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 )
64 static x264_frame_t *x264_frame_new( x264_t *h, int b_fdec )
67 int i_csp = x264_frame_internal_csp( h->param.i_csp );
68 int i_mb_count = h->mb.i_mb_count;
69 int i_stride, i_width, i_lines;
70 int i_padv = PADV << PARAM_INTERLACED;
71 int align = h->param.cpu&X264_CPU_CACHELINE_64 ? 64 : h->param.cpu&X264_CPU_CACHELINE_32 ? 32 : 16;
72 int disalign = h->param.cpu&X264_CPU_ALTIVEC ? 1<<9 : 1<<10;
73 int luma_plane_count = i_csp == X264_CSP_NV12 ? 1 : 3;
75 CHECKED_MALLOCZERO( frame, sizeof(x264_frame_t) );
77 /* allocate frame data (+64 for extra data for me) */
78 i_width = h->mb.i_mb_width*16;
79 i_lines = h->mb.i_mb_height*16;
80 i_stride = align_stride( i_width + 2*PADH, align, disalign );
82 if( i_csp == X264_CSP_NV12 )
85 for( int i = 0; i < 2; i++ )
87 frame->i_width[i] = i_width >> i;
88 frame->i_lines[i] = i_lines >> i;
89 frame->i_stride[i] = i_stride;
92 else if( i_csp == X264_CSP_I444 )
95 for( int i = 0; i < 3; i++ )
97 frame->i_width[i] = i_width;
98 frame->i_lines[i] = i_lines;
99 frame->i_stride[i] = i_stride;
105 frame->i_csp = i_csp;
106 frame->i_width_lowres = frame->i_width[0]/2;
107 frame->i_lines_lowres = frame->i_lines[0]/2;
108 frame->i_stride_lowres = align_stride( frame->i_width_lowres + 2*PADH, align, disalign<<1 );
110 for( int i = 0; i < h->param.i_bframe + 2; i++ )
111 for( int j = 0; j < h->param.i_bframe + 2; j++ )
112 CHECKED_MALLOC( frame->i_row_satds[i][j], i_lines/16 * sizeof(int) );
115 frame->i_type = X264_TYPE_AUTO;
116 frame->i_qpplus1 = X264_QP_AUTO;
119 frame->i_frame_num = -1;
120 frame->i_lines_completed = -1;
121 frame->b_fdec = b_fdec;
122 frame->i_pic_struct = PIC_STRUCT_AUTO;
123 frame->i_field_cnt = -1;
125 frame->i_cpb_duration =
126 frame->i_dpb_output_delay =
127 frame->i_cpb_delay = 0;
128 frame->i_coded_fields_lookahead =
129 frame->i_cpb_delay_lookahead = -1;
133 if( i_csp == X264_CSP_NV12 )
135 int chroma_plane_size = (frame->i_stride[1] * (frame->i_lines[1] + i_padv));
136 CHECKED_MALLOC( frame->buffer[1], chroma_plane_size * sizeof(pixel) );
137 frame->plane[1] = frame->buffer[1] + frame->i_stride[1] * i_padv/2 + PADH;
138 if( PARAM_INTERLACED )
140 CHECKED_MALLOC( frame->buffer_fld[1], chroma_plane_size * sizeof(pixel) );
141 frame->plane_fld[1] = frame->buffer_fld[1] + frame->i_stride[1] * i_padv/2 + PADH;
145 /* all 4 luma planes allocated together, since the cacheline split code
146 * requires them to be in-phase wrt cacheline alignment. */
148 for( int p = 0; p < luma_plane_count; p++ )
150 int luma_plane_size = align_plane_size( frame->i_stride[p] * (frame->i_lines[p] + 2*i_padv), disalign );
151 if( h->param.analyse.i_subpel_refine && b_fdec )
153 /* FIXME: Don't allocate both buffers in non-adaptive MBAFF. */
154 CHECKED_MALLOC( frame->buffer[p], 4*luma_plane_size * sizeof(pixel) );
155 if( PARAM_INTERLACED )
156 CHECKED_MALLOC( frame->buffer_fld[p], 4*luma_plane_size * sizeof(pixel) );
157 for( int i = 0; i < 4; i++ )
159 frame->filtered[p][i] = frame->buffer[p] + i*luma_plane_size + frame->i_stride[p] * i_padv + PADH;
160 frame->filtered_fld[p][i] = frame->buffer_fld[p] + i*luma_plane_size + frame->i_stride[p] * i_padv + PADH;
162 frame->plane[p] = frame->filtered[p][0];
163 frame->plane_fld[p] = frame->filtered_fld[p][0];
167 CHECKED_MALLOC( frame->buffer[p], luma_plane_size * sizeof(pixel) );
168 if( PARAM_INTERLACED )
169 CHECKED_MALLOC( frame->buffer_fld[p], luma_plane_size * sizeof(pixel) );
170 frame->filtered[p][0] = frame->plane[p] = frame->buffer[p] + frame->i_stride[p] * i_padv + PADH;
171 frame->filtered_fld[p][0] = frame->plane_fld[p] = frame->buffer_fld[p] + frame->i_stride[p] * i_padv + PADH;
175 frame->b_duplicate = 0;
177 if( b_fdec ) /* fdec frame */
179 CHECKED_MALLOC( frame->mb_type, i_mb_count * sizeof(int8_t));
180 CHECKED_MALLOC( frame->mb_partition, i_mb_count * sizeof(uint8_t));
181 CHECKED_MALLOC( frame->mv[0], 2*16 * i_mb_count * sizeof(int16_t) );
182 CHECKED_MALLOC( frame->mv16x16, 2*(i_mb_count+1) * sizeof(int16_t) );
183 M32( frame->mv16x16[0] ) = 0;
185 CHECKED_MALLOC( frame->ref[0], 4 * i_mb_count * sizeof(int8_t) );
186 if( h->param.i_bframe )
188 CHECKED_MALLOC( frame->mv[1], 2*16 * i_mb_count * sizeof(int16_t) );
189 CHECKED_MALLOC( frame->ref[1], 4 * i_mb_count * sizeof(int8_t) );
194 frame->ref[1] = NULL;
196 CHECKED_MALLOC( frame->i_row_bits, i_lines/16 * sizeof(int) );
197 CHECKED_MALLOC( frame->f_row_qp, i_lines/16 * sizeof(float) );
198 CHECKED_MALLOC( frame->f_row_qscale, i_lines/16 * sizeof(float) );
199 if( h->param.analyse.i_me_method >= X264_ME_ESA )
201 CHECKED_MALLOC( frame->buffer[3],
202 frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) << h->frames.b_have_sub8x8_esa );
203 frame->integral = (uint16_t*)frame->buffer[3] + frame->i_stride[0] * i_padv + PADH;
205 if( PARAM_INTERLACED )
206 CHECKED_MALLOC( frame->field, i_mb_count * sizeof(uint8_t) );
208 else /* fenc frame */
210 if( h->frames.b_have_lowres )
212 int luma_plane_size = align_plane_size( frame->i_stride_lowres * (frame->i_lines[0]/2 + 2*PADV), disalign );
214 CHECKED_MALLOC( frame->buffer_lowres[0], 4 * luma_plane_size * sizeof(pixel) );
215 for( int i = 0; i < 4; i++ )
216 frame->lowres[i] = frame->buffer_lowres[0] + (frame->i_stride_lowres * PADV + PADH) + i * luma_plane_size;
218 for( int j = 0; j <= !!h->param.i_bframe; j++ )
219 for( int i = 0; i <= h->param.i_bframe; i++ )
221 CHECKED_MALLOCZERO( frame->lowres_mvs[j][i], 2*h->mb.i_mb_count*sizeof(int16_t) );
222 CHECKED_MALLOC( frame->lowres_mv_costs[j][i], h->mb.i_mb_count*sizeof(int) );
224 CHECKED_MALLOC( frame->i_propagate_cost, (i_mb_count+7) * sizeof(uint16_t) );
225 for( int j = 0; j <= h->param.i_bframe+1; j++ )
226 for( int i = 0; i <= h->param.i_bframe+1; i++ )
227 CHECKED_MALLOC( frame->lowres_costs[j][i], (i_mb_count+3) * sizeof(uint16_t) );
228 frame->i_intra_cost = frame->lowres_costs[0][0];
229 memset( frame->i_intra_cost, -1, (i_mb_count+3) * sizeof(uint16_t) );
231 if( h->param.rc.i_aq_mode )
233 CHECKED_MALLOC( frame->f_qp_offset, h->mb.i_mb_count * sizeof(float) );
234 CHECKED_MALLOC( frame->f_qp_offset_aq, h->mb.i_mb_count * sizeof(float) );
235 if( h->frames.b_have_lowres )
236 /* shouldn't really be initialized, just silences a valgrind false-positive in x264_mbtree_propagate_cost_sse2 */
237 CHECKED_MALLOCZERO( frame->i_inv_qscale_factor, (h->mb.i_mb_count+3) * sizeof(uint16_t) );
241 if( x264_pthread_mutex_init( &frame->mutex, NULL ) )
243 if( x264_pthread_cond_init( &frame->cv, NULL ) )
253 void x264_frame_delete( x264_frame_t *frame )
255 /* Duplicate frames are blank copies of real frames (including pointers),
256 * so freeing those pointers would cause a double free later. */
257 if( !frame->b_duplicate )
259 for( int i = 0; i < 4; i++ )
261 x264_free( frame->buffer[i] );
262 x264_free( frame->buffer_fld[i] );
264 for( int i = 0; i < 4; i++ )
265 x264_free( frame->buffer_lowres[i] );
266 for( int i = 0; i < X264_BFRAME_MAX+2; i++ )
267 for( int j = 0; j < X264_BFRAME_MAX+2; j++ )
268 x264_free( frame->i_row_satds[i][j] );
269 for( int j = 0; j < 2; j++ )
270 for( int i = 0; i <= X264_BFRAME_MAX; i++ )
272 x264_free( frame->lowres_mvs[j][i] );
273 x264_free( frame->lowres_mv_costs[j][i] );
275 x264_free( frame->i_propagate_cost );
276 for( int j = 0; j <= X264_BFRAME_MAX+1; j++ )
277 for( int i = 0; i <= X264_BFRAME_MAX+1; i++ )
278 x264_free( frame->lowres_costs[j][i] );
279 x264_free( frame->f_qp_offset );
280 x264_free( frame->f_qp_offset_aq );
281 x264_free( frame->i_inv_qscale_factor );
282 x264_free( frame->i_row_bits );
283 x264_free( frame->f_row_qp );
284 x264_free( frame->f_row_qscale );
285 x264_free( frame->field );
286 x264_free( frame->mb_type );
287 x264_free( frame->mb_partition );
288 x264_free( frame->mv[0] );
289 x264_free( frame->mv[1] );
291 x264_free( frame->mv16x16-1 );
292 x264_free( frame->ref[0] );
293 x264_free( frame->ref[1] );
294 x264_pthread_mutex_destroy( &frame->mutex );
295 x264_pthread_cond_destroy( &frame->cv );
300 static int get_plane_ptr( x264_t *h, x264_picture_t *src, uint8_t **pix, int *stride, int plane, int xshift, int yshift )
302 int width = h->param.i_width >> xshift;
303 int height = h->param.i_height >> yshift;
304 *pix = src->img.plane[plane];
305 *stride = src->img.i_stride[plane];
306 if( src->img.i_csp & X264_CSP_VFLIP )
308 *pix += (height-1) * *stride;
311 if( width > abs(*stride) )
313 x264_log( h, X264_LOG_ERROR, "Input picture width (%d) is greater than stride (%d)\n", width, *stride );
319 #define get_plane_ptr(...) do{ if( get_plane_ptr(__VA_ARGS__) < 0 ) return -1; }while(0)
321 int x264_frame_copy_picture( x264_t *h, x264_frame_t *dst, x264_picture_t *src )
323 int i_csp = src->img.i_csp & X264_CSP_MASK;
324 if( i_csp <= X264_CSP_NONE || i_csp >= X264_CSP_MAX )
326 x264_log( h, X264_LOG_ERROR, "Invalid input colorspace\n" );
331 if( !(src->img.i_csp & X264_CSP_HIGH_DEPTH) )
333 x264_log( h, X264_LOG_ERROR, "This build of x264 requires high depth input. Rebuild to support 8-bit input.\n" );
337 if( src->img.i_csp & X264_CSP_HIGH_DEPTH )
339 x264_log( h, X264_LOG_ERROR, "This build of x264 requires 8-bit input. Rebuild to support high depth input.\n" );
344 dst->i_type = src->i_type;
345 dst->i_qpplus1 = src->i_qpplus1;
346 dst->i_pts = dst->i_reordered_pts = src->i_pts;
347 dst->param = src->param;
348 dst->i_pic_struct = src->i_pic_struct;
349 dst->extra_sei = src->extra_sei;
353 if ( i_csp >= X264_CSP_BGR )
355 stride[0] = src->img.i_stride[0];
356 pix[0] = src->img.plane[0];
357 if( src->img.i_csp & X264_CSP_VFLIP )
359 pix[0] += (h->param.i_height-1) * stride[0];
360 stride[0] = -stride[0];
362 int b = i_csp==X264_CSP_RGB ? 2 : 0;
363 h->mc.plane_copy_deinterleave_rgb( dst->plane[1], dst->i_stride[1],
364 dst->plane[b], dst->i_stride[b],
365 dst->plane[2-b], dst->i_stride[2-b],
366 (pixel*)pix[0], stride[0]/sizeof(pixel), i_csp==X264_CSP_BGRA ? 4 : 3, h->param.i_width, h->param.i_height );
370 get_plane_ptr( h, src, &pix[0], &stride[0], 0, 0, 0 );
371 h->mc.plane_copy( dst->plane[0], dst->i_stride[0], (pixel*)pix[0],
372 stride[0]/sizeof(pixel), h->param.i_width, h->param.i_height );
373 if( i_csp == X264_CSP_NV12 )
375 get_plane_ptr( h, src, &pix[1], &stride[1], 1, 0, 1 );
376 h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
377 stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height>>1 );
379 else if( i_csp == X264_CSP_I420 || i_csp == X264_CSP_YV12 )
381 get_plane_ptr( h, src, &pix[1], &stride[1], i_csp==X264_CSP_I420 ? 1 : 2, 1, 1 );
382 get_plane_ptr( h, src, &pix[2], &stride[2], i_csp==X264_CSP_I420 ? 2 : 1, 1, 1 );
383 h->mc.plane_copy_interleave( dst->plane[1], dst->i_stride[1],
384 (pixel*)pix[1], stride[1]/sizeof(pixel),
385 (pixel*)pix[2], stride[2]/sizeof(pixel),
386 h->param.i_width>>1, h->param.i_height>>1 );
388 else //if( i_csp == X264_CSP_I444 || i_csp == X264_CSP_YV24 )
390 get_plane_ptr( h, src, &pix[1], &stride[1], i_csp==X264_CSP_I444 ? 1 : 2, 0, 0 );
391 get_plane_ptr( h, src, &pix[2], &stride[2], i_csp==X264_CSP_I444 ? 2 : 1, 0, 0 );
392 h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
393 stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height );
394 h->mc.plane_copy( dst->plane[2], dst->i_stride[2], (pixel*)pix[2],
395 stride[2]/sizeof(pixel), h->param.i_width, h->param.i_height );
401 static void ALWAYS_INLINE pixel_memset( pixel *dst, pixel *src, int len, int size )
403 uint8_t *dstp = (uint8_t*)dst;
405 uint16_t v2 = size == 1 ? v1 + (v1 << 8) : M16( src );
406 uint32_t v4 = size <= 2 ? v2 + (v2 << 16) : M32( src );
410 /* Align the input pointer if it isn't already */
411 if( (intptr_t)dstp & (WORD_SIZE - 1) )
413 if( size <= 2 && ((intptr_t)dstp & 3) )
415 if( size == 1 && ((intptr_t)dstp & 1) )
417 if( (intptr_t)dstp & 2 )
423 if( WORD_SIZE == 8 && (intptr_t)dstp & 4 )
433 uint64_t v8 = v4 + ((uint64_t)v4<<32);
434 for( ; i < len - 7; i+=8 )
437 for( ; i < len - 3; i+=4 )
440 /* Finish up the last few bytes */
448 if( size == 1 && i != len )
453 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 )
455 #define PPIXEL(x, y) ( pix + (x) + (y)*i_stride )
456 for( int y = 0; y < i_height; y++ )
459 pixel_memset( PPIXEL(-i_padh, y), PPIXEL(0, y), i_padh>>b_chroma, sizeof(pixel)<<b_chroma );
461 pixel_memset( PPIXEL(i_width, y), PPIXEL(i_width-1-b_chroma, y), i_padh>>b_chroma, sizeof(pixel)<<b_chroma );
465 for( int y = 0; y < i_padv; y++ )
466 memcpy( PPIXEL(-i_padh, -y-1), PPIXEL(-i_padh, 0), (i_width+2*i_padh) * sizeof(pixel) );
469 for( int y = 0; y < i_padv; y++ )
470 memcpy( PPIXEL(-i_padh, i_height+y), PPIXEL(-i_padh, i_height-1), (i_width+2*i_padh) * sizeof(pixel) );
474 void x264_frame_expand_border( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
477 if( mb_y & SLICE_MBAFF )
479 for( int i = 0; i < frame->i_plane; i++ )
481 int shift = i && !CHROMA444;
482 int stride = frame->i_stride[i];
483 int width = 16*h->mb.i_mb_width;
484 int height = (b_end ? 16*(h->mb.i_mb_height - mb_y) >> SLICE_MBAFF : 16) >> shift;
486 int padv = PADV >> shift;
487 // buffer: 2 chroma, 3 luma (rounded to 4) because deblocking goes beyond the top of the mb
488 if( b_end && !b_start )
489 height += 4 >> (shift + SLICE_MBAFF);
493 // border samples for each field are extended separately
494 pix = frame->plane_fld[i] + X264_MAX(0, (16*mb_y-4)*stride >> shift);
495 plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end, shift );
496 plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end, shift );
498 height = (b_end ? 16*(h->mb.i_mb_height - mb_y) : 32) >> shift;
499 if( b_end && !b_start )
500 height += 4 >> shift;
501 pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> shift);
502 plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end, shift );
506 pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> shift);
507 plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end, shift );
512 void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
514 /* during filtering, 8 extra pixels were filtered on each edge,
515 * but up to 3 of the horizontal ones may be wrong.
516 we want to expand border from the last filtered pixel */
518 int width = 16*h->mb.i_mb_width + 8;
519 int height = b_end ? (16*(h->mb.i_mb_height - mb_y) >> SLICE_MBAFF) + 16 : 16;
522 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
523 for( int i = 1; i < 4; i++ )
525 int stride = frame->i_stride[p];
526 // buffer: 8 luma, to match the hpel filter
530 pix = frame->filtered_fld[p][i] + (16*mb_y - 16) * stride - 4;
531 plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end, 0 );
532 plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end, 0 );
535 pix = frame->filtered[p][i] + (16*mb_y - 8) * stride - 4;
536 plane_expand_border( pix, stride, width, height << SLICE_MBAFF, padh, padv, b_start, b_end, 0 );
540 void x264_frame_expand_border_lowres( x264_frame_t *frame )
542 for( int i = 0; i < 4; i++ )
543 plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_width_lowres, frame->i_lines_lowres, PADH, PADV, 1, 1, 0 );
546 void x264_frame_expand_border_chroma( x264_t *h, x264_frame_t *frame, int plane )
548 int shift = !CHROMA444;
549 plane_expand_border( frame->plane[plane], frame->i_stride[plane], 16*h->mb.i_mb_width, 16*h->mb.i_mb_height>>shift,
550 PADH, PADV>>shift, 1, 1, shift );
553 void x264_frame_expand_border_mod16( x264_t *h, x264_frame_t *frame )
555 for( int i = 0; i < frame->i_plane; i++ )
557 int i_width = h->param.i_width;
558 int shift = i && !CHROMA444;
559 int i_height = h->param.i_height >> shift;
560 int i_padx = (h->mb.i_mb_width * 16 - h->param.i_width);
561 int i_pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> shift;
565 for( int y = 0; y < i_height; y++ )
566 pixel_memset( &frame->plane[i][y*frame->i_stride[i] + i_width],
567 &frame->plane[i][y*frame->i_stride[i] + i_width - 1-shift],
568 i_padx>>shift, sizeof(pixel)<<shift );
572 for( int y = i_height; y < i_height + i_pady; y++ )
573 memcpy( &frame->plane[i][y*frame->i_stride[i]],
574 &frame->plane[i][(i_height-(~y&PARAM_INTERLACED)-1)*frame->i_stride[i]],
575 (i_width + i_padx) * sizeof(pixel) );
580 void x264_expand_border_mbpair( x264_t *h, int mb_x, int mb_y )
582 for( int i = 0; i < h->fenc->i_plane; i++ )
584 int shift = i && !CHROMA444;
585 int stride = h->fenc->i_stride[i];
586 int height = h->param.i_height >> shift;
587 int pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> shift;
588 pixel *fenc = h->fenc->plane[i] + 16*mb_x;
589 for( int y = height; y < height + pady; y++ )
590 memcpy( fenc + y*stride, fenc + (height-1)*stride, 16*sizeof(pixel) );
595 void x264_frame_cond_broadcast( x264_frame_t *frame, int i_lines_completed )
597 x264_pthread_mutex_lock( &frame->mutex );
598 frame->i_lines_completed = i_lines_completed;
599 x264_pthread_cond_broadcast( &frame->cv );
600 x264_pthread_mutex_unlock( &frame->mutex );
603 void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
605 x264_pthread_mutex_lock( &frame->mutex );
606 while( frame->i_lines_completed < i_lines_completed )
607 x264_pthread_cond_wait( &frame->cv, &frame->mutex );
608 x264_pthread_mutex_unlock( &frame->mutex );
613 void x264_frame_push( x264_frame_t **list, x264_frame_t *frame )
616 while( list[i] ) i++;
620 x264_frame_t *x264_frame_pop( x264_frame_t **list )
625 while( list[i+1] ) i++;
631 void x264_frame_unshift( x264_frame_t **list, x264_frame_t *frame )
634 while( list[i] ) i++;
640 x264_frame_t *x264_frame_shift( x264_frame_t **list )
642 x264_frame_t *frame = list[0];
644 for( i = 0; list[i]; i++ )
650 void x264_frame_push_unused( x264_t *h, x264_frame_t *frame )
652 assert( frame->i_reference_count > 0 );
653 frame->i_reference_count--;
654 if( frame->i_reference_count == 0 )
655 x264_frame_push( h->frames.unused[frame->b_fdec], frame );
658 x264_frame_t *x264_frame_pop_unused( x264_t *h, int b_fdec )
661 if( h->frames.unused[b_fdec][0] )
662 frame = x264_frame_pop( h->frames.unused[b_fdec] );
664 frame = x264_frame_new( h, b_fdec );
667 frame->b_last_minigop_bframe = 0;
668 frame->i_reference_count = 1;
669 frame->b_intra_calculated = 0;
670 frame->b_scenecut = 1;
671 frame->b_keyframe = 0;
672 frame->b_corrupt = 0;
674 memset( frame->weight, 0, sizeof(frame->weight) );
675 memset( frame->f_weighted_cost_delta, 0, sizeof(frame->f_weighted_cost_delta) );
680 void x264_frame_push_blank_unused( x264_t *h, x264_frame_t *frame )
682 assert( frame->i_reference_count > 0 );
683 frame->i_reference_count--;
684 if( frame->i_reference_count == 0 )
685 x264_frame_push( h->frames.blank_unused, frame );
688 x264_frame_t *x264_frame_pop_blank_unused( x264_t *h )
691 if( h->frames.blank_unused[0] )
692 frame = x264_frame_pop( h->frames.blank_unused );
694 frame = x264_malloc( sizeof(x264_frame_t) );
697 frame->b_duplicate = 1;
698 frame->i_reference_count = 1;
702 void x264_weight_scale_plane( x264_t *h, pixel *dst, int i_dst_stride, pixel *src, int i_src_stride,
703 int i_width, int i_height, x264_weight_t *w )
705 /* Weight horizontal strips of height 16. This was found to be the optimal height
706 * in terms of the cache loads. */
707 while( i_height > 0 )
709 for( int x = 0; x < i_width; x += 16 )
710 w->weightfn[16>>2]( dst+x, i_dst_stride, src+x, i_src_stride, w, X264_MIN( i_height, 16 ) );
712 dst += 16 * i_dst_stride;
713 src += 16 * i_src_stride;
717 void x264_frame_delete_list( x264_frame_t **list )
723 x264_frame_delete( list[i++] );
727 int x264_sync_frame_list_init( x264_sync_frame_list_t *slist, int max_size )
731 slist->i_max_size = max_size;
733 CHECKED_MALLOCZERO( slist->list, (max_size+1) * sizeof(x264_frame_t*) );
734 if( x264_pthread_mutex_init( &slist->mutex, NULL ) ||
735 x264_pthread_cond_init( &slist->cv_fill, NULL ) ||
736 x264_pthread_cond_init( &slist->cv_empty, NULL ) )
743 void x264_sync_frame_list_delete( x264_sync_frame_list_t *slist )
745 x264_pthread_mutex_destroy( &slist->mutex );
746 x264_pthread_cond_destroy( &slist->cv_fill );
747 x264_pthread_cond_destroy( &slist->cv_empty );
748 x264_frame_delete_list( slist->list );
751 void x264_sync_frame_list_push( x264_sync_frame_list_t *slist, x264_frame_t *frame )
753 x264_pthread_mutex_lock( &slist->mutex );
754 while( slist->i_size == slist->i_max_size )
755 x264_pthread_cond_wait( &slist->cv_empty, &slist->mutex );
756 slist->list[ slist->i_size++ ] = frame;
757 x264_pthread_mutex_unlock( &slist->mutex );
758 x264_pthread_cond_broadcast( &slist->cv_fill );
761 x264_frame_t *x264_sync_frame_list_pop( x264_sync_frame_list_t *slist )
764 x264_pthread_mutex_lock( &slist->mutex );
765 while( !slist->i_size )
766 x264_pthread_cond_wait( &slist->cv_fill, &slist->mutex );
767 frame = slist->list[ --slist->i_size ];
768 slist->list[ slist->i_size ] = NULL;
769 x264_pthread_cond_broadcast( &slist->cv_empty );
770 x264_pthread_mutex_unlock( &slist->mutex );