1 /*****************************************************************************
2 * frame.c: frame handling
3 *****************************************************************************
4 * Copyright (C) 2003-2011 x264 project
6 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7 * Loren Merritt <lorenm@u.washington.edu>
8 * Fiona Glaser <fiona@x264.com>
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License as published by
12 * the Free Software Foundation; either version 2 of the License, or
13 * (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
24 * This program is also available under a commercial proprietary license.
25 * For more information, contact us at licensing@x264.com.
26 *****************************************************************************/
30 static int align_stride( int x, int align, int disalign )
32 x = ALIGN( x, align );
33 if( !(x&(disalign-1)) )
38 static int align_plane_size( int x, int disalign )
40 if( !(x&(disalign-1)) )
45 static int x264_frame_internal_csp( int external_csp )
47 switch( external_csp & X264_CSP_MASK )
61 static x264_frame_t *x264_frame_new( x264_t *h, int b_fdec )
64 int i_csp = x264_frame_internal_csp( h->param.i_csp );
65 int i_mb_count = h->mb.i_mb_count;
66 int i_stride, i_width, i_lines;
67 int i_padv = PADV << PARAM_INTERLACED;
68 int align = h->param.cpu&X264_CPU_CACHELINE_64 ? 64 : h->param.cpu&X264_CPU_CACHELINE_32 ? 32 : 16;
69 int disalign = h->param.cpu&X264_CPU_ALTIVEC ? 1<<9 : 1<<10;
70 int luma_plane_count = i_csp == X264_CSP_NV12 ? 1 : 3;
72 CHECKED_MALLOCZERO( frame, sizeof(x264_frame_t) );
74 /* allocate frame data (+64 for extra data for me) */
75 i_width = h->mb.i_mb_width*16;
76 i_lines = h->mb.i_mb_height*16;
77 i_stride = align_stride( i_width + 2*PADH, align, disalign );
79 if( i_csp == X264_CSP_NV12 )
82 for( int i = 0; i < 2; i++ )
84 frame->i_width[i] = i_width >> i;
85 frame->i_lines[i] = i_lines >> i;
86 frame->i_stride[i] = i_stride;
89 else if( i_csp == X264_CSP_I444 )
92 for( int i = 0; i < 3; i++ )
94 frame->i_width[i] = i_width;
95 frame->i_lines[i] = i_lines;
96 frame->i_stride[i] = i_stride;
102 frame->i_width_lowres = frame->i_width[0]/2;
103 frame->i_lines_lowres = frame->i_lines[0]/2;
104 frame->i_stride_lowres = align_stride( frame->i_width_lowres + 2*PADH, align, disalign<<1 );
106 for( int i = 0; i < h->param.i_bframe + 2; i++ )
107 for( int j = 0; j < h->param.i_bframe + 2; j++ )
108 CHECKED_MALLOC( frame->i_row_satds[i][j], i_lines/16 * sizeof(int) );
111 frame->i_type = X264_TYPE_AUTO;
112 frame->i_qpplus1 = X264_QP_AUTO;
115 frame->i_frame_num = -1;
116 frame->i_lines_completed = -1;
117 frame->b_fdec = b_fdec;
118 frame->i_pic_struct = PIC_STRUCT_AUTO;
119 frame->i_field_cnt = -1;
121 frame->i_cpb_duration =
122 frame->i_dpb_output_delay =
123 frame->i_cpb_delay = 0;
124 frame->i_coded_fields_lookahead =
125 frame->i_cpb_delay_lookahead = -1;
129 if( i_csp == X264_CSP_NV12 )
131 int chroma_plane_size = (frame->i_stride[1] * (frame->i_lines[1] + i_padv));
132 CHECKED_MALLOC( frame->buffer[1], chroma_plane_size * sizeof(pixel) );
133 frame->plane[1] = frame->buffer[1] + frame->i_stride[1] * i_padv/2 + PADH;
134 if( PARAM_INTERLACED )
136 CHECKED_MALLOC( frame->buffer_fld[1], chroma_plane_size * sizeof(pixel) );
137 frame->plane_fld[1] = frame->buffer_fld[1] + frame->i_stride[1] * i_padv/2 + PADH;
141 /* all 4 luma planes allocated together, since the cacheline split code
142 * requires them to be in-phase wrt cacheline alignment. */
144 for( int p = 0; p < luma_plane_count; p++ )
146 int luma_plane_size = align_plane_size( frame->i_stride[p] * (frame->i_lines[p] + 2*i_padv), disalign );
147 if( h->param.analyse.i_subpel_refine && b_fdec )
149 /* FIXME: Don't allocate both buffers in non-adaptive MBAFF. */
150 CHECKED_MALLOC( frame->buffer[p], 4*luma_plane_size * sizeof(pixel) );
151 if( PARAM_INTERLACED )
152 CHECKED_MALLOC( frame->buffer_fld[p], 4*luma_plane_size * sizeof(pixel) );
153 for( int i = 0; i < 4; i++ )
155 frame->filtered[p][i] = frame->buffer[p] + i*luma_plane_size + frame->i_stride[p] * i_padv + PADH;
156 frame->filtered_fld[p][i] = frame->buffer_fld[p] + i*luma_plane_size + frame->i_stride[p] * i_padv + PADH;
158 frame->plane[p] = frame->filtered[p][0];
159 frame->plane_fld[p] = frame->filtered_fld[p][0];
163 CHECKED_MALLOC( frame->buffer[p], luma_plane_size * sizeof(pixel) );
164 if( PARAM_INTERLACED )
165 CHECKED_MALLOC( frame->buffer_fld[p], luma_plane_size * sizeof(pixel) );
166 frame->filtered[p][0] = frame->plane[p] = frame->buffer[p] + frame->i_stride[p] * i_padv + PADH;
167 frame->filtered_fld[p][0] = frame->plane_fld[p] = frame->buffer_fld[p] + frame->i_stride[p] * i_padv + PADH;
171 frame->b_duplicate = 0;
173 if( b_fdec ) /* fdec frame */
175 CHECKED_MALLOC( frame->mb_type, i_mb_count * sizeof(int8_t));
176 CHECKED_MALLOC( frame->mb_partition, i_mb_count * sizeof(uint8_t));
177 CHECKED_MALLOC( frame->mv[0], 2*16 * i_mb_count * sizeof(int16_t) );
178 CHECKED_MALLOC( frame->mv16x16, 2*(i_mb_count+1) * sizeof(int16_t) );
179 M32( frame->mv16x16[0] ) = 0;
181 CHECKED_MALLOC( frame->ref[0], 4 * i_mb_count * sizeof(int8_t) );
182 if( h->param.i_bframe )
184 CHECKED_MALLOC( frame->mv[1], 2*16 * i_mb_count * sizeof(int16_t) );
185 CHECKED_MALLOC( frame->ref[1], 4 * i_mb_count * sizeof(int8_t) );
190 frame->ref[1] = NULL;
192 CHECKED_MALLOC( frame->i_row_bits, i_lines/16 * sizeof(int) );
193 CHECKED_MALLOC( frame->f_row_qp, i_lines/16 * sizeof(float) );
194 CHECKED_MALLOC( frame->f_row_qscale, i_lines/16 * sizeof(float) );
195 if( h->param.analyse.i_me_method >= X264_ME_ESA )
197 CHECKED_MALLOC( frame->buffer[3],
198 frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) << h->frames.b_have_sub8x8_esa );
199 frame->integral = (uint16_t*)frame->buffer[3] + frame->i_stride[0] * i_padv + PADH;
201 if( PARAM_INTERLACED )
202 CHECKED_MALLOC( frame->field, i_mb_count * sizeof(uint8_t) );
204 else /* fenc frame */
206 if( h->frames.b_have_lowres )
208 int luma_plane_size = align_plane_size( frame->i_stride_lowres * (frame->i_lines[0]/2 + 2*PADV), disalign );
210 CHECKED_MALLOC( frame->buffer_lowres[0], 4 * luma_plane_size * sizeof(pixel) );
211 for( int i = 0; i < 4; i++ )
212 frame->lowres[i] = frame->buffer_lowres[0] + (frame->i_stride_lowres * PADV + PADH) + i * luma_plane_size;
214 for( int j = 0; j <= !!h->param.i_bframe; j++ )
215 for( int i = 0; i <= h->param.i_bframe; i++ )
217 CHECKED_MALLOCZERO( frame->lowres_mvs[j][i], 2*h->mb.i_mb_count*sizeof(int16_t) );
218 CHECKED_MALLOC( frame->lowres_mv_costs[j][i], h->mb.i_mb_count*sizeof(int) );
220 CHECKED_MALLOC( frame->i_propagate_cost, (i_mb_count+7) * sizeof(uint16_t) );
221 for( int j = 0; j <= h->param.i_bframe+1; j++ )
222 for( int i = 0; i <= h->param.i_bframe+1; i++ )
223 CHECKED_MALLOC( frame->lowres_costs[j][i], (i_mb_count+3) * sizeof(uint16_t) );
224 frame->i_intra_cost = frame->lowres_costs[0][0];
225 memset( frame->i_intra_cost, -1, (i_mb_count+3) * sizeof(uint16_t) );
227 if( h->param.rc.i_aq_mode )
229 CHECKED_MALLOC( frame->f_qp_offset, h->mb.i_mb_count * sizeof(float) );
230 CHECKED_MALLOC( frame->f_qp_offset_aq, h->mb.i_mb_count * sizeof(float) );
231 if( h->frames.b_have_lowres )
232 /* shouldn't really be initialized, just silences a valgrind false-positive in x264_mbtree_propagate_cost_sse2 */
233 CHECKED_MALLOCZERO( frame->i_inv_qscale_factor, (h->mb.i_mb_count+3) * sizeof(uint16_t) );
237 if( x264_pthread_mutex_init( &frame->mutex, NULL ) )
239 if( x264_pthread_cond_init( &frame->cv, NULL ) )
249 void x264_frame_delete( x264_frame_t *frame )
251 /* Duplicate frames are blank copies of real frames (including pointers),
252 * so freeing those pointers would cause a double free later. */
253 if( !frame->b_duplicate )
255 for( int i = 0; i < 4; i++ )
257 x264_free( frame->buffer[i] );
258 x264_free( frame->buffer_fld[i] );
260 for( int i = 0; i < 4; i++ )
261 x264_free( frame->buffer_lowres[i] );
262 for( int i = 0; i < X264_BFRAME_MAX+2; i++ )
263 for( int j = 0; j < X264_BFRAME_MAX+2; j++ )
264 x264_free( frame->i_row_satds[i][j] );
265 for( int j = 0; j < 2; j++ )
266 for( int i = 0; i <= X264_BFRAME_MAX; i++ )
268 x264_free( frame->lowres_mvs[j][i] );
269 x264_free( frame->lowres_mv_costs[j][i] );
271 x264_free( frame->i_propagate_cost );
272 for( int j = 0; j <= X264_BFRAME_MAX+1; j++ )
273 for( int i = 0; i <= X264_BFRAME_MAX+1; i++ )
274 x264_free( frame->lowres_costs[j][i] );
275 x264_free( frame->f_qp_offset );
276 x264_free( frame->f_qp_offset_aq );
277 x264_free( frame->i_inv_qscale_factor );
278 x264_free( frame->i_row_bits );
279 x264_free( frame->f_row_qp );
280 x264_free( frame->f_row_qscale );
281 x264_free( frame->field );
282 x264_free( frame->mb_type );
283 x264_free( frame->mb_partition );
284 x264_free( frame->mv[0] );
285 x264_free( frame->mv[1] );
287 x264_free( frame->mv16x16-1 );
288 x264_free( frame->ref[0] );
289 x264_free( frame->ref[1] );
290 x264_pthread_mutex_destroy( &frame->mutex );
291 x264_pthread_cond_destroy( &frame->cv );
296 static int get_plane_ptr( x264_t *h, x264_picture_t *src, uint8_t **pix, int *stride, int plane, int xshift, int yshift )
298 int width = h->param.i_width >> xshift;
299 int height = h->param.i_height >> yshift;
300 *pix = src->img.plane[plane];
301 *stride = src->img.i_stride[plane];
302 if( src->img.i_csp & X264_CSP_VFLIP )
304 *pix += (height-1) * *stride;
307 if( width > abs(*stride) )
309 x264_log( h, X264_LOG_ERROR, "Input picture width (%d) is greater than stride (%d)\n", width, *stride );
315 #define get_plane_ptr(...) do{ if( get_plane_ptr(__VA_ARGS__) < 0 ) return -1; }while(0)
317 int x264_frame_copy_picture( x264_t *h, x264_frame_t *dst, x264_picture_t *src )
319 int i_csp = src->img.i_csp & X264_CSP_MASK;
320 if( i_csp <= X264_CSP_NONE || i_csp >= X264_CSP_MAX )
322 x264_log( h, X264_LOG_ERROR, "Invalid input colorspace\n" );
327 if( !(src->img.i_csp & X264_CSP_HIGH_DEPTH) )
329 x264_log( h, X264_LOG_ERROR, "This build of x264 requires high depth input. Rebuild to support 8-bit input.\n" );
333 if( src->img.i_csp & X264_CSP_HIGH_DEPTH )
335 x264_log( h, X264_LOG_ERROR, "This build of x264 requires 8-bit input. Rebuild to support high depth input.\n" );
340 dst->i_type = src->i_type;
341 dst->i_qpplus1 = src->i_qpplus1;
342 dst->i_pts = dst->i_reordered_pts = src->i_pts;
343 dst->param = src->param;
344 dst->i_pic_struct = src->i_pic_struct;
345 dst->extra_sei = src->extra_sei;
349 get_plane_ptr( h, src, &pix[0], &stride[0], 0, 0, 0 );
350 h->mc.plane_copy( dst->plane[0], dst->i_stride[0], (pixel*)pix[0],
351 stride[0]/sizeof(pixel), h->param.i_width, h->param.i_height );
352 if( i_csp == X264_CSP_NV12 )
354 get_plane_ptr( h, src, &pix[1], &stride[1], 1, 0, 1 );
355 h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
356 stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height>>1 );
358 else if( i_csp == X264_CSP_I420 || i_csp == X264_CSP_YV12 )
360 get_plane_ptr( h, src, &pix[1], &stride[1], i_csp==X264_CSP_I420 ? 1 : 2, 1, 1 );
361 get_plane_ptr( h, src, &pix[2], &stride[2], i_csp==X264_CSP_I420 ? 2 : 1, 1, 1 );
362 h->mc.plane_copy_interleave( dst->plane[1], dst->i_stride[1],
363 (pixel*)pix[1], stride[1]/sizeof(pixel),
364 (pixel*)pix[2], stride[2]/sizeof(pixel),
365 h->param.i_width>>1, h->param.i_height>>1 );
367 else //if( i_csp == X264_CSP_I444 || i_csp == X264_CSP_YV24 )
369 get_plane_ptr( h, src, &pix[1], &stride[1], i_csp==X264_CSP_I444 ? 1 : 2, 0, 0 );
370 get_plane_ptr( h, src, &pix[2], &stride[2], i_csp==X264_CSP_I444 ? 2 : 1, 0, 0 );
371 h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
372 stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height );
373 h->mc.plane_copy( dst->plane[2], dst->i_stride[2], (pixel*)pix[2],
374 stride[2]/sizeof(pixel), h->param.i_width, h->param.i_height );
379 static void ALWAYS_INLINE pixel_memset( pixel *dst, pixel *src, int len, int size )
381 uint8_t *dstp = (uint8_t*)dst;
383 uint16_t v2 = size == 1 ? v1 + (v1 << 8) : M16( src );
384 uint32_t v4 = size <= 2 ? v2 + (v2 << 16) : M32( src );
388 /* Align the input pointer if it isn't already */
389 if( (intptr_t)dstp & (WORD_SIZE - 1) )
391 if( size <= 2 && ((intptr_t)dstp & 3) )
393 if( size == 1 && ((intptr_t)dstp & 1) )
395 if( (intptr_t)dstp & 2 )
401 if( WORD_SIZE == 8 && (intptr_t)dstp & 4 )
411 uint64_t v8 = v4 + ((uint64_t)v4<<32);
412 for( ; i < len - 7; i+=8 )
415 for( ; i < len - 3; i+=4 )
418 /* Finish up the last few bytes */
426 if( size == 1 && i != len )
431 static void ALWAYS_INLINE plane_expand_border( pixel *pix, int i_stride, int i_width, int i_height, int i_padh, int i_padv, int b_pad_top, int b_pad_bottom, int b_chroma )
433 #define PPIXEL(x, y) ( pix + (x) + (y)*i_stride )
434 for( int y = 0; y < i_height; y++ )
437 pixel_memset( PPIXEL(-i_padh, y), PPIXEL(0, y), i_padh>>b_chroma, sizeof(pixel)<<b_chroma );
439 pixel_memset( PPIXEL(i_width, y), PPIXEL(i_width-1-b_chroma, y), i_padh>>b_chroma, sizeof(pixel)<<b_chroma );
443 for( int y = 0; y < i_padv; y++ )
444 memcpy( PPIXEL(-i_padh, -y-1), PPIXEL(-i_padh, 0), (i_width+2*i_padh) * sizeof(pixel) );
447 for( int y = 0; y < i_padv; y++ )
448 memcpy( PPIXEL(-i_padh, i_height+y), PPIXEL(-i_padh, i_height-1), (i_width+2*i_padh) * sizeof(pixel) );
452 void x264_frame_expand_border( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
455 if( mb_y & SLICE_MBAFF )
457 for( int i = 0; i < frame->i_plane; i++ )
459 int shift = i && !CHROMA444;
460 int stride = frame->i_stride[i];
461 int width = 16*h->mb.i_mb_width;
462 int height = (b_end ? 16*(h->mb.i_mb_height - mb_y) >> SLICE_MBAFF : 16) >> shift;
464 int padv = PADV >> shift;
465 // buffer: 2 chroma, 3 luma (rounded to 4) because deblocking goes beyond the top of the mb
466 if( b_end && !b_start )
467 height += 4 >> (shift + SLICE_MBAFF);
471 // border samples for each field are extended separately
472 pix = frame->plane_fld[i] + X264_MAX(0, (16*mb_y-4)*stride >> shift);
473 plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end, shift );
474 plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end, shift );
476 height = (b_end ? 16*(h->mb.i_mb_height - mb_y) : 32) >> shift;
477 if( b_end && !b_start )
478 height += 4 >> shift;
479 pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> shift);
480 plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end, shift );
484 pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> shift);
485 plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end, shift );
490 void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
492 /* during filtering, 8 extra pixels were filtered on each edge,
493 * but up to 3 of the horizontal ones may be wrong.
494 we want to expand border from the last filtered pixel */
496 int width = 16*h->mb.i_mb_width + 8;
497 int height = b_end ? (16*(h->mb.i_mb_height - mb_y) >> SLICE_MBAFF) + 16 : 16;
500 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
501 for( int i = 1; i < 4; i++ )
503 int stride = frame->i_stride[p];
504 // buffer: 8 luma, to match the hpel filter
508 pix = frame->filtered_fld[p][i] + (16*mb_y - 16) * stride - 4;
509 plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end, 0 );
510 plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end, 0 );
513 pix = frame->filtered[p][i] + (16*mb_y - 8) * stride - 4;
514 plane_expand_border( pix, stride, width, height << SLICE_MBAFF, padh, padv, b_start, b_end, 0 );
518 void x264_frame_expand_border_lowres( x264_frame_t *frame )
520 for( int i = 0; i < 4; i++ )
521 plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_width_lowres, frame->i_lines_lowres, PADH, PADV, 1, 1, 0 );
524 void x264_frame_expand_border_mod16( x264_t *h, x264_frame_t *frame )
526 for( int i = 0; i < frame->i_plane; i++ )
528 int i_width = h->param.i_width;
529 int shift = i && !CHROMA444;
530 int i_height = h->param.i_height >> shift;
531 int i_padx = (h->mb.i_mb_width * 16 - h->param.i_width);
532 int i_pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> shift;
536 for( int y = 0; y < i_height; y++ )
537 pixel_memset( &frame->plane[i][y*frame->i_stride[i] + i_width],
538 &frame->plane[i][y*frame->i_stride[i] + i_width - 1-shift],
539 i_padx>>shift, sizeof(pixel)<<shift );
543 for( int y = i_height; y < i_height + i_pady; y++ )
544 memcpy( &frame->plane[i][y*frame->i_stride[i]],
545 &frame->plane[i][(i_height-(~y&PARAM_INTERLACED)-1)*frame->i_stride[i]],
546 (i_width + i_padx) * sizeof(pixel) );
551 void x264_expand_border_mbpair( x264_t *h, int mb_x, int mb_y )
553 for( int i = 0; i < h->fenc->i_plane; i++ )
555 int shift = i && !CHROMA444;
556 int stride = h->fenc->i_stride[i];
557 int height = h->param.i_height >> shift;
558 int pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> shift;
559 int mbsize = 16>>shift;
560 pixel *fenc = h->fenc->plane[i] + mbsize * mb_x;
561 for( int y = height; y < height + pady; y++ )
562 memcpy( fenc + y*stride,
563 fenc + (height-1)*stride,
564 mbsize * sizeof(pixel) );
569 void x264_frame_cond_broadcast( x264_frame_t *frame, int i_lines_completed )
571 x264_pthread_mutex_lock( &frame->mutex );
572 frame->i_lines_completed = i_lines_completed;
573 x264_pthread_cond_broadcast( &frame->cv );
574 x264_pthread_mutex_unlock( &frame->mutex );
577 void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
579 x264_pthread_mutex_lock( &frame->mutex );
580 while( frame->i_lines_completed < i_lines_completed )
581 x264_pthread_cond_wait( &frame->cv, &frame->mutex );
582 x264_pthread_mutex_unlock( &frame->mutex );
587 void x264_frame_push( x264_frame_t **list, x264_frame_t *frame )
590 while( list[i] ) i++;
594 x264_frame_t *x264_frame_pop( x264_frame_t **list )
599 while( list[i+1] ) i++;
605 void x264_frame_unshift( x264_frame_t **list, x264_frame_t *frame )
608 while( list[i] ) i++;
614 x264_frame_t *x264_frame_shift( x264_frame_t **list )
616 x264_frame_t *frame = list[0];
618 for( i = 0; list[i]; i++ )
624 void x264_frame_push_unused( x264_t *h, x264_frame_t *frame )
626 assert( frame->i_reference_count > 0 );
627 frame->i_reference_count--;
628 if( frame->i_reference_count == 0 )
629 x264_frame_push( h->frames.unused[frame->b_fdec], frame );
632 x264_frame_t *x264_frame_pop_unused( x264_t *h, int b_fdec )
635 if( h->frames.unused[b_fdec][0] )
636 frame = x264_frame_pop( h->frames.unused[b_fdec] );
638 frame = x264_frame_new( h, b_fdec );
641 frame->b_last_minigop_bframe = 0;
642 frame->i_reference_count = 1;
643 frame->b_intra_calculated = 0;
644 frame->b_scenecut = 1;
645 frame->b_keyframe = 0;
646 frame->b_corrupt = 0;
648 memset( frame->weight, 0, sizeof(frame->weight) );
649 memset( frame->f_weighted_cost_delta, 0, sizeof(frame->f_weighted_cost_delta) );
654 void x264_frame_push_blank_unused( x264_t *h, x264_frame_t *frame )
656 assert( frame->i_reference_count > 0 );
657 frame->i_reference_count--;
658 if( frame->i_reference_count == 0 )
659 x264_frame_push( h->frames.blank_unused, frame );
662 x264_frame_t *x264_frame_pop_blank_unused( x264_t *h )
665 if( h->frames.blank_unused[0] )
666 frame = x264_frame_pop( h->frames.blank_unused );
668 frame = x264_malloc( sizeof(x264_frame_t) );
671 frame->b_duplicate = 1;
672 frame->i_reference_count = 1;
676 void x264_frame_sort( x264_frame_t **list, int b_dts )
681 for( int i = 0; list[i+1]; i++ )
683 int dtype = list[i]->i_type - list[i+1]->i_type;
684 int dtime = list[i]->i_frame - list[i+1]->i_frame;
685 int swap = b_dts ? dtype > 0 || ( dtype == 0 && dtime > 0 )
689 XCHG( x264_frame_t*, list[i], list[i+1] );
696 void x264_weight_scale_plane( x264_t *h, pixel *dst, int i_dst_stride, pixel *src, int i_src_stride,
697 int i_width, int i_height, x264_weight_t *w )
699 /* Weight horizontal strips of height 16. This was found to be the optimal height
700 * in terms of the cache loads. */
701 while( i_height > 0 )
703 for( int x = 0; x < i_width; x += 16 )
704 w->weightfn[16>>2]( dst+x, i_dst_stride, src+x, i_src_stride, w, X264_MIN( i_height, 16 ) );
706 dst += 16 * i_dst_stride;
707 src += 16 * i_src_stride;
711 void x264_frame_delete_list( x264_frame_t **list )
717 x264_frame_delete( list[i++] );
721 int x264_sync_frame_list_init( x264_sync_frame_list_t *slist, int max_size )
725 slist->i_max_size = max_size;
727 CHECKED_MALLOCZERO( slist->list, (max_size+1) * sizeof(x264_frame_t*) );
728 if( x264_pthread_mutex_init( &slist->mutex, NULL ) ||
729 x264_pthread_cond_init( &slist->cv_fill, NULL ) ||
730 x264_pthread_cond_init( &slist->cv_empty, NULL ) )
737 void x264_sync_frame_list_delete( x264_sync_frame_list_t *slist )
739 x264_pthread_mutex_destroy( &slist->mutex );
740 x264_pthread_cond_destroy( &slist->cv_fill );
741 x264_pthread_cond_destroy( &slist->cv_empty );
742 x264_frame_delete_list( slist->list );
745 void x264_sync_frame_list_push( x264_sync_frame_list_t *slist, x264_frame_t *frame )
747 x264_pthread_mutex_lock( &slist->mutex );
748 while( slist->i_size == slist->i_max_size )
749 x264_pthread_cond_wait( &slist->cv_empty, &slist->mutex );
750 slist->list[ slist->i_size++ ] = frame;
751 x264_pthread_mutex_unlock( &slist->mutex );
752 x264_pthread_cond_broadcast( &slist->cv_fill );
755 x264_frame_t *x264_sync_frame_list_pop( x264_sync_frame_list_t *slist )
758 x264_pthread_mutex_lock( &slist->mutex );
759 while( !slist->i_size )
760 x264_pthread_cond_wait( &slist->cv_fill, &slist->mutex );
761 frame = slist->list[ --slist->i_size ];
762 slist->list[ slist->i_size ] = NULL;
763 x264_pthread_cond_broadcast( &slist->cv_empty );
764 x264_pthread_mutex_unlock( &slist->mutex );