] + minpix_y * h->fdec->i_stride[p], h->fdec->i_stride[p],
2482                     h->fenc->plane[p] + minpix_y * h->fenc->i_stride[p], h->fenc->i_stride[p],
2483                     h->param.i_width, maxpix_y-minpix_y );
2484             if( !CHROMA444 )
2485             {
2486                 uint64_t ssd_u, ssd_v;
2487                 int v_shift = CHROMA_V_SHIFT;
2488                 x264_pixel_ssd_nv12( &h->pixf,
2489                     h->fdec->plane[1] + (minpix_y>>v_shift) * h->fdec->i_stride[1], h->fdec->i_stride[1],
2490                     h->fenc->plane[1] + (minpix_y>>v_shift) * h->fenc->i_stride[1], h->fenc->i_stride[1],
2491                     h->param.i_width>>1, (maxpix_y-minpix_y)>>v_shift, &ssd_u, &ssd_v );
2492                 h->stat.frame.i_ssd[1] += ssd_u;
2493                 h->stat.frame.i_ssd[2] += ssd_v;
2494             }
2495         }
2496 
2497         if( h->param.analyse.b_ssim )
2498         {
2499             int ssim_cnt;
2500             x264_emms();
2501             /* offset by 2 pixels to avoid alignment of ssim blocks with dct blocks,
2502              * and overlap by 4 */
2503             minpix_y += b_start ? 2 : -6;
2504             h->stat.frame.f_ssim +=
2505                 x264_pixel_ssim_wxh( &h->pixf,
2506                     h->fdec->plane[0] + 2+minpix_y*h->fdec->i_stride[0], h->fdec->i_stride[0],
2507                     h->fenc->plane[0] + 2+minpix_y*h->fenc->i_stride[0], h->fenc->i_stride[0],
2508                     h->param.i_width-2, maxpix_y-minpix_y, h->scratch_buffer, &ssim_cnt );
2509             h->stat.frame.i_ssim_cnt += ssim_cnt;
2510         }
2511     }
2512 }
2513 
reference_update(x264_t * h)2514 static inline int reference_update( x264_t *h )
2515 {
2516     if( !h->fdec->b_kept_as_ref )
2517     {
2518         if( h->i_thread_frames > 1 )
2519         {
2520             x264_frame_push_unused( h, h->fdec );
2521             h->fdec = x264_frame_pop_unused( h, 1 );
2522             if( !h->fdec )
2523                 return -1;
2524         }
2525         return 0;
2526     }
2527 
2528     /* apply mmco from previous frame. */
2529     for( int i = 0; i < h->sh.i_mmco_command_count; i++ )
2530         for( int j = 0; h->frames.reference[j]; j++ )
2531             if( h->frames.reference[j]->i_poc == h->sh.mmco[i].i_poc )
2532                 x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[j] ) );
2533 
2534     /* move frame in the buffer */
2535     x264_frame_push( h->frames.reference, h->fdec );
2536     if( h->frames.reference[h->sps->i_num_ref_frames] )
2537         x264_frame_push_unused( h, x264_frame_shift( h->frames.reference ) );
2538     h->fdec = x264_frame_pop_unused( h, 1 );
2539     if( !h->fdec )
2540         return -1;
2541     return 0;
2542 }
2543 
reference_reset(x264_t * h)2544 static inline void reference_reset( x264_t *h )
2545 {
2546     while( h->frames.reference[0] )
2547         x264_frame_push_unused( h, x264_frame_pop( h->frames.reference ) );
2548     h->fdec->i_poc =
2549     h->fenc->i_poc = 0;
2550 }
2551 
reference_hierarchy_reset(x264_t * h)2552 static inline void reference_hierarchy_reset( x264_t *h )
2553 {
2554     int ref;
2555     int b_hasdelayframe = 0;
2556 
2557     /* look for delay frames -- chain must only contain frames that are disposable */
2558     for( int i = 0; h->frames.current[i] && IS_DISPOSABLE( h->frames.current[i]->i_type ); i++ )
2559         b_hasdelayframe |= h->frames.current[i]->i_coded
2560                         != h->frames.current[i]->i_frame + h->sps->vui.i_num_reorder_frames;
2561 
2562     /* This function must handle b-pyramid and clear frames for open-gop */
2563     if( h->param.i_bframe_pyramid != X264_B_PYRAMID_STRICT && !b_hasdelayframe && h->frames.i_poc_last_open_gop == -1 )
2564         return;
2565 
2566     /* Remove last BREF. There will never be old BREFs in the
2567      * dpb during a BREF decode when pyramid == STRICT */
2568     for( ref = 0; h->frames.reference[ref]; ref++ )
2569     {
2570         if( ( h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT
2571             && h->frames.reference[ref]->i_type == X264_TYPE_BREF )
2572             || ( h->frames.reference[ref]->i_poc < h->frames.i_poc_last_open_gop
2573             && h->sh.i_type != SLICE_TYPE_B ) )
2574         {
2575             int diff = h->i_frame_num - h->frames.reference[ref]->i_frame_num;
2576             h->sh.mmco[h->sh.i_mmco_command_count].i_difference_of_pic_nums = diff;
2577             h->sh.mmco[h->sh.i_mmco_command_count++].i_poc = h->frames.reference[ref]->i_poc;
2578             x264_frame_push_unused( h, x264_frame_shift( &h->frames.reference[ref] ) );
2579             h->b_ref_reorder[0] = 1;
2580             ref--;
2581         }
2582     }
2583 
2584     /* Prepare room in the dpb for the delayed display time of the later b-frame's */
2585     if( h->param.i_bframe_pyramid )
2586         h->sh.i_mmco_remove_from_end = X264_MAX( ref + 2 - h->frames.i_max_dpb, 0 );
2587 }
2588 
slice_init(x264_t * h,int i_nal_type,int i_global_qp)2589 static inline void slice_init( x264_t *h, int i_nal_type, int i_global_qp )
2590 {
2591     /* ------------------------ Create slice header  ----------------------- */
2592     if( i_nal_type == NAL_SLICE_IDR )
2593     {
2594         slice_header_init( h, &h->sh, h->sps, h->pps, h->i_idr_pic_id, h->i_frame_num, i_global_qp );
2595 
2596         /* alternate id */
2597         if( h->param.i_avcintra_class )
2598         {
2599             switch( h->i_idr_pic_id )
2600             {
2601                 case 5:
2602                     h->i_idr_pic_id = 3;
2603                     break;
2604                 case 3:
2605                     h->i_idr_pic_id = 4;
2606                     break;
2607                 case 4:
2608                 default:
2609                     h->i_idr_pic_id = 5;
2610                     break;
2611             }
2612         }
2613         else
2614             h->i_idr_pic_id ^= 1;
2615     }
2616     else
2617     {
2618         slice_header_init( h, &h->sh, h->sps, h->pps, -1, h->i_frame_num, i_global_qp );
2619 
2620         h->sh.i_num_ref_idx_l0_active = h->i_ref[0] <= 0 ? 1 : h->i_ref[0];
2621         h->sh.i_num_ref_idx_l1_active = h->i_ref[1] <= 0 ? 1 : h->i_ref[1];
2622         if( h->sh.i_num_ref_idx_l0_active != h->pps->i_num_ref_idx_l0_default_active ||
2623             (h->sh.i_type == SLICE_TYPE_B && h->sh.i_num_ref_idx_l1_active != h->pps->i_num_ref_idx_l1_default_active) )
2624         {
2625             h->sh.b_num_ref_idx_override = 1;
2626         }
2627     }
2628 
2629     if( h->fenc->i_type == X264_TYPE_BREF && h->param.b_bluray_compat && h->sh.i_mmco_command_count )
2630     {
2631         h->b_sh_backup = 1;
2632         h->sh_backup = h->sh;
2633     }
2634 
2635     h->fdec->i_frame_num = h->sh.i_frame_num;
2636 
2637     if( h->sps->i_poc_type == 0 )
2638     {
2639         h->sh.i_poc = h->fdec->i_poc;
2640         if( PARAM_INTERLACED )
2641         {
2642             h->sh.i_delta_poc_bottom = h->param.b_tff ? 1 : -1;
2643             h->sh.i_poc += h->sh.i_delta_poc_bottom == -1;
2644         }
2645         else
2646             h->sh.i_delta_poc_bottom = 0;
2647         h->fdec->i_delta_poc[0] = h->sh.i_delta_poc_bottom == -1;
2648         h->fdec->i_delta_poc[1] = h->sh.i_delta_poc_bottom ==  1;
2649     }
2650     else
2651     {
2652         /* Nothing to do ? */
2653     }
2654 
2655     x264_macroblock_slice_init( h );
2656 }
2657 
2658 typedef struct
2659 {
2660     int skip;
2661     uint8_t cabac_prevbyte;
2662     bs_t bs;
2663     x264_cabac_t cabac;
2664     x264_frame_stat_t stat;
2665     int last_qp;
2666     int last_dqp;
2667     int field_decoding_flag;
2668 } x264_bs_bak_t;
2669 
bitstream_backup(x264_t * h,x264_bs_bak_t * bak,int i_skip,int full)2670 static ALWAYS_INLINE void bitstream_backup( x264_t *h, x264_bs_bak_t *bak, int i_skip, int full )
2671 {
2672     if( full )
2673     {
2674         bak->stat = h->stat.frame;
2675         bak->last_qp = h->mb.i_last_qp;
2676         bak->last_dqp = h->mb.i_last_dqp;
2677         bak->field_decoding_flag = h->mb.field_decoding_flag;
2678     }
2679     else
2680     {
2681         bak->stat.i_mv_bits = h->stat.frame.i_mv_bits;
2682         bak->stat.i_tex_bits = h->stat.frame.i_tex_bits;
2683     }
2684     /* In the per-MB backup, we don't need the contexts because flushing the CABAC
2685      * encoder has no context dependency and in this case, a slice is ended (and
2686      * thus the content of all contexts are thrown away). */
2687     if( h->param.b_cabac )
2688     {
2689         if( full )
2690             memcpy( &bak->cabac, &h->cabac, sizeof(x264_cabac_t) );
2691         else
2692             memcpy( &bak->cabac, &h->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
2693         /* x264's CABAC writer modifies the previous byte during carry, so it has to be
2694          * backed up. */
2695         bak->cabac_prevbyte = h->cabac.p[-1];
2696     }
2697     else
2698     {
2699         bak->bs = h->out.bs;
2700         bak->skip = i_skip;
2701     }
2702 }
2703 
bitstream_restore(x264_t * h,x264_bs_bak_t * bak,int * skip,int full)2704 static ALWAYS_INLINE void bitstream_restore( x264_t *h, x264_bs_bak_t *bak, int *skip, int full )
2705 {
2706     if( full )
2707     {
2708         h->stat.frame = bak->stat;
2709         h->mb.i_last_qp = bak->last_qp;
2710         h->mb.i_last_dqp = bak->last_dqp;
2711         h->mb.field_decoding_flag = bak->field_decoding_flag;
2712     }
2713     else
2714     {
2715         h->stat.frame.i_mv_bits = bak->stat.i_mv_bits;
2716         h->stat.frame.i_tex_bits = bak->stat.i_tex_bits;
2717     }
2718     if( h->param.b_cabac )
2719     {
2720         if( full )
2721             memcpy( &h->cabac, &bak->cabac, sizeof(x264_cabac_t) );
2722         else
2723             memcpy( &h->cabac, &bak->cabac, offsetof(x264_cabac_t, f8_bits_encoded) );
2724         h->cabac.p[-1] = bak->cabac_prevbyte;
2725     }
2726     else
2727     {
2728         h->out.bs = bak->bs;
2729         *skip = bak->skip;
2730     }
2731 }
2732 
slice_write(x264_t * h)2733 static intptr_t slice_write( x264_t *h )
2734 {
2735     int i_skip;
2736     int mb_xy, i_mb_x, i_mb_y;
2737     /* NALUs other than the first use a 3-byte startcode.
2738      * Add one extra byte for the rbsp, and one more for the final CABAC putbyte.
2739      * Then add an extra 5 bytes just in case, to account for random NAL escapes and
2740      * other inaccuracies. */
2741     int overhead_guess = (NALU_OVERHEAD - (h->param.b_annexb && h->out.i_nal)) + 1 + h->param.b_cabac + 5;
2742     int slice_max_size = h->param.i_slice_max_size > 0 ? (h->param.i_slice_max_size-overhead_guess)*8 : 0;
2743     int back_up_bitstream_cavlc = !h->param.b_cabac && h->sps->i_profile_idc < PROFILE_HIGH;
2744     int back_up_bitstream = slice_max_size || back_up_bitstream_cavlc;
2745     int starting_bits = bs_pos(&h->out.bs);
2746     int b_deblock = h->sh.i_disable_deblocking_filter_idc != 1;
2747     int b_hpel = h->fdec->b_kept_as_ref;
2748     int orig_last_mb = h->sh.i_last_mb;
2749     int thread_last_mb = h->i_threadslice_end * h->mb.i_mb_width - 1;
2750     uint8_t *last_emu_check;
2751 #define BS_BAK_SLICE_MAX_SIZE 0
2752 #define BS_BAK_CAVLC_OVERFLOW 1
2753 #define BS_BAK_SLICE_MIN_MBS  2
2754 #define BS_BAK_ROW_VBV        3
2755     x264_bs_bak_t bs_bak[4];
2756     b_deblock &= b_hpel || h->param.b_full_recon || h->param.psz_dump_yuv;
2757     bs_realign( &h->out.bs );
2758 
2759     /* Slice */
2760     nal_start( h, h->i_nal_type, h->i_nal_ref_idc );
2761     h->out.nal[h->out.i_nal].i_first_mb = h->sh.i_first_mb;
2762 
2763     /* Slice header */
2764     x264_macroblock_thread_init( h );
2765 
2766     /* Set the QP equal to the first QP in the slice for more accurate CABAC initialization. */
2767     h->mb.i_mb_xy = h->sh.i_first_mb;
2768     h->sh.i_qp = x264_ratecontrol_mb_qp( h );
2769     h->sh.i_qp = SPEC_QP( h->sh.i_qp );
2770     h->sh.i_qp_delta = h->sh.i_qp - h->pps->i_pic_init_qp;
2771 
2772     slice_header_write( &h->out.bs, &h->sh, h->i_nal_ref_idc );
2773     if( h->param.b_cabac )
2774     {
2775         /* alignment needed */
2776         bs_align_1( &h->out.bs );
2777 
2778         /* init cabac */
2779         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 );
2780         x264_cabac_encode_init ( &h->cabac, h->out.bs.p, h->out.bs.p_end );
2781         last_emu_check = h->cabac.p;
2782     }
2783     else
2784         last_emu_check = h->out.bs.p;
2785     h->mb.i_last_qp = h->sh.i_qp;
2786     h->mb.i_last_dqp = 0;
2787     h->mb.field_decoding_flag = 0;
2788 
2789     i_mb_y = h->sh.i_first_mb / h->mb.i_mb_width;
2790     i_mb_x = h->sh.i_first_mb % h->mb.i_mb_width;
2791     i_skip = 0;
2792 
2793     while( 1 )
2794     {
2795         mb_xy = i_mb_x + i_mb_y * h->mb.i_mb_width;
2796         int mb_spos = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
2797 
2798         if( i_mb_x == 0 )
2799         {
2800             if( bitstream_check_buffer( h ) )
2801                 return -1;
2802             if( !(i_mb_y & SLICE_MBAFF) && h->param.rc.i_vbv_buffer_size )
2803                 bitstream_backup( h, &bs_bak[BS_BAK_ROW_VBV], i_skip, 1 );
2804             if( !h->mb.b_reencode_mb )
2805                 fdec_filter_row( h, i_mb_y, 0 );
2806         }
2807 
2808         if( back_up_bitstream )
2809         {
2810             if( back_up_bitstream_cavlc )
2811                 bitstream_backup( h, &bs_bak[BS_BAK_CAVLC_OVERFLOW], i_skip, 0 );
2812             if( slice_max_size && !(i_mb_y & SLICE_MBAFF) )
2813             {
2814                 bitstream_backup( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], i_skip, 0 );
2815                 if( (thread_last_mb+1-mb_xy) == h->param.i_slice_min_mbs )
2816                     bitstream_backup( h, &bs_bak[BS_BAK_SLICE_MIN_MBS], i_skip, 0 );
2817             }
2818         }
2819 
2820         if( PARAM_INTERLACED )
2821         {
2822             if( h->mb.b_adaptive_mbaff )
2823             {
2824                 if( !(i_mb_y&1) )
2825                 {
2826                     /* FIXME: VSAD is fast but fairly poor at choosing the best interlace type. */
2827                     h->mb.b_interlaced = x264_field_vsad( h, i_mb_x, i_mb_y );
2828                     memcpy( &h->zigzagf, MB_INTERLACED ? &h->zigzagf_interlaced : &h->zigzagf_progressive, sizeof(h->zigzagf) );
2829                     if( !MB_INTERLACED && (i_mb_y+2) == h->mb.i_mb_height )
2830                         x264_expand_border_mbpair( h, i_mb_x, i_mb_y );
2831                 }
2832             }
2833             h->mb.field[mb_xy] = MB_INTERLACED;
2834         }
2835 
2836         /* load cache */
2837         if( SLICE_MBAFF )
2838             x264_macroblock_cache_load_interlaced( h, i_mb_x, i_mb_y );
2839         else
2840             x264_macroblock_cache_load_progressive( h, i_mb_x, i_mb_y );
2841 
2842         x264_macroblock_analyse( h );
2843 
2844         /* encode this macroblock -> be careful it can change the mb type to P_SKIP if needed */
2845 reencode:
2846         x264_macroblock_encode( h );
2847 
2848         if( h->param.b_cabac )
2849         {
2850             if( mb_xy > h->sh.i_first_mb && !(SLICE_MBAFF && (i_mb_y&1)) )
2851                 x264_cabac_encode_terminal( &h->cabac );
2852 
2853             if( IS_SKIP( h->mb.i_type ) )
2854                 x264_cabac_mb_skip( h, 1 );
2855             else
2856             {
2857                 if( h->sh.i_type != SLICE_TYPE_I )
2858                     x264_cabac_mb_skip( h, 0 );
2859                 x264_macroblock_write_cabac( h, &h->cabac );
2860             }
2861         }
2862         else
2863         {
2864             if( IS_SKIP( h->mb.i_type ) )
2865                 i_skip++;
2866             else
2867             {
2868                 if( h->sh.i_type != SLICE_TYPE_I )
2869                 {
2870                     bs_write_ue( &h->out.bs, i_skip );  /* skip run */
2871                     i_skip = 0;
2872                 }
2873                 x264_macroblock_write_cavlc( h );
2874                 /* If there was a CAVLC level code overflow, try again at a higher QP. */
2875                 if( h->mb.b_overflow )
2876                 {
2877                     h->mb.i_chroma_qp = h->chroma_qp_table[++h->mb.i_qp];
2878                     h->mb.i_skip_intra = 0;
2879                     h->mb.b_skip_mc = 0;
2880                     h->mb.b_overflow = 0;
2881                     bitstream_restore( h, &bs_bak[BS_BAK_CAVLC_OVERFLOW], &i_skip, 0 );
2882                     goto reencode;
2883                 }
2884             }
2885         }
2886 
2887         int total_bits = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
2888         int mb_size = total_bits - mb_spos;
2889 
2890         if( slice_max_size && (!SLICE_MBAFF || (i_mb_y&1)) )
2891         {
2892             /* Count the skip run, just in case. */
2893             if( !h->param.b_cabac )
2894                 total_bits += bs_size_ue_big( i_skip );
2895             /* Check for escape bytes. */
2896             uint8_t *end = h->param.b_cabac ? h->cabac.p : h->out.bs.p;
2897             for( ; last_emu_check < end - 2; last_emu_check++ )
2898                 if( last_emu_check[0] == 0 && last_emu_check[1] == 0 && last_emu_check[2] <= 3 )
2899                 {
2900                     slice_max_size -= 8;
2901                     last_emu_check++;
2902                 }
2903             /* We'll just re-encode this last macroblock if we go over the max slice size. */
2904             if( total_bits - starting_bits > slice_max_size && !h->mb.b_reencode_mb )
2905             {
2906                 if( !x264_frame_new_slice( h, h->fdec ) )
2907                 {
2908                     /* Handle the most obnoxious slice-min-mbs edge case: we need to end the slice
2909                      * because it's gone over the maximum size, but doing so would violate slice-min-mbs.
2910                      * If possible, roll back to the last checkpoint and try again.
2911                      * We could try raising QP, but that would break in the case where a slice spans multiple
2912                      * rows, which the re-encoding infrastructure can't currently handle. */
2913                     if( mb_xy <= thread_last_mb && (thread_last_mb+1-mb_xy) < h->param.i_slice_min_mbs )
2914                     {
2915                         if( thread_last_mb-h->param.i_slice_min_mbs < h->sh.i_first_mb+h->param.i_slice_min_mbs )
2916                         {
2917                             x264_log( h, X264_LOG_WARNING, "slice-max-size violated (frame %d, cause: slice-min-mbs)\n", h->i_frame );
2918                             slice_max_size = 0;
2919                             goto cont;
2920                         }
2921                         bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MIN_MBS], &i_skip, 0 );
2922                         h->mb.b_reencode_mb = 1;
2923                         h->sh.i_last_mb = thread_last_mb-h->param.i_slice_min_mbs;
2924                         break;
2925                     }
2926                     if( mb_xy-SLICE_MBAFF*h->mb.i_mb_stride != h->sh.i_first_mb )
2927                     {
2928                         bitstream_restore( h, &bs_bak[BS_BAK_SLICE_MAX_SIZE], &i_skip, 0 );
2929                         h->mb.b_reencode_mb = 1;
2930                         if( SLICE_MBAFF )
2931                         {
2932                             // set to bottom of previous mbpair
2933                             if( i_mb_x )
2934                                 h->sh.i_last_mb = mb_xy-1+h->mb.i_mb_stride*(!(i_mb_y&1));
2935                             else
2936                                 h->sh.i_last_mb = (i_mb_y-2+!(i_mb_y&1))*h->mb.i_mb_stride + h->mb.i_mb_width - 1;
2937                         }
2938                         else
2939                             h->sh.i_last_mb = mb_xy-1;
2940                         break;
2941                     }
2942                     else
2943                         h->sh.i_last_mb = mb_xy;
2944                 }
2945                 else
2946                     slice_max_size = 0;
2947             }
2948         }
2949 cont:
2950         h->mb.b_reencode_mb = 0;
2951 
2952         /* save cache */
2953         x264_macroblock_cache_save( h );
2954 
2955         if( x264_ratecontrol_mb( h, mb_size ) < 0 )
2956         {
2957             bitstream_restore( h, &bs_bak[BS_BAK_ROW_VBV], &i_skip, 1 );
2958             h->mb.b_reencode_mb = 1;
2959             i_mb_x = 0;
2960             i_mb_y = i_mb_y - SLICE_MBAFF;
2961             h->mb.i_mb_prev_xy = i_mb_y * h->mb.i_mb_stride - 1;
2962             h->sh.i_last_mb = orig_last_mb;
2963             continue;
2964         }
2965 
2966         /* accumulate mb stats */
2967         h->stat.frame.i_mb_count[h->mb.i_type]++;
2968 
2969         int b_intra = IS_INTRA( h->mb.i_type );
2970         int b_skip = IS_SKIP( h->mb.i_type );
2971         if( h->param.i_log_level >= X264_LOG_INFO || h->param.rc.b_stat_write )
2972         {
2973             if( !b_intra && !b_skip && !IS_DIRECT( h->mb.i_type ) )
2974             {
2975                 if( h->mb.i_partition != D_8x8 )
2976                         h->stat.frame.i_mb_partition[h->mb.i_partition] += 4;
2977                     else
2978                         for( int i = 0; i < 4; i++ )
2979                             h->stat.frame.i_mb_partition[h->mb.i_sub_partition[i]] ++;
2980                 if( h->param.i_frame_reference > 1 )
2981                     for( int i_list = 0; i_list <= (h->sh.i_type == SLICE_TYPE_B); i_list++ )
2982                         for( int i = 0; i < 4; i++ )
2983                         {
2984                             int i_ref = h->mb.cache.ref[i_list][ x264_scan8[4*i] ];
2985                             if( i_ref >= 0 )
2986                                 h->stat.frame.i_mb_count_ref[i_list][i_ref] ++;
2987                         }
2988             }
2989         }
2990 
2991         if( h->param.i_log_level >= X264_LOG_INFO )
2992         {
2993             if( h->mb.i_cbp_luma | h->mb.i_cbp_chroma )
2994             {
2995                 if( CHROMA444 )
2996                 {
2997                     for( int i = 0; i < 4; i++ )
2998                         if( h->mb.i_cbp_luma & (1 << i) )
2999                             for( int p = 0; p < 3; p++ )
3000                             {
3001                                 int s8 = i*4+p*16;
3002                                 int nnz8x8 = M16( &h->mb.cache.non_zero_count[x264_scan8[s8]+0] )
3003                                            | M16( &h->mb.cache.non_zero_count[x264_scan8[s8]+8] );
3004                                 h->stat.frame.i_mb_cbp[!b_intra + p*2] += !!nnz8x8;
3005                             }
3006                 }
3007                 else
3008                 {
3009                     int cbpsum = (h->mb.i_cbp_luma&1) + ((h->mb.i_cbp_luma>>1)&1)
3010                                + ((h->mb.i_cbp_luma>>2)&1) + (h->mb.i_cbp_luma>>3);
3011                     h->stat.frame.i_mb_cbp[!b_intra + 0] += cbpsum;
3012                     h->stat.frame.i_mb_cbp[!b_intra + 2] += !!h->mb.i_cbp_chroma;
3013                     h->stat.frame.i_mb_cbp[!b_intra + 4] += h->mb.i_cbp_chroma >> 1;
3014                 }
3015             }
3016             if( h->mb.i_cbp_luma && !b_intra )
3017             {
3018                 h->stat.frame.i_mb_count_8x8dct[0] ++;
3019                 h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
3020             }
3021             if( b_intra && h->mb.i_type != I_PCM )
3022             {
3023                 if( h->mb.i_type == I_16x16 )
3024                     h->stat.frame.i_mb_pred_mode[0][h->mb.i_intra16x16_pred_mode]++;
3025                 else if( h->mb.i_type == I_8x8 )
3026                     for( int i = 0; i < 16; i += 4 )
3027                         h->stat.frame.i_mb_pred_mode[1][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
3028                 else //if( h->mb.i_type == I_4x4 )
3029                     for( int i = 0; i < 16; i++ )
3030                         h->stat.frame.i_mb_pred_mode[2][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
3031                 h->stat.frame.i_mb_pred_mode[3][x264_mb_chroma_pred_mode_fix[h->mb.i_chroma_pred_mode]]++;
3032             }
3033             h->stat.frame.i_mb_field[b_intra?0:b_skip?2:1] += MB_INTERLACED;
3034         }
3035 
3036         /* calculate deblock strength values (actual deblocking is done per-row along with hpel) */
3037         if( b_deblock )
3038             x264_macroblock_deblock_strength( h );
3039 
3040         if( mb_xy == h->sh.i_last_mb )
3041             break;
3042 
3043         if( SLICE_MBAFF )
3044         {
3045             i_mb_x += i_mb_y & 1;
3046             i_mb_y ^= i_mb_x < h->mb.i_mb_width;
3047         }
3048         else
3049             i_mb_x++;
3050         if( i_mb_x == h->mb.i_mb_width )
3051         {
3052             i_mb_y++;
3053             i_mb_x = 0;
3054         }
3055     }
3056     if( h->sh.i_last_mb < h->sh.i_first_mb )
3057         return 0;
3058 
3059     h->out.nal[h->out.i_nal].i_last_mb = h->sh.i_last_mb;
3060 
3061     if( h->param.b_cabac )
3062     {
3063         x264_cabac_encode_flush( h, &h->cabac );
3064         h->out.bs.p = h->cabac.p;
3065     }
3066     else
3067     {
3068         if( i_skip > 0 )
3069             bs_write_ue( &h->out.bs, i_skip );  /* last skip run */
3070         /* rbsp_slice_trailing_bits */
3071         bs_rbsp_trailing( &h->out.bs );
3072         bs_flush( &h->out.bs );
3073     }
3074     if( nal_end( h ) )
3075         return -1;
3076 
3077     if( h->sh.i_last_mb == (h->i_threadslice_end * h->mb.i_mb_width - 1) )
3078     {
3079         h->stat.frame.i_misc_bits = bs_pos( &h->out.bs )
3080                                   + (h->out.i_nal*NALU_OVERHEAD * 8)
3081                                   - h->stat.frame.i_tex_bits
3082                                   - h->stat.frame.i_mv_bits;
3083         fdec_filter_row( h, h->i_threadslice_end, 0 );
3084 
3085         if( h->param.b_sliced_threads )
3086         {
3087             /* Tell the main thread we're done. */
3088             x264_threadslice_cond_broadcast( h, 1 );
3089             /* Do hpel now */
3090             for( int mb_y = h->i_threadslice_start; mb_y <= h->i_threadslice_end; mb_y++ )
3091                 fdec_filter_row( h, mb_y, 1 );
3092             x264_threadslice_cond_broadcast( h, 2 );
3093             /* Do the first row of hpel, now that the previous slice is done */
3094             if( h->i_thread_idx > 0 )
3095             {
3096                 x264_threadslice_cond_wait( h->thread[h->i_thread_idx-1], 2 );
3097                 fdec_filter_row( h, h->i_threadslice_start + (1 << SLICE_MBAFF), 2 );
3098             }
3099         }
3100 
3101         /* Free mb info after the last thread's done using it */
3102         if( h->fdec->mb_info_free && (!h->param.b_sliced_threads || h->i_thread_idx == (h->param.i_threads-1)) )
3103         {
3104             h->fdec->mb_info_free( h->fdec->mb_info );
3105             h->fdec->mb_info = NULL;
3106             h->fdec->mb_info_free = NULL;
3107         }
3108     }
3109 
3110     return 0;
3111 }
3112 
thread_sync_context(x264_t * dst,x264_t * src)3113 static void thread_sync_context( x264_t *dst, x264_t *src )
3114 {
3115     if( dst == src )
3116         return;
3117 
3118     // reference counting
3119     for( x264_frame_t **f = src->frames.reference; *f; f++ )
3120         (*f)->i_reference_count++;
3121     for( x264_frame_t **f = dst->frames.reference; *f; f++ )
3122         x264_frame_push_unused( src, *f );
3123     src->fdec->i_reference_count++;
3124     x264_frame_push_unused( src, dst->fdec );
3125 
3126     // copy everything except the per-thread pointers and the constants.
3127     memcpy( &dst->i_frame, &src->i_frame, offsetof(x264_t, mb.base) - offsetof(x264_t, i_frame) );
3128     dst->param = src->param;
3129     dst->stat = src->stat;
3130     dst->pixf = src->pixf;
3131     dst->reconfig = src->reconfig;
3132 }
3133 
thread_sync_stat(x264_t * dst,x264_t * src)3134 static void thread_sync_stat( x264_t *dst, x264_t *src )
3135 {
3136     if( dst != src )
3137         memcpy( &dst->stat, &src->stat, offsetof(x264_t, stat.frame) - offsetof(x264_t, stat) );
3138 }
3139 
slices_write(x264_t * h)3140 static void *slices_write( x264_t *h )
3141 {
3142     int i_slice_num = 0;
3143     int last_thread_mb = h->sh.i_last_mb;
3144     int round_bias = h->param.i_avcintra_class ? 0 : h->param.i_slice_count/2;
3145 
3146     /* init stats */
3147     memset( &h->stat.frame, 0, sizeof(h->stat.frame) );
3148     h->mb.b_reencode_mb = 0;
3149     while( h->sh.i_first_mb + SLICE_MBAFF*h->mb.i_mb_stride <= last_thread_mb )
3150     {
3151         h->sh.i_last_mb = last_thread_mb;
3152         if( !i_slice_num || !x264_frame_new_slice( h, h->fdec ) )
3153         {
3154             if( h->param.i_slice_max_mbs )
3155             {
3156                 if( SLICE_MBAFF )
3157                 {
3158                     // convert first to mbaff form, add slice-max-mbs, then convert back to normal form
3159                     int last_mbaff = 2*(h->sh.i_first_mb % h->mb.i_mb_width)
3160                         + h->mb.i_mb_width*(h->sh.i_first_mb / h->mb.i_mb_width)
3161                         + h->param.i_slice_max_mbs - 1;
3162                     int last_x = (last_mbaff % (2*h->mb.i_mb_width))/2;
3163                     int last_y = (last_mbaff / (2*h->mb.i_mb_width))*2 + 1;
3164                     h->sh.i_last_mb = last_x + h->mb.i_mb_stride*last_y;
3165                 }
3166                 else
3167                 {
3168                     h->sh.i_last_mb = h->sh.i_first_mb + h->param.i_slice_max_mbs - 1;
3169                     if( h->sh.i_last_mb < last_thread_mb && last_thread_mb - h->sh.i_last_mb < h->param.i_slice_min_mbs )
3170                         h->sh.i_last_mb = last_thread_mb - h->param.i_slice_min_mbs;
3171                 }
3172                 i_slice_num++;
3173             }
3174             else if( h->param.i_slice_count && !h->param.b_sliced_threads )
3175             {
3176                 int height = h->mb.i_mb_height >> PARAM_INTERLACED;
3177                 int width = h->mb.i_mb_width << PARAM_INTERLACED;
3178                 i_slice_num++;
3179                 h->sh.i_last_mb = (height * i_slice_num + round_bias) / h->param.i_slice_count * width - 1;
3180             }
3181         }
3182         h->sh.i_last_mb = X264_MIN( h->sh.i_last_mb, last_thread_mb );
3183         if( slice_write( h ) )
3184             goto fail;
3185         h->sh.i_first_mb = h->sh.i_last_mb + 1;
3186         // if i_first_mb is not the last mb in a row then go to the next mb in MBAFF order
3187         if( SLICE_MBAFF && h->sh.i_first_mb % h->mb.i_mb_width )
3188             h->sh.i_first_mb -= h->mb.i_mb_stride;
3189     }
3190 
3191     return (void *)0;
3192 
3193 fail:
3194     /* Tell other threads we're done, so they wouldn't wait for it */
3195     if( h->param.b_sliced_threads )
3196         x264_threadslice_cond_broadcast( h, 2 );
3197     return (void *)-1;
3198 }
3199 
threaded_slices_write(x264_t * h)3200 static int threaded_slices_write( x264_t *h )
3201 {
3202     int round_bias = h->param.i_avcintra_class ? 0 : h->param.i_slice_count/2;
3203 
3204     /* set first/last mb and sync contexts */
3205     for( int i = 0; i < h->param.i_threads; i++ )
3206     {
3207         x264_t *t = h->thread[i];
3208         if( i )
3209         {
3210             t->param = h->param;
3211             memcpy( &t->i_frame, &h->i_frame, offsetof(x264_t, rc) - offsetof(x264_t, i_frame) );
3212         }
3213         int height = h->mb.i_mb_height >> PARAM_INTERLACED;
3214         t->i_threadslice_start = ((height *  i    + round_bias) / h->param.i_threads) << PARAM_INTERLACED;
3215         t->i_threadslice_end   = ((height * (i+1) + round_bias) / h->param.i_threads) << PARAM_INTERLACED;
3216         t->sh.i_first_mb = t->i_threadslice_start * h->mb.i_mb_width;
3217         t->sh.i_last_mb  =   t->i_threadslice_end * h->mb.i_mb_width - 1;
3218     }
3219 
3220     x264_analyse_weight_frame( h, h->mb.i_mb_height*16 + 16 );
3221 
3222     x264_threads_distribute_ratecontrol( h );
3223 
3224     /* setup */
3225     for( int i = 0; i < h->param.i_threads; i++ )
3226     {
3227         h->thread[i]->i_thread_idx = i;
3228         h->thread[i]->b_thread_active = 1;
3229         x264_threadslice_cond_broadcast( h->thread[i], 0 );
3230     }
3231     /* dispatch */
3232     for( int i = 0; i < h->param.i_threads; i++ )
3233         x264_threadpool_run( h->threadpool, (void*)slices_write, h->thread[i] );
3234     /* wait */
3235     for( int i = 0; i < h->param.i_threads; i++ )
3236         x264_threadslice_cond_wait( h->thread[i], 1 );
3237 
3238     x264_threads_merge_ratecontrol( h );
3239 
3240     for( int i = 1; i < h->param.i_threads; i++ )
3241     {
3242         x264_t *t = h->thread[i];
3243         for( int j = 0; j < t->out.i_nal; j++ )
3244         {
3245             h->out.nal[h->out.i_nal] = t->out.nal[j];
3246             h->out.i_nal++;
3247             nal_check_buffer( h );
3248         }
3249         /* All entries in stat.frame are ints except for ssd/ssim. */
3250         for( size_t j = 0; j < (offsetof(x264_t,stat.frame.i_ssd) - offsetof(x264_t,stat.frame.i_mv_bits)) / sizeof(int); j++ )
3251             ((int*)&h->stat.frame)[j] += ((int*)&t->stat.frame)[j];
3252         for( int j = 0; j < 3; j++ )
3253             h->stat.frame.i_ssd[j] += t->stat.frame.i_ssd[j];
3254         h->stat.frame.f_ssim += t->stat.frame.f_ssim;
3255         h->stat.frame.i_ssim_cnt += t->stat.frame.i_ssim_cnt;
3256     }
3257 
3258     return 0;
3259 }
3260 
x264_encoder_intra_refresh(x264_t * h)3261 void x264_encoder_intra_refresh( x264_t *h )
3262 {
3263     h = h->thread[h->i_thread_phase];
3264     h->b_queued_intra_refresh = 1;
3265 }
3266 
x264_encoder_invalidate_reference(x264_t * h,int64_t pts)3267 int x264_encoder_invalidate_reference( x264_t *h, int64_t pts )
3268 {
3269     if( h->param.i_bframe )
3270     {
3271         x264_log( h, X264_LOG_ERROR, "x264_encoder_invalidate_reference is not supported with B-frames enabled\n" );
3272         return -1;
3273     }
3274     if( h->param.b_intra_refresh )
3275     {
3276         x264_log( h, X264_LOG_ERROR, "x264_encoder_invalidate_reference is not supported with intra refresh enabled\n" );
3277         return -1;
3278     }
3279     h = h->thread[h->i_thread_phase];
3280     if( pts >= h->i_last_idr_pts )
3281     {
3282         for( int i = 0; h->frames.reference[i]; i++ )
3283             if( pts <= h->frames.reference[i]->i_pts )
3284                 h->frames.reference[i]->b_corrupt = 1;
3285         if( pts <= h->fdec->i_pts )
3286             h->fdec->b_corrupt = 1;
3287     }
3288     return 0;
3289 }
3290 
3291 /****************************************************************************
3292  * x264_encoder_encode:
3293  *  XXX: i_poc   : is the poc of the current given picture
3294  *       i_frame : is the number of the frame being coded
3295  *  ex:  type frame poc
3296  *       I      0   2*0
3297  *       P      1   2*3
3298  *       B      2   2*1
3299  *       B      3   2*2
3300  *       P      4   2*6
3301  *       B      5   2*4
3302  *       B      6   2*5
3303  ****************************************************************************/
x264_encoder_encode(x264_t * h,x264_nal_t ** pp_nal,int * pi_nal,x264_picture_t * pic_in,x264_picture_t * pic_out)3304 int     x264_encoder_encode( x264_t *h,
3305                              x264_nal_t **pp_nal, int *pi_nal,
3306                              x264_picture_t *pic_in,
3307                              x264_picture_t *pic_out )
3308 {
3309     x264_t *thread_current, *thread_prev, *thread_oldest;
3310     int i_nal_type, i_nal_ref_idc, i_global_qp;
3311     int overhead = NALU_OVERHEAD;
3312 
3313 #if HAVE_OPENCL
3314     if( h->opencl.b_fatal_error )
3315         return -1;
3316 #endif
3317 
3318     if( h->i_thread_frames > 1 )
3319     {
3320         thread_prev    = h->thread[ h->i_thread_phase ];
3321         h->i_thread_phase = (h->i_thread_phase + 1) % h->i_thread_frames;
3322         thread_current = h->thread[ h->i_thread_phase ];
3323         thread_oldest  = h->thread[ (h->i_thread_phase + 1) % h->i_thread_frames ];
3324         thread_sync_context( thread_current, thread_prev );
3325         x264_thread_sync_ratecontrol( thread_current, thread_prev, thread_oldest );
3326         h = thread_current;
3327     }
3328     else
3329     {
3330         thread_current =
3331         thread_oldest  = h;
3332     }
3333     h->i_cpb_delay_pir_offset = h->i_cpb_delay_pir_offset_next;
3334 
3335     /* no data out */
3336     *pi_nal = 0;
3337     *pp_nal = NULL;
3338 
3339     /* ------------------- Setup new frame from picture -------------------- */
3340     if( pic_in != NULL )
3341     {
3342         if( h->lookahead->b_exit_thread )
3343         {
3344             x264_log( h, X264_LOG_ERROR, "lookahead thread is already stopped\n" );
3345             return -1;
3346         }
3347 
3348         /* 1: Copy the picture to a frame and move it to a buffer */
3349         x264_frame_t *fenc = x264_frame_pop_unused( h, 0 );
3350         if( !fenc )
3351             return -1;
3352 
3353         if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 )
3354             return -1;
3355 
3356         if( h->param.i_width != 16 * h->mb.i_mb_width ||
3357             h->param.i_height != 16 * h->mb.i_mb_height )
3358             x264_frame_expand_border_mod16( h, fenc );
3359 
3360         fenc->i_frame = h->frames.i_input++;
3361 
3362         if( fenc->i_frame == 0 )
3363             h->frames.i_first_pts = fenc->i_pts;
3364         if( h->frames.i_bframe_delay && fenc->i_frame == h->frames.i_bframe_delay )
3365             h->frames.i_bframe_delay_time = fenc->i_pts - h->frames.i_first_pts;
3366 
3367         if( h->param.b_vfr_input && fenc->i_pts <= h->frames.i_largest_pts )
3368             x264_log( h, X264_LOG_WARNING, "non-strictly-monotonic PTS\n" );
3369 
3370         h->frames.i_second_largest_pts = h->frames.i_largest_pts;
3371         h->frames.i_largest_pts = fenc->i_pts;
3372 
3373         if( (fenc->i_pic_struct < PIC_STRUCT_AUTO) || (fenc->i_pic_struct > PIC_STRUCT_TRIPLE) )
3374             fenc->i_pic_struct = PIC_STRUCT_AUTO;
3375 
3376         if( fenc->i_pic_struct == PIC_STRUCT_AUTO )
3377         {
3378 #if HAVE_INTERLACED
3379             int b_interlaced = fenc->param ? fenc->param->b_interlaced : h->param.b_interlaced;
3380 #else
3381             int b_interlaced = 0;
3382 #endif
3383             if( b_interlaced )
3384             {
3385                 int b_tff = fenc->param ? fenc->param->b_tff : h->param.b_tff;
3386                 fenc->i_pic_struct = b_tff ? PIC_STRUCT_TOP_BOTTOM : PIC_STRUCT_BOTTOM_TOP;
3387             }
3388             else
3389                 fenc->i_pic_struct = PIC_STRUCT_PROGRESSIVE;
3390         }
3391 
3392         if( h->param.rc.b_mb_tree && h->param.rc.b_stat_read )
3393         {
3394             if( x264_macroblock_tree_read( h, fenc, pic_in->prop.quant_offsets ) )
3395                 return -1;
3396         }
3397         else
3398             x264_adaptive_quant_frame( h, fenc, pic_in->prop.quant_offsets );
3399 
3400         if( pic_in->prop.quant_offsets_free )
3401             pic_in->prop.quant_offsets_free( pic_in->prop.quant_offsets );
3402 
3403         if( h->frames.b_have_lowres )
3404             x264_frame_init_lowres( h, fenc );
3405 
3406         /* 2: Place the frame into the queue for its slice type decision */
3407         x264_lookahead_put_frame( h, fenc );
3408 
3409         if( h->frames.i_input <= h->frames.i_delay + 1 - h->i_thread_frames )
3410         {
3411             /* Nothing yet to encode, waiting for filling of buffers */
3412             pic_out->i_type = X264_TYPE_AUTO;
3413             return 0;
3414         }
3415     }
3416     else
3417     {
3418         /* signal kills for lookahead thread */
3419         x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
3420         h->lookahead->b_exit_thread = 1;
3421         x264_pthread_cond_broadcast( &h->lookahead->ifbuf.cv_fill );
3422         x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
3423     }
3424 
3425     h->i_frame++;
3426     /* 3: The picture is analyzed in the lookahead */
3427     if( !h->frames.current[0] )
3428         x264_lookahead_get_frames( h );
3429 
3430     if( !h->frames.current[0] && x264_lookahead_is_empty( h ) )
3431         return encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
3432 
3433     /* ------------------- Get frame to be encoded ------------------------- */
3434     /* 4: get picture to encode */
3435     h->fenc = x264_frame_shift( h->frames.current );
3436 
3437     /* If applicable, wait for previous frame reconstruction to finish */
3438     if( h->param.b_sliced_threads )
3439         if( threadpool_wait_all( h ) < 0 )
3440             return -1;
3441 
3442     if( h->i_frame == 0 )
3443         h->i_reordered_pts_delay = h->fenc->i_reordered_pts;
3444     if( h->reconfig )
3445     {
3446         x264_encoder_reconfig_apply( h, &h->reconfig_h->param );
3447         h->reconfig = 0;
3448     }
3449     if( h->fenc->param )
3450     {
3451         x264_encoder_reconfig_apply( h, h->fenc->param );
3452         if( h->fenc->param->param_free )
3453         {
3454             x264_param_cleanup( h->fenc->param );
3455             h->fenc->param->param_free( h->fenc->param );
3456             h->fenc->param = NULL;
3457         }
3458     }
3459     x264_ratecontrol_zone_init( h );
3460 
3461     // ok to call this before encoding any frames, since the initial values of fdec have b_kept_as_ref=0
3462     if( reference_update( h ) )
3463         return -1;
3464     h->fdec->i_lines_completed = -1;
3465 
3466     if( !IS_X264_TYPE_I( h->fenc->i_type ) )
3467     {
3468         int valid_refs_left = 0;
3469         for( int i = 0; h->frames.reference[i]; i++ )
3470             if( !h->frames.reference[i]->b_corrupt )
3471                 valid_refs_left++;
3472         /* No valid reference frames left: force an IDR. */
3473         if( !valid_refs_left )
3474         {
3475             h->fenc->b_keyframe = 1;
3476             h->fenc->i_type = X264_TYPE_IDR;
3477         }
3478     }
3479 
3480     if( h->fenc->b_keyframe )
3481     {
3482         h->frames.i_last_keyframe = h->fenc->i_frame;
3483         if( h->fenc->i_type == X264_TYPE_IDR )
3484         {
3485             h->i_frame_num = 0;
3486             h->frames.i_last_idr = h->fenc->i_frame;
3487         }
3488     }
3489     h->sh.i_mmco_command_count =
3490     h->sh.i_mmco_remove_from_end = 0;
3491     h->b_ref_reorder[0] =
3492     h->b_ref_reorder[1] = 0;
3493     h->fdec->i_poc =
3494     h->fenc->i_poc = 2 * ( h->fenc->i_frame - X264_MAX( h->frames.i_last_idr, 0 ) );
3495 
3496     /* ------------------- Setup frame context ----------------------------- */
3497     /* 5: Init data dependent of frame type */
3498     if( h->fenc->i_type == X264_TYPE_IDR )
3499     {
3500         /* reset ref pictures */
3501         i_nal_type    = NAL_SLICE_IDR;
3502         i_nal_ref_idc = NAL_PRIORITY_HIGHEST;
3503         h->sh.i_type = SLICE_TYPE_I;
3504         reference_reset( h );
3505         h->frames.i_poc_last_open_gop = -1;
3506     }
3507     else if( h->fenc->i_type == X264_TYPE_I )
3508     {
3509         i_nal_type    = NAL_SLICE;
3510         i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
3511         h->sh.i_type = SLICE_TYPE_I;
3512         reference_hierarchy_reset( h );
3513         if( h->param.b_open_gop )
3514             h->frames.i_poc_last_open_gop = h->fenc->b_keyframe ? h->fenc->i_poc : -1;
3515     }
3516     else if( h->fenc->i_type == X264_TYPE_P )
3517     {
3518         i_nal_type    = NAL_SLICE;
3519         i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
3520         h->sh.i_type = SLICE_TYPE_P;
3521         reference_hierarchy_reset( h );
3522         h->frames.i_poc_last_open_gop = -1;
3523     }
3524     else if( h->fenc->i_type == X264_TYPE_BREF )
3525     {
3526         i_nal_type    = NAL_SLICE;
3527         i_nal_ref_idc = h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT ? NAL_PRIORITY_LOW : NAL_PRIORITY_HIGH;
3528         h->sh.i_type = SLICE_TYPE_B;
3529         reference_hierarchy_reset( h );
3530     }
3531     else    /* B frame */
3532     {
3533         i_nal_type    = NAL_SLICE;
3534         i_nal_ref_idc = NAL_PRIORITY_DISPOSABLE;
3535         h->sh.i_type = SLICE_TYPE_B;
3536     }
3537 
3538     h->fdec->i_type = h->fenc->i_type;
3539     h->fdec->i_frame = h->fenc->i_frame;
3540     h->fenc->b_kept_as_ref =
3541     h->fdec->b_kept_as_ref = i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE && h->param.i_keyint_max > 1;
3542 
3543     h->fdec->mb_info = h->fenc->mb_info;
3544     h->fdec->mb_info_free = h->fenc->mb_info_free;
3545     h->fenc->mb_info = NULL;
3546     h->fenc->mb_info_free = NULL;
3547 
3548     h->fdec->i_pts = h->fenc->i_pts;
3549     if( h->frames.i_bframe_delay )
3550     {
3551         int64_t *prev_reordered_pts = thread_current->frames.i_prev_reordered_pts;
3552         h->fdec->i_dts = h->i_frame > h->frames.i_bframe_delay
3553                        ? prev_reordered_pts[ (h->i_frame - h->frames.i_bframe_delay) % h->frames.i_bframe_delay ]
3554                        : h->fenc->i_reordered_pts - h->frames.i_bframe_delay_time;
3555         prev_reordered_pts[ h->i_frame % h->frames.i_bframe_delay ] = h->fenc->i_reordered_pts;
3556     }
3557     else
3558         h->fdec->i_dts = h->fenc->i_reordered_pts;
3559     if( h->fenc->i_type == X264_TYPE_IDR )
3560         h->i_last_idr_pts = h->fdec->i_pts;
3561 
3562     /* ------------------- Init                ----------------------------- */
3563     /* build ref list 0/1 */
3564     reference_build_list( h, h->fdec->i_poc );
3565 
3566     /* ---------------------- Write the bitstream -------------------------- */
3567     /* Init bitstream context */
3568     if( h->param.b_sliced_threads )
3569     {
3570         for( int i = 0; i < h->param.i_threads; i++ )
3571         {
3572             bs_init( &h->thread[i]->out.bs, h->thread[i]->out.p_bitstream, h->thread[i]->out.i_bitstream );
3573             h->thread[i]->out.i_nal = 0;
3574         }
3575     }
3576     else
3577     {
3578         bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
3579         h->out.i_nal = 0;
3580     }
3581 
3582     if( h->param.b_aud )
3583     {
3584         int pic_type;
3585 
3586         if( h->sh.i_type == SLICE_TYPE_I )
3587             pic_type = 0;
3588         else if( h->sh.i_type == SLICE_TYPE_P )
3589             pic_type = 1;
3590         else if( h->sh.i_type == SLICE_TYPE_B )
3591             pic_type = 2;
3592         else
3593             pic_type = 7;
3594 
3595         nal_start( h, NAL_AUD, NAL_PRIORITY_DISPOSABLE );
3596         bs_write( &h->out.bs, 3, pic_type );
3597         bs_rbsp_trailing( &h->out.bs );
3598         bs_flush( &h->out.bs );
3599         if( nal_end( h ) )
3600             return -1;
3601         overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3602     }
3603 
3604     h->i_nal_type = i_nal_type;
3605     h->i_nal_ref_idc = i_nal_ref_idc;
3606 
3607     if( h->param.b_intra_refresh )
3608     {
3609         if( IS_X264_TYPE_I( h->fenc->i_type ) )
3610         {
3611             h->fdec->i_frames_since_pir = 0;
3612             h->b_queued_intra_refresh = 0;
3613             /* PIR is currently only supported with ref == 1, so any intra frame effectively refreshes
3614              * the whole frame and counts as an intra refresh. */
3615             h->fdec->f_pir_position = h->mb.i_mb_width;
3616         }
3617         else if( h->fenc->i_type == X264_TYPE_P )
3618         {
3619             int pocdiff = (h->fdec->i_poc - h->fref[0][0]->i_poc)/2;
3620             float increment = X264_MAX( ((float)h->mb.i_mb_width-1) / h->param.i_keyint_max, 1 );
3621             h->fdec->f_pir_position = h->fref[0][0]->f_pir_position;
3622             h->fdec->i_frames_since_pir = h->fref[0][0]->i_frames_since_pir + pocdiff;
3623             if( h->fdec->i_frames_since_pir >= h->param.i_keyint_max ||
3624                 (h->b_queued_intra_refresh && h->fdec->f_pir_position + 0.5 >= h->mb.i_mb_width) )
3625             {
3626                 h->fdec->f_pir_position = 0;
3627                 h->fdec->i_frames_since_pir = 0;
3628                 h->b_queued_intra_refresh = 0;
3629                 h->fenc->b_keyframe = 1;
3630             }
3631             h->fdec->i_pir_start_col = h->fdec->f_pir_position+0.5;
3632             h->fdec->f_pir_position += increment * pocdiff;
3633             h->fdec->i_pir_end_col = h->fdec->f_pir_position+0.5;
3634             /* If our intra refresh has reached the right side of the frame, we're done. */
3635             if( h->fdec->i_pir_end_col >= h->mb.i_mb_width - 1 )
3636             {
3637                 h->fdec->f_pir_position = h->mb.i_mb_width;
3638                 h->fdec->i_pir_end_col = h->mb.i_mb_width - 1;
3639             }
3640         }
3641     }
3642 
3643     if( h->fenc->b_keyframe )
3644     {
3645         /* Write SPS and PPS */
3646         if( h->param.b_repeat_headers )
3647         {
3648             /* generate sequence parameters */
3649             nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST );
3650             x264_sps_write( &h->out.bs, h->sps );
3651             if( nal_end( h ) )
3652                 return -1;
3653             /* Pad AUD/SPS to 256 bytes like Panasonic */
3654             if( h->param.i_avcintra_class )
3655                 h->out.nal[h->out.i_nal-1].i_padding = 256 - bs_pos( &h->out.bs ) / 8 - 2*NALU_OVERHEAD;
3656             overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + NALU_OVERHEAD;
3657 
3658             /* generate picture parameters */
3659             nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
3660             x264_pps_write( &h->out.bs, h->sps, h->pps );
3661             if( nal_end( h ) )
3662                 return -1;
3663             if( h->param.i_avcintra_class )
3664             {
3665                 int total_len = 256;
3666                 /* Sony XAVC uses an oversized PPS instead of SEI padding */
3667                 if( h->param.i_avcintra_flavor == X264_AVCINTRA_FLAVOR_SONY )
3668                     total_len += h->param.i_height == 1080 ? 18*512 : 10*512;
3669                 h->out.nal[h->out.i_nal-1].i_padding = total_len - h->out.nal[h->out.i_nal-1].i_payload - NALU_OVERHEAD;
3670             }
3671             overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + NALU_OVERHEAD;
3672         }
3673 
3674         /* when frame threading is used, buffering period sei is written in encoder_frame_end */
3675         if( h->i_thread_frames == 1 && h->sps->vui.b_nal_hrd_parameters_present )
3676         {
3677             x264_hrd_fullness( h );
3678             nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3679             x264_sei_buffering_period_write( h, &h->out.bs );
3680             if( nal_end( h ) )
3681                return -1;
3682             overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3683         }
3684     }
3685 
3686     /* write extra sei */
3687     for( int i = 0; i < h->fenc->extra_sei.num_payloads; i++ )
3688     {
3689         nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3690         x264_sei_write( &h->out.bs, h->fenc->extra_sei.payloads[i].payload, h->fenc->extra_sei.payloads[i].payload_size,
3691                         h->fenc->extra_sei.payloads[i].payload_type );
3692         if( nal_end( h ) )
3693             return -1;
3694         overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3695         if( h->fenc->extra_sei.sei_free )
3696         {
3697             h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads[i].payload );
3698             h->fenc->extra_sei.payloads[i].payload = NULL;
3699         }
3700     }
3701 
3702     if( h->fenc->extra_sei.sei_free )
3703     {
3704         h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads );
3705         h->fenc->extra_sei.payloads = NULL;
3706         h->fenc->extra_sei.sei_free = NULL;
3707     }
3708 
3709     if( h->fenc->b_keyframe )
3710     {
3711         /* Avid's decoder strictly wants two SEIs for AVC-Intra so we can't insert the x264 SEI */
3712         if( h->param.b_repeat_headers && h->fenc->i_frame == 0 && !h->param.i_avcintra_class )
3713         {
3714             /* identify ourself */
3715             nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3716             if( x264_sei_version_write( h, &h->out.bs ) )
3717                 return -1;
3718             if( nal_end( h ) )
3719                 return -1;
3720             overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3721         }
3722 
3723         if( h->fenc->i_type != X264_TYPE_IDR )
3724         {
3725             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;
3726             nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3727             x264_sei_recovery_point_write( h, &h->out.bs, time_to_recovery );
3728             if( nal_end( h ) )
3729                 return -1;
3730             overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3731         }
3732 
3733         if( h->param.mastering_display.b_mastering_display )
3734         {
3735             nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3736             x264_sei_mastering_display_write( h, &h->out.bs );
3737             if( nal_end( h ) )
3738                 return -1;
3739             overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3740         }
3741 
3742         if( h->param.content_light_level.b_cll )
3743         {
3744             nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3745             x264_sei_content_light_level_write( h, &h->out.bs );
3746             if( nal_end( h ) )
3747                 return -1;
3748             overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3749         }
3750 
3751         if( h->param.i_alternative_transfer != 2 )
3752         {
3753             nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3754             x264_sei_alternative_transfer_write( h, &h->out.bs );
3755             if( nal_end( h ) )
3756                 return -1;
3757             overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3758         }
3759     }
3760 
3761     if( h->param.i_frame_packing >= 0 && (h->fenc->b_keyframe || h->param.i_frame_packing == 5) )
3762     {
3763         nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3764         x264_sei_frame_packing_write( h, &h->out.bs );
3765         if( nal_end( h ) )
3766             return -1;
3767         overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3768     }
3769 
3770     /* generate sei pic timing */
3771     if( h->sps->vui.b_pic_struct_present || h->sps->vui.b_nal_hrd_parameters_present )
3772     {
3773         nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3774         x264_sei_pic_timing_write( h, &h->out.bs );
3775         if( nal_end( h ) )
3776             return -1;
3777         overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3778     }
3779 
3780     /* As required by Blu-ray. */
3781     if( !IS_X264_TYPE_B( h->fenc->i_type ) && h->b_sh_backup )
3782     {
3783         h->b_sh_backup = 0;
3784         nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3785         x264_sei_dec_ref_pic_marking_write( h, &h->out.bs );
3786         if( nal_end( h ) )
3787             return -1;
3788         overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3789     }
3790 
3791     if( h->fenc->b_keyframe && h->param.b_intra_refresh )
3792         h->i_cpb_delay_pir_offset_next = h->fenc->i_cpb_delay;
3793 
3794     /* Filler space: 10 or 18 SEIs' worth of space, depending on resolution */
3795     if( h->param.i_avcintra_class && h->param.i_avcintra_flavor != X264_AVCINTRA_FLAVOR_SONY )
3796     {
3797         /* Write an empty filler NAL to mimic the AUD in the P2 format*/
3798         nal_start( h, NAL_FILLER, NAL_PRIORITY_DISPOSABLE );
3799         x264_filler_write( h, &h->out.bs, 0 );
3800         if( nal_end( h ) )
3801             return -1;
3802         overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
3803 
3804         /* All lengths are magic lengths that decoders expect to see */
3805         /* "UMID" SEI */
3806         nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3807         if( x264_sei_avcintra_umid_write( h, &h->out.bs ) < 0 )
3808             return -1;
3809         if( nal_end( h ) )
3810             return -1;
3811         overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
3812 
3813         int unpadded_len;
3814         int total_len;
3815         if( h->param.i_height == 1080 )
3816         {
3817             unpadded_len = 5780;
3818             total_len = 17*512;
3819         }
3820         else
3821         {
3822             unpadded_len = 2900;
3823             total_len = 9*512;
3824         }
3825         /* "VANC" SEI */
3826         nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3827         if( x264_sei_avcintra_vanc_write( h, &h->out.bs, unpadded_len ) < 0 )
3828             return -1;
3829         if( nal_end( h ) )
3830             return -1;
3831 
3832         h->out.nal[h->out.i_nal-1].i_padding = total_len - h->out.nal[h->out.i_nal-1].i_payload - SEI_OVERHEAD;
3833         overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + SEI_OVERHEAD;
3834     }
3835 
3836     /* Init the rate control */
3837     /* FIXME: Include slice header bit cost. */
3838     x264_ratecontrol_start( h, h->fenc->i_qpplus1, overhead*8 );
3839     i_global_qp = x264_ratecontrol_qp( h );
3840 
3841     pic_out->i_qpplus1 =
3842     h->fdec->i_qpplus1 = i_global_qp + 1;
3843 
3844     if( h->param.rc.b_stat_read && h->sh.i_type != SLICE_TYPE_I )
3845     {
3846         x264_reference_build_list_optimal( h );
3847         reference_check_reorder( h );
3848     }
3849 
3850     if( h->i_ref[0] )
3851         h->fdec->i_poc_l0ref0 = h->fref[0][0]->i_poc;
3852 
3853     /* ------------------------ Create slice header  ----------------------- */
3854     slice_init( h, i_nal_type, i_global_qp );
3855 
3856     /*------------------------- Weights -------------------------------------*/
3857     if( h->sh.i_type == SLICE_TYPE_B )
3858         x264_macroblock_bipred_init( h );
3859 
3860     weighted_pred_init( h );
3861 
3862     if( i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE )
3863         h->i_frame_num++;
3864 
3865     /* Write frame */
3866     h->i_threadslice_start = 0;
3867     h->i_threadslice_end = h->mb.i_mb_height;
3868     if( h->i_thread_frames > 1 )
3869     {
3870         x264_threadpool_run( h->threadpool, (void*)slices_write, h );
3871         h->b_thread_active = 1;
3872     }
3873     else if( h->param.b_sliced_threads )
3874     {
3875         if( threaded_slices_write( h ) )
3876             return -1;
3877     }
3878     else
3879         if( (intptr_t)slices_write( h ) )
3880             return -1;
3881 
3882     return encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
3883 }
3884 
encoder_frame_end(x264_t * h,x264_t * thread_current,x264_nal_t ** pp_nal,int * pi_nal,x264_picture_t * pic_out)3885 static int encoder_frame_end( x264_t *h, x264_t *thread_current,
3886                               x264_nal_t **pp_nal, int *pi_nal,
3887                               x264_picture_t *pic_out )
3888 {
3889     char psz_message[80];
3890 
3891     if( !h->param.b_sliced_threads && h->b_thread_active )
3892     {
3893         h->b_thread_active = 0;
3894         if( (intptr_t)x264_threadpool_wait( h->threadpool, h ) )
3895             return -1;
3896     }
3897     if( !h->out.i_nal )
3898     {
3899         pic_out->i_type = X264_TYPE_AUTO;
3900         return 0;
3901     }
3902 
3903     x264_emms();
3904 
3905     /* generate buffering period sei and insert it into place */
3906     if( h->i_thread_frames > 1 && h->fenc->b_keyframe && h->sps->vui.b_nal_hrd_parameters_present )
3907     {
3908         x264_hrd_fullness( h );
3909         nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
3910         x264_sei_buffering_period_write( h, &h->out.bs );
3911         if( nal_end( h ) )
3912            return -1;
3913         /* buffering period sei must follow AUD, SPS and PPS and precede all other SEIs */
3914         int idx = 0;
3915         while( h->out.nal[idx].i_type == NAL_AUD ||
3916                h->out.nal[idx].i_type == NAL_SPS ||
3917                h->out.nal[idx].i_type == NAL_PPS )
3918             idx++;
3919         x264_nal_t nal_tmp = h->out.nal[h->out.i_nal-1];
3920         memmove( &h->out.nal[idx+1], &h->out.nal[idx], (h->out.i_nal-idx-1)*sizeof(x264_nal_t) );
3921         h->out.nal[idx] = nal_tmp;
3922     }
3923 
3924     int frame_size = encoder_encapsulate_nals( h, 0 );
3925     if( frame_size < 0 )
3926         return -1;
3927 
3928     /* Set output picture properties */
3929     pic_out->i_type = h->fenc->i_type;
3930 
3931     pic_out->b_keyframe = h->fenc->b_keyframe;
3932     pic_out->i_pic_struct = h->fenc->i_pic_struct;
3933 
3934     pic_out->i_pts = h->fdec->i_pts;
3935     pic_out->i_dts = h->fdec->i_dts;
3936 
3937     if( pic_out->i_pts < pic_out->i_dts )
3938         x264_log( h, X264_LOG_WARNING, "invalid DTS: PTS is less than DTS\n" );
3939 
3940     pic_out->opaque = h->fenc->opaque;
3941 
3942     pic_out->img.i_csp = h->fdec->i_csp;
3943 #if HIGH_BIT_DEPTH
3944     pic_out->img.i_csp |= X264_CSP_HIGH_DEPTH;
3945 #endif
3946     pic_out->img.i_plane = h->fdec->i_plane;
3947     for( int i = 0; i < pic_out->img.i_plane; i++ )
3948     {
3949         pic_out->img.i_stride[i] = h->fdec->i_stride[i] * SIZEOF_PIXEL;
3950         pic_out->img.plane[i] = (uint8_t*)h->fdec->plane[i];
3951     }
3952 
3953     x264_frame_push_unused( thread_current, h->fenc );
3954 
3955     /* ---------------------- Update encoder state ------------------------- */
3956 
3957     /* update rc */
3958     int filler = 0;
3959     if( x264_ratecontrol_end( h, frame_size * 8, &filler ) < 0 )
3960         return -1;
3961 
3962     pic_out->hrd_timing = h->fenc->hrd_timing;
3963     pic_out->prop.f_crf_avg = h->fdec->f_crf_avg;
3964 
3965     /* Filler in AVC-Intra mode is written as zero bytes to the last slice
3966      * We don't know the size of the last slice until encapsulation so we add filler to the encapsulated NAL */
3967     if( h->param.i_avcintra_class )
3968     {
3969         if( check_encapsulated_buffer( h, h->thread[0], h->out.i_nal, frame_size, (int64_t)frame_size + filler ) < 0 )
3970             return -1;
3971 
3972         x264_nal_t *nal = &h->out.nal[h->out.i_nal-1];
3973         memset( nal->p_payload + nal->i_payload, 0, filler );
3974         nal->i_payload += filler;
3975         nal->i_padding = filler;
3976         frame_size += filler;
3977 
3978         /* Fix up the size header for mp4/etc */
3979         if( !h->param.b_annexb )
3980         {
3981             /* Size doesn't include the size of the header we're writing now. */
3982             uint8_t *nal_data = nal->p_payload;
3983             int chunk_size = nal->i_payload - 4;
3984             nal_data[0] = chunk_size >> 24;
3985             nal_data[1] = chunk_size >> 16;
3986             nal_data[2] = chunk_size >> 8;
3987             nal_data[3] = chunk_size >> 0;
3988         }
3989     }
3990     else
3991     {
3992         while( filler > 0 )
3993         {
3994             int f, overhead = FILLER_OVERHEAD - h->param.b_annexb;
3995             if( h->param.i_slice_max_size && filler > h->param.i_slice_max_size )
3996             {
3997                 int next_size = filler - h->param.i_slice_max_size;
3998                 int overflow = X264_MAX( overhead - next_size, 0 );
3999                 f = h->param.i_slice_max_size - overhead - overflow;
4000             }
4001             else
4002                 f = X264_MAX( 0, filler - overhead );
4003 
4004             if( bitstream_check_buffer_filler( h, f ) )
4005                 return -1;
4006             nal_start( h, NAL_FILLER, NAL_PRIORITY_DISPOSABLE );
4007             x264_filler_write( h, &h->out.bs, f );
4008             if( nal_end( h ) )
4009                 return -1;
4010             int total_size = encoder_encapsulate_nals( h, h->out.i_nal-1 );
4011             if( total_size < 0 )
4012                 return -1;
4013             frame_size += total_size;
4014             filler -= total_size;
4015         }
4016     }
4017 
4018     /* End bitstream, set output  */
4019     *pi_nal = h->out.i_nal;
4020     *pp_nal = h->out.nal;
4021 
4022     h->out.i_nal = 0;
4023 
4024     x264_noise_reduction_update( h );
4025 
4026     /* ---------------------- Compute/Print statistics --------------------- */
4027     thread_sync_stat( h, h->thread[0] );
4028 
4029     /* Slice stat */
4030     h->stat.i_frame_count[h->sh.i_type]++;
4031     h->stat.i_frame_size[h->sh.i_type] += frame_size;
4032     h->stat.f_frame_qp[h->sh.i_type] += h->fdec->f_qp_avg_aq;
4033 
4034     for( int i = 0; i < X264_MBTYPE_MAX; i++ )
4035         h->stat.i_mb_count[h->sh.i_type][i] += h->stat.frame.i_mb_count[i];
4036     for( int i = 0; i < 2; i++ )
4037         h->stat.i_mb_count_8x8dct[i] += h->stat.frame.i_mb_count_8x8dct[i];
4038     for( int i = 0; i < 6; i++ )
4039         h->stat.i_mb_cbp[i] += h->stat.frame.i_mb_cbp[i];
4040     for( int i = 0; i < 4; i++ )
4041         for( int j = 0; j < 13; j++ )
4042             h->stat.i_mb_pred_mode[i][j] += h->stat.frame.i_mb_pred_mode[i][j];
4043     if( h->sh.i_type != SLICE_TYPE_I )
4044     {
4045         for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
4046             h->stat.i_mb_partition[h->sh.i_type][i] += h->stat.frame.i_mb_partition[i];
4047         for( int i_list = 0; i_list < 2; i_list++ )
4048             for( int i = 0; i < X264_REF_MAX*2; i++ )
4049                 h->stat.i_mb_count_ref[h->sh.i_type][i_list][i] += h->stat.frame.i_mb_count_ref[i_list][i];
4050     }
4051     for( int i = 0; i < 3; i++ )
4052         h->stat.i_mb_field[i] += h->stat.frame.i_mb_field[i];
4053     if( h->sh.i_type == SLICE_TYPE_P && h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE )
4054     {
4055         h->stat.i_wpred[0] += !!h->sh.weight[0][0].weightfn;
4056         h->stat.i_wpred[1] += !!h->sh.weight[0][1].weightfn || !!h->sh.weight[0][2].weightfn;
4057     }
4058     if( h->sh.i_type == SLICE_TYPE_B )
4059     {
4060         h->stat.i_direct_frames[ h->sh.b_direct_spatial_mv_pred ] ++;
4061         if( h->mb.b_direct_auto_write )
4062         {
4063             //FIXME somewhat arbitrary time constants
4064             if( h->stat.i_direct_score[0] + h->stat.i_direct_score[1] > h->mb.i_mb_count )
4065                 for( int i = 0; i < 2; i++ )
4066                     h->stat.i_direct_score[i] = h->stat.i_direct_score[i] * 9/10;
4067             for( int i = 0; i < 2; i++ )
4068                 h->stat.i_direct_score[i] += h->stat.frame.i_direct_score[i];
4069         }
4070     }
4071     else
4072         h->stat.i_consecutive_bframes[h->fenc->i_bframes]++;
4073 
4074     psz_message[0] = '\0';
4075     double dur = h->fenc->f_duration;
4076     h->stat.f_frame_duration[h->sh.i_type] += dur;
4077     if( h->param.analyse.b_psnr )
4078     {
4079         int64_t ssd[3] =
4080         {
4081             h->stat.frame.i_ssd[0],
4082             h->stat.frame.i_ssd[1],
4083             h->stat.frame.i_ssd[2],
4084         };
4085         int luma_size = h->param.i_width * h->param.i_height;
4086         int chroma_size = CHROMA_SIZE( luma_size );
4087         pic_out->prop.f_psnr[0] = calc_psnr( ssd[0], luma_size );
4088         pic_out->prop.f_psnr[1] = calc_psnr( ssd[1], chroma_size );
4089         pic_out->prop.f_psnr[2] = calc_psnr( ssd[2], chroma_size );
4090         pic_out->prop.f_psnr_avg = calc_psnr( ssd[0] + ssd[1] + ssd[2], luma_size + chroma_size*2 );
4091 
4092         h->stat.f_ssd_global[h->sh.i_type]   += dur * (ssd[0] + ssd[1] + ssd[2]);
4093         h->stat.f_psnr_average[h->sh.i_type] += dur * pic_out->prop.f_psnr_avg;
4094         h->stat.f_psnr_mean_y[h->sh.i_type]  += dur * pic_out->prop.f_psnr[0];
4095         h->stat.f_psnr_mean_u[h->sh.i_type]  += dur * pic_out->prop.f_psnr[1];
4096         h->stat.f_psnr_mean_v[h->sh.i_type]  += dur * pic_out->prop.f_psnr[2];
4097 
4098         snprintf( psz_message, 80, " PSNR Y:%5.2f U:%5.2f V:%5.2f", pic_out->prop.f_psnr[0],
4099                                                                     pic_out->prop.f_psnr[1],
4100                                                                     pic_out->prop.f_psnr[2] );
4101     }
4102 
4103     if( h->param.analyse.b_ssim )
4104     {
4105         pic_out->prop.f_ssim = h->stat.frame.f_ssim / h->stat.frame.i_ssim_cnt;
4106         h->stat.f_ssim_mean_y[h->sh.i_type] += pic_out->prop.f_ssim * dur;
4107         int msg_len = strlen(psz_message);
4108         snprintf( psz_message + msg_len, 80 - msg_len, " SSIM Y:%.5f", pic_out->prop.f_ssim );
4109     }
4110     psz_message[79] = '\0';
4111 
4112     x264_log( h, X264_LOG_DEBUG,
4113               "frame=%4d QP=%.2f NAL=%d Slice:%c Poc:%-3d I:%-4d P:%-4d SKIP:%-4d size=%d bytes%s\n",
4114               h->i_frame,
4115               h->fdec->f_qp_avg_aq,
4116               h->i_nal_ref_idc,
4117               h->sh.i_type == SLICE_TYPE_I ? 'I' : (h->sh.i_type == SLICE_TYPE_P ? 'P' : 'B' ),
4118               h->fdec->i_poc,
4119               h->stat.frame.i_mb_count_i,
4120               h->stat.frame.i_mb_count_p,
4121               h->stat.frame.i_mb_count_skip,
4122               frame_size,
4123               psz_message );
4124 
4125     // keep stats all in one place
4126     thread_sync_stat( h->thread[0], h );
4127     // for the use of the next frame
4128     thread_sync_stat( thread_current, h );
4129 
4130 #ifdef DEBUG_MB_TYPE
4131 {
4132     static const char mb_chars[] = { 'i', 'i', 'I', 'C', 'P', '8', 'S',
4133         'D', '<', 'X', 'B', 'X', '>', 'B', 'B', 'B', 'B', '8', 'S' };
4134     for( int mb_xy = 0; mb_xy < h->mb.i_mb_width * h->mb.i_mb_height; mb_xy++ )
4135     {
4136         if( h->mb.type[mb_xy] < X264_MBTYPE_MAX && h->mb.type[mb_xy] >= 0 )
4137             fprintf( stderr, "%c ", mb_chars[ h->mb.type[mb_xy] ] );
4138         else
4139             fprintf( stderr, "? " );
4140 
4141         if( (mb_xy+1) % h->mb.i_mb_width == 0 )
4142             fprintf( stderr, "\n" );
4143     }
4144 }
4145 #endif
4146 
4147     /* Remove duplicates, must be done near the end as breaks h->fref0 array
4148      * by freeing some of its pointers. */
4149     for( int i = 0; i < h->i_ref[0]; i++ )
4150         if( h->fref[0][i] && h->fref[0][i]->b_duplicate )
4151         {
4152             x264_frame_push_blank_unused( h, h->fref[0][i] );
4153             h->fref[0][i] = 0;
4154         }
4155 
4156     if( h->param.psz_dump_yuv )
4157         frame_dump( h );
4158     x264_emms();
4159 
4160     return frame_size;
4161 }
4162 
print_intra(int64_t * i_mb_count,double i_count,int b_print_pcm,char * intra)4163 static void print_intra( int64_t *i_mb_count, double i_count, int b_print_pcm, char *intra )
4164 {
4165     intra += sprintf( intra, "I16..4%s: %4.1f%% %4.1f%% %4.1f%%",
4166         b_print_pcm ? "..PCM" : "",
4167         i_mb_count[I_16x16]/ i_count,
4168         i_mb_count[I_8x8]  / i_count,
4169         i_mb_count[I_4x4]  / i_count );
4170     if( b_print_pcm )
4171         sprintf( intra, " %4.1f%%", i_mb_count[I_PCM]  / i_count );
4172 }
4173 
4174 /****************************************************************************
4175  * x264_encoder_close:
4176  ****************************************************************************/
x264_encoder_close(x264_t * h)4177 void    x264_encoder_close  ( x264_t *h )
4178 {
4179     int64_t i_yuv_size = FRAME_SIZE( h->param.i_width * h->param.i_height );
4180     int64_t i_mb_count_size[2][7] = {{0}};
4181     char buf[200];
4182     int b_print_pcm = h->stat.i_mb_count[SLICE_TYPE_I][I_PCM]
4183                    || h->stat.i_mb_count[SLICE_TYPE_P][I_PCM]
4184                    || h->stat.i_mb_count[SLICE_TYPE_B][I_PCM];
4185 
4186     x264_lookahead_delete( h );
4187 
4188 #if HAVE_OPENCL
4189     x264_opencl_lookahead_delete( h );
4190     x264_opencl_function_t *ocl = h->opencl.ocl;
4191 #endif
4192 
4193     if( h->param.b_sliced_threads )
4194         threadpool_wait_all( h );
4195     if( h->param.i_threads > 1 )
4196         x264_threadpool_delete( h->threadpool );
4197     if( h->param.i_lookahead_threads > 1 )
4198         x264_threadpool_delete( h->lookaheadpool );
4199     if( h->i_thread_frames > 1 )
4200     {
4201         for( int i = 0; i < h->i_thread_frames; i++ )
4202             if( h->thread[i]->b_thread_active )
4203             {
4204                 assert( h->thread[i]->fenc->i_reference_count == 1 );
4205                 x264_frame_delete( h->thread[i]->fenc );
4206             }
4207 
4208         x264_t *thread_prev = h->thread[h->i_thread_phase];
4209         x264_thread_sync_ratecontrol( h, thread_prev, h );
4210         x264_thread_sync_ratecontrol( thread_prev, thread_prev, h );
4211         h->i_frame = thread_prev->i_frame + 1 - h->i_thread_frames;
4212     }
4213     h->i_frame++;
4214 
4215     /* Slices used and PSNR */
4216     for( int i = 0; i < 3; i++ )
4217     {
4218         static const uint8_t slice_order[] = { SLICE_TYPE_I, SLICE_TYPE_P, SLICE_TYPE_B };
4219         int i_slice = slice_order[i];
4220 
4221         if( h->stat.i_frame_count[i_slice] > 0 )
4222         {
4223             int i_count = h->stat.i_frame_count[i_slice];
4224             double dur =  h->stat.f_frame_duration[i_slice];
4225             if( h->param.analyse.b_psnr )
4226             {
4227                 x264_log( h, X264_LOG_INFO,
4228                           "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",
4229                           slice_type_to_char[i_slice],
4230                           i_count,
4231                           h->stat.f_frame_qp[i_slice] / i_count,
4232                           (double)h->stat.i_frame_size[i_slice] / i_count,
4233                           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,
4234                           h->stat.f_psnr_average[i_slice] / dur,
4235                           calc_psnr( h->stat.f_ssd_global[i_slice], dur * i_yuv_size ) );
4236             }
4237             else
4238             {
4239                 x264_log( h, X264_LOG_INFO,
4240                           "frame %c:%-5d Avg QP:%5.2f  size:%6.0f\n",
4241                           slice_type_to_char[i_slice],
4242                           i_count,
4243                           h->stat.f_frame_qp[i_slice] / i_count,
4244                           (double)h->stat.i_frame_size[i_slice] / i_count );
4245             }
4246         }
4247     }
4248     if( h->param.i_bframe && h->stat.i_frame_count[SLICE_TYPE_B] )
4249     {
4250         char *p = buf;
4251         int den = 0;
4252         // weight by number of frames (including the I/P-frames) that are in a sequence of N B-frames
4253         for( int i = 0; i <= h->param.i_bframe; i++ )
4254             den += (i+1) * h->stat.i_consecutive_bframes[i];
4255         for( int i = 0; i <= h->param.i_bframe; i++ )
4256             p += sprintf( p, " %4.1f%%", 100. * (i+1) * h->stat.i_consecutive_bframes[i] / den );
4257         x264_log( h, X264_LOG_INFO, "consecutive B-frames:%s\n", buf );
4258     }
4259 
4260     for( int i_type = 0; i_type < 2; i_type++ )
4261         for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
4262         {
4263             if( i == D_DIRECT_8x8 ) continue; /* direct is counted as its own type */
4264             i_mb_count_size[i_type][x264_mb_partition_pixel_table[i]] += h->stat.i_mb_partition[i_type][i];
4265         }
4266 
4267     /* MB types used */
4268     if( h->stat.i_frame_count[SLICE_TYPE_I] > 0 )
4269     {
4270         int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_I];
4271         double i_count = (double)h->stat.i_frame_count[SLICE_TYPE_I] * h->mb.i_mb_count / 100.0;
4272         print_intra( i_mb_count, i_count, b_print_pcm, buf );
4273         x264_log( h, X264_LOG_INFO, "mb I  %s\n", buf );
4274     }
4275     if( h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
4276     {
4277         int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_P];
4278         double i_count = (double)h->stat.i_frame_count[SLICE_TYPE_P] * h->mb.i_mb_count / 100.0;
4279         int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_P];
4280         print_intra( i_mb_count, i_count, b_print_pcm, buf );
4281         x264_log( h, X264_LOG_INFO,
4282                   "mb P  %s  P16..4: %4.1f%% %4.1f%% %4.1f%% %4.1f%% %4.1f%%    skip:%4.1f%%\n",
4283                   buf,
4284                   i_mb_size[PIXEL_16x16] / (i_count*4),
4285                   (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
4286                   i_mb_size[PIXEL_8x8] / (i_count*4),
4287                   (i_mb_size[PIXEL_8x4] + i_mb_size[PIXEL_4x8]) / (i_count*4),
4288                   i_mb_size[PIXEL_4x4] / (i_count*4),
4289                   i_mb_count[P_SKIP] / i_count );
4290     }
4291     if( h->stat.i_frame_count[SLICE_TYPE_B] > 0 )
4292     {
4293         int64_t *i_mb_count = h->stat.i_mb_count[SLICE_TYPE_B];
4294         double i_count = (double)h->stat.i_frame_count[SLICE_TYPE_B] * h->mb.i_mb_count / 100.0;
4295         double i_mb_list_count;
4296         int64_t *i_mb_size = i_mb_count_size[SLICE_TYPE_B];
4297         int64_t list_count[3] = {0}; /* 0 == L0, 1 == L1, 2 == BI */
4298         print_intra( i_mb_count, i_count, b_print_pcm, buf );
4299         for( int i = 0; i < X264_PARTTYPE_MAX; i++ )
4300             for( int j = 0; j < 2; j++ )
4301             {
4302                 int l0 = x264_mb_type_list_table[i][0][j];
4303                 int l1 = x264_mb_type_list_table[i][1][j];
4304                 if( l0 || l1 )
4305                     list_count[l1+l0*l1] += h->stat.i_mb_count[SLICE_TYPE_B][i] * 2;
4306             }
4307         list_count[0] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L0_8x8];
4308         list_count[1] += h->stat.i_mb_partition[SLICE_TYPE_B][D_L1_8x8];
4309         list_count[2] += h->stat.i_mb_partition[SLICE_TYPE_B][D_BI_8x8];
4310         i_mb_count[B_DIRECT] += (h->stat.i_mb_partition[SLICE_TYPE_B][D_DIRECT_8x8]+2)/4;
4311         i_mb_list_count = (list_count[0] + list_count[1] + list_count[2]) / 100.0;
4312         sprintf( buf + strlen(buf), "  B16..8: %4.1f%% %4.1f%% %4.1f%%  direct:%4.1f%%  skip:%4.1f%%",
4313                  i_mb_size[PIXEL_16x16] / (i_count*4),
4314                  (i_mb_size[PIXEL_16x8] + i_mb_size[PIXEL_8x16]) / (i_count*4),
4315                  i_mb_size[PIXEL_8x8] / (i_count*4),
4316                  i_mb_count[B_DIRECT] / i_count,
4317                  i_mb_count[B_SKIP]   / i_count );
4318         if( i_mb_list_count != 0 )
4319             sprintf( buf + strlen(buf), "  L0:%4.1f%% L1:%4.1f%% BI:%4.1f%%",
4320                      list_count[0] / i_mb_list_count,
4321                      list_count[1] / i_mb_list_count,
4322                      list_count[2] / i_mb_list_count );
4323         x264_log( h, X264_LOG_INFO, "mb B  %s\n", buf );
4324     }
4325 
4326     x264_ratecontrol_summary( h );
4327 
4328     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 )
4329     {
4330 #define SUM3(p) (p[SLICE_TYPE_I] + p[SLICE_TYPE_P] + p[SLICE_TYPE_B])
4331 #define SUM3b(p,o) (p[SLICE_TYPE_I][o] + p[SLICE_TYPE_P][o] + p[SLICE_TYPE_B][o])
4332         int64_t i_i8x8 = SUM3b( h->stat.i_mb_count, I_8x8 );
4333         int64_t i_intra = i_i8x8 + SUM3b( h->stat.i_mb_count, I_4x4 )
4334                                  + SUM3b( h->stat.i_mb_count, I_16x16 );
4335         int64_t i_all_intra = i_intra + SUM3b( h->stat.i_mb_count, I_PCM );
4336         int64_t i_skip = SUM3b( h->stat.i_mb_count, P_SKIP )
4337                        + SUM3b( h->stat.i_mb_count, B_SKIP );
4338         const int i_count = h->stat.i_frame_count[SLICE_TYPE_I] +
4339                             h->stat.i_frame_count[SLICE_TYPE_P] +
4340                             h->stat.i_frame_count[SLICE_TYPE_B];
4341         int64_t i_mb_count = (int64_t)i_count * h->mb.i_mb_count;
4342         int64_t i_inter = i_mb_count - i_skip - i_all_intra;
4343         const double duration = h->stat.f_frame_duration[SLICE_TYPE_I] +
4344                                 h->stat.f_frame_duration[SLICE_TYPE_P] +
4345                                 h->stat.f_frame_duration[SLICE_TYPE_B];
4346         float f_bitrate = SUM3(h->stat.i_frame_size) / duration / 125;
4347 
4348         if( PARAM_INTERLACED )
4349         {
4350             char *fieldstats = buf;
4351             fieldstats[0] = 0;
4352             if( i_inter )
4353                 fieldstats += sprintf( fieldstats, " inter:%.1f%%", h->stat.i_mb_field[1] * 100.0 / i_inter );
4354             if( i_skip )
4355                 fieldstats += sprintf( fieldstats, " skip:%.1f%%", h->stat.i_mb_field[2] * 100.0 / i_skip );
4356             x264_log( h, X264_LOG_INFO, "field mbs: intra: %.1f%%%s\n",
4357                       h->stat.i_mb_field[0] * 100.0 / i_all_intra, buf );
4358         }
4359 
4360         if( h->pps->b_transform_8x8_mode )
4361         {
4362             buf[0] = 0;
4363             if( h->stat.i_mb_count_8x8dct[0] )
4364                 sprintf( buf, " inter:%.1f%%", 100. * h->stat.i_mb_count_8x8dct[1] / h->stat.i_mb_count_8x8dct[0] );
4365             x264_log( h, X264_LOG_INFO, "8x8 transform intra:%.1f%%%s\n", 100. * i_i8x8 / X264_MAX( i_intra, 1 ), buf );
4366         }
4367 
4368         if( (h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO ||
4369             (h->stat.i_direct_frames[0] && h->stat.i_direct_frames[1]))
4370             && h->stat.i_frame_count[SLICE_TYPE_B] )
4371         {
4372             x264_log( h, X264_LOG_INFO, "direct mvs  spatial:%.1f%% temporal:%.1f%%\n",
4373                       h->stat.i_direct_frames[1] * 100. / h->stat.i_frame_count[SLICE_TYPE_B],
4374                       h->stat.i_direct_frames[0] * 100. / h->stat.i_frame_count[SLICE_TYPE_B] );
4375         }
4376 
4377         buf[0] = 0;
4378         if( CHROMA_FORMAT )
4379         {
4380             int csize = CHROMA444 ? 4 : 1;
4381             if( i_mb_count != i_all_intra )
4382                 sprintf( buf, " inter: %.1f%% %.1f%% %.1f%%",
4383                          h->stat.i_mb_cbp[1] * 100.0 / ((i_mb_count - i_all_intra)*4),
4384                          h->stat.i_mb_cbp[3] * 100.0 / ((i_mb_count - i_all_intra)*csize),
4385                          h->stat.i_mb_cbp[5] * 100.0 / ((i_mb_count - i_all_intra)*csize) );
4386             x264_log( h, X264_LOG_INFO, "coded y,%s,%s intra: %.1f%% %.1f%% %.1f%%%s\n",
4387                       CHROMA444?"u":"uvDC", CHROMA444?"v":"uvAC",
4388                       h->stat.i_mb_cbp[0] * 100.0 / (i_all_intra*4),
4389                       h->stat.i_mb_cbp[2] * 100.0 / (i_all_intra*csize),
4390                       h->stat.i_mb_cbp[4] * 100.0 / (i_all_intra*csize), buf );
4391         }
4392         else
4393         {
4394             if( i_mb_count != i_all_intra )
4395                 sprintf( buf, " inter: %.1f%%", h->stat.i_mb_cbp[1] * 100.0 / ((i_mb_count - i_all_intra)*4) );
4396             x264_log( h, X264_LOG_INFO, "coded y intra: %.1f%%%s\n",
4397                       h->stat.i_mb_cbp[0] * 100.0 / (i_all_intra*4), buf );
4398         }
4399 
4400         int64_t fixed_pred_modes[4][9] = {{0}};
4401         int64_t sum_pred_modes[4] = {0};
4402         for( int i = 0; i <= I_PRED_16x16_DC_128; i++ )
4403         {
4404             fixed_pred_modes[0][x264_mb_pred_mode16x16_fix[i]] += h->stat.i_mb_pred_mode[0][i];
4405             sum_pred_modes[0] += h->stat.i_mb_pred_mode[0][i];
4406         }
4407         if( sum_pred_modes[0] )
4408             x264_log( h, X264_LOG_INFO, "i16 v,h,dc,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
4409                       fixed_pred_modes[0][0] * 100.0 / sum_pred_modes[0],
4410                       fixed_pred_modes[0][1] * 100.0 / sum_pred_modes[0],
4411                       fixed_pred_modes[0][2] * 100.0 / sum_pred_modes[0],
4412                       fixed_pred_modes[0][3] * 100.0 / sum_pred_modes[0] );
4413         for( int i = 1; i <= 2; i++ )
4414         {
4415             for( int j = 0; j <= I_PRED_8x8_DC_128; j++ )
4416             {
4417                 fixed_pred_modes[i][x264_mb_pred_mode4x4_fix(j)] += h->stat.i_mb_pred_mode[i][j];
4418                 sum_pred_modes[i] += h->stat.i_mb_pred_mode[i][j];
4419             }
4420             if( sum_pred_modes[i] )
4421                 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,
4422                           fixed_pred_modes[i][0] * 100.0 / sum_pred_modes[i],
4423                           fixed_pred_modes[i][1] * 100.0 / sum_pred_modes[i],
4424                           fixed_pred_modes[i][2] * 100.0 / sum_pred_modes[i],
4425                           fixed_pred_modes[i][3] * 100.0 / sum_pred_modes[i],
4426                           fixed_pred_modes[i][4] * 100.0 / sum_pred_modes[i],
4427                           fixed_pred_modes[i][5] * 100.0 / sum_pred_modes[i],
4428                           fixed_pred_modes[i][6] * 100.0 / sum_pred_modes[i],
4429                           fixed_pred_modes[i][7] * 100.0 / sum_pred_modes[i],
4430                           fixed_pred_modes[i][8] * 100.0 / sum_pred_modes[i] );
4431         }
4432         for( int i = 0; i <= I_PRED_CHROMA_DC_128; i++ )
4433         {
4434             fixed_pred_modes[3][x264_mb_chroma_pred_mode_fix[i]] += h->stat.i_mb_pred_mode[3][i];
4435             sum_pred_modes[3] += h->stat.i_mb_pred_mode[3][i];
4436         }
4437         if( sum_pred_modes[3] && !CHROMA444 )
4438             x264_log( h, X264_LOG_INFO, "i8c dc,h,v,p: %2.0f%% %2.0f%% %2.0f%% %2.0f%%\n",
4439                       fixed_pred_modes[3][0] * 100.0 / sum_pred_modes[3],
4440                       fixed_pred_modes[3][1] * 100.0 / sum_pred_modes[3],
4441                       fixed_pred_modes[3][2] * 100.0 / sum_pred_modes[3],
4442                       fixed_pred_modes[3][3] * 100.0 / sum_pred_modes[3] );
4443 
4444         if( h->param.analyse.i_weighted_pred >= X264_WEIGHTP_SIMPLE && h->stat.i_frame_count[SLICE_TYPE_P] > 0 )
4445         {
4446             buf[0] = 0;
4447             if( CHROMA_FORMAT )
4448                 sprintf( buf, " UV:%.1f%%", h->stat.i_wpred[1] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P] );
4449             x264_log( h, X264_LOG_INFO, "Weighted P-Frames: Y:%.1f%%%s\n",
4450                       h->stat.i_wpred[0] * 100.0 / h->stat.i_frame_count[SLICE_TYPE_P], buf );
4451         }
4452 
4453         for( int i_list = 0; i_list < 2; i_list++ )
4454             for( int i_slice = 0; i_slice < 2; i_slice++ )
4455             {
4456                 char *p = buf;
4457                 int64_t i_den = 0;
4458                 int i_max = 0;
4459                 for( int i = 0; i < X264_REF_MAX*2; i++ )
4460                     if( h->stat.i_mb_count_ref[i_slice][i_list][i] )
4461                     {
4462                         i_den += h->stat.i_mb_count_ref[i_slice][i_list][i];
4463                         i_max = i;
4464                     }
4465                 if( i_max == 0 )
4466                     continue;
4467                 for( int i = 0; i <= i_max; i++ )
4468                     p += sprintf( p, " %4.1f%%", 100. * h->stat.i_mb_count_ref[i_slice][i_list][i] / i_den );
4469                 x264_log( h, X264_LOG_INFO, "ref %c L%d:%s\n", "PB"[i_slice], i_list, buf );
4470             }
4471 
4472         if( h->param.analyse.b_ssim )
4473         {
4474             float ssim = SUM3( h->stat.f_ssim_mean_y ) / duration;
4475             x264_log( h, X264_LOG_INFO, "SSIM Mean Y:%.7f (%6.3fdb)\n", ssim, calc_ssim_db( ssim ) );
4476         }
4477         if( h->param.analyse.b_psnr )
4478         {
4479             x264_log( h, X264_LOG_INFO,
4480                       "PSNR Mean Y:%6.3f U:%6.3f V:%6.3f Avg:%6.3f Global:%6.3f kb/s:%.2f\n",
4481                       SUM3( h->stat.f_psnr_mean_y ) / duration,
4482                       SUM3( h->stat.f_psnr_mean_u ) / duration,
4483                       SUM3( h->stat.f_psnr_mean_v ) / duration,
4484                       SUM3( h->stat.f_psnr_average ) / duration,
4485                       calc_psnr( SUM3( h->stat.f_ssd_global ), duration * i_yuv_size ),
4486                       f_bitrate );
4487         }
4488         else
4489             x264_log( h, X264_LOG_INFO, "kb/s:%.2f\n", f_bitrate );
4490     }
4491 
4492     /* rc */
4493     x264_ratecontrol_delete( h );
4494 
4495     /* param */
4496     x264_param_cleanup( &h->param );
4497 
4498     x264_cqm_delete( h );
4499     x264_free( h->nal_buffer );
4500     x264_free( h->reconfig_h );
4501     x264_analyse_free_costs( h );
4502     x264_free( h->cost_table );
4503 
4504     if( h->i_thread_frames > 1 )
4505         h = h->thread[h->i_thread_phase];
4506 
4507     /* frames */
4508     x264_frame_delete_list( h->frames.unused[0] );
4509     x264_frame_delete_list( h->frames.unused[1] );
4510     x264_frame_delete_list( h->frames.current );
4511     x264_frame_delete_list( h->frames.blank_unused );
4512 
4513     h = h->thread[0];
4514 
4515     for( int i = 0; i < h->i_thread_frames; i++ )
4516         if( h->thread[i]->b_thread_active )
4517             for( int j = 0; j < h->thread[i]->i_ref[0]; j++ )
4518                 if( h->thread[i]->fref[0][j] && h->thread[i]->fref[0][j]->b_duplicate )
4519                     x264_frame_delete( h->thread[i]->fref[0][j] );
4520 
4521     if( h->param.i_lookahead_threads > 1 )
4522         for( int i = 0; i < h->param.i_lookahead_threads; i++ )
4523             x264_free( h->lookahead_thread[i] );
4524 
4525     for( int i = h->param.i_threads - 1; i >= 0; i-- )
4526     {
4527         x264_frame_t **frame;
4528 
4529         if( !h->param.b_sliced_threads || i == 0 )
4530         {
4531             for( frame = h->thread[i]->frames.reference; *frame; frame++ )
4532             {
4533                 assert( (*frame)->i_reference_count > 0 );
4534                 (*frame)->i_reference_count--;
4535                 if( (*frame)->i_reference_count == 0 )
4536                     x264_frame_delete( *frame );
4537             }
4538             frame = &h->thread[i]->fdec;
4539             if( *frame )
4540             {
4541                 assert( (*frame)->i_reference_count > 0 );
4542                 (*frame)->i_reference_count--;
4543                 if( (*frame)->i_reference_count == 0 )
4544                     x264_frame_delete( *frame );
4545             }
4546             x264_macroblock_cache_free( h->thread[i] );
4547         }
4548         x264_macroblock_thread_free( h->thread[i], 0 );
4549         x264_free( h->thread[i]->out.p_bitstream );
4550         x264_free( h->thread[i]->out.nal );
4551         x264_pthread_mutex_destroy( &h->thread[i]->mutex );
4552         x264_pthread_cond_destroy( &h->thread[i]->cv );
4553         x264_free( h->thread[i] );
4554     }
4555 #if HAVE_OPENCL
4556     x264_opencl_close_library( ocl );
4557 #endif
4558 }
4559 
x264_encoder_delayed_frames(x264_t * h)4560 int x264_encoder_delayed_frames( x264_t *h )
4561 {
4562     int delayed_frames = 0;
4563     if( h->i_thread_frames > 1 )
4564     {
4565         for( int i = 0; i < h->i_thread_frames; i++ )
4566             delayed_frames += h->thread[i]->b_thread_active;
4567         h = h->thread[h->i_thread_phase];
4568     }
4569     for( int i = 0; h->frames.current[i]; i++ )
4570         delayed_frames++;
4571     x264_pthread_mutex_lock( &h->lookahead->ofbuf.mutex );
4572     x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
4573     x264_pthread_mutex_lock( &h->lookahead->next.mutex );
4574     delayed_frames += h->lookahead->ifbuf.i_size + h->lookahead->next.i_size + h->lookahead->ofbuf.i_size;
4575     x264_pthread_mutex_unlock( &h->lookahead->next.mutex );
4576     x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
4577     x264_pthread_mutex_unlock( &h->lookahead->ofbuf.mutex );
4578     return delayed_frames;
4579 }
4580 
x264_encoder_maximum_delayed_frames(x264_t * h)4581 int x264_encoder_maximum_delayed_frames( x264_t *h )
4582 {
4583     return h->frames.i_delay;
4584 }
4585