1 /*****************************************************************************
2 * frame.c: frame handling
3 *****************************************************************************
4 * Copyright (C) 2003-2012 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 )
68 static x264_frame_t *x264_frame_new( x264_t *h, int b_fdec )
71 int i_csp = x264_frame_internal_csp( h->param.i_csp );
72 int i_mb_count = h->mb.i_mb_count;
73 int i_stride, i_width, i_lines, luma_plane_count;
74 int i_padv = PADV << PARAM_INTERLACED;
75 int align = h->param.cpu&X264_CPU_CACHELINE_64 ? 64 : h->param.cpu&X264_CPU_CACHELINE_32 ? 32 : 16;
76 int disalign = h->param.cpu&X264_CPU_ALTIVEC ? 1<<9 : 1<<10;
78 CHECKED_MALLOCZERO( frame, sizeof(x264_frame_t) );
80 /* allocate frame data (+64 for extra data for me) */
81 i_width = h->mb.i_mb_width*16;
82 i_lines = h->mb.i_mb_height*16;
83 i_stride = align_stride( i_width + 2*PADH, align, disalign );
85 if( i_csp == X264_CSP_NV12 || i_csp == X264_CSP_NV16 )
89 for( int i = 0; i < 2; i++ )
91 frame->i_width[i] = i_width >> i;
92 frame->i_lines[i] = i_lines >> (i && i_csp == X264_CSP_NV12);
93 frame->i_stride[i] = i_stride;
96 else if( i_csp == X264_CSP_I444 )
100 for( int i = 0; i < 3; i++ )
102 frame->i_width[i] = i_width;
103 frame->i_lines[i] = i_lines;
104 frame->i_stride[i] = i_stride;
110 frame->i_csp = i_csp;
111 frame->i_width_lowres = frame->i_width[0]/2;
112 frame->i_lines_lowres = frame->i_lines[0]/2;
113 frame->i_stride_lowres = align_stride( frame->i_width_lowres + 2*PADH, align, disalign<<1 );
115 for( int i = 0; i < h->param.i_bframe + 2; i++ )
116 for( int j = 0; j < h->param.i_bframe + 2; j++ )
117 CHECKED_MALLOC( frame->i_row_satds[i][j], i_lines/16 * sizeof(int) );
120 frame->i_type = X264_TYPE_AUTO;
121 frame->i_qpplus1 = X264_QP_AUTO;
124 frame->i_frame_num = -1;
125 frame->i_lines_completed = -1;
126 frame->b_fdec = b_fdec;
127 frame->i_pic_struct = PIC_STRUCT_AUTO;
128 frame->i_field_cnt = -1;
130 frame->i_cpb_duration =
131 frame->i_dpb_output_delay =
132 frame->i_cpb_delay = 0;
133 frame->i_coded_fields_lookahead =
134 frame->i_cpb_delay_lookahead = -1;
138 if( i_csp == X264_CSP_NV12 || i_csp == X264_CSP_NV16 )
140 int chroma_padv = i_padv >> (i_csp == X264_CSP_NV12);
141 int chroma_plane_size = (frame->i_stride[1] * (frame->i_lines[1] + 2*chroma_padv));
142 CHECKED_MALLOC( frame->buffer[1], chroma_plane_size * sizeof(pixel) );
143 frame->plane[1] = frame->buffer[1] + frame->i_stride[1] * chroma_padv + PADH;
144 if( PARAM_INTERLACED )
146 CHECKED_MALLOC( frame->buffer_fld[1], chroma_plane_size * sizeof(pixel) );
147 frame->plane_fld[1] = frame->buffer_fld[1] + frame->i_stride[1] * chroma_padv + PADH;
151 /* all 4 luma planes allocated together, since the cacheline split code
152 * requires them to be in-phase wrt cacheline alignment. */
154 for( int p = 0; p < luma_plane_count; p++ )
156 int luma_plane_size = align_plane_size( frame->i_stride[p] * (frame->i_lines[p] + 2*i_padv), disalign );
157 if( h->param.analyse.i_subpel_refine && b_fdec )
159 /* FIXME: Don't allocate both buffers in non-adaptive MBAFF. */
160 CHECKED_MALLOC( frame->buffer[p], 4*luma_plane_size * sizeof(pixel) );
161 if( PARAM_INTERLACED )
162 CHECKED_MALLOC( frame->buffer_fld[p], 4*luma_plane_size * sizeof(pixel) );
163 for( int i = 0; i < 4; i++ )
165 frame->filtered[p][i] = frame->buffer[p] + i*luma_plane_size + frame->i_stride[p] * i_padv + PADH;
166 frame->filtered_fld[p][i] = frame->buffer_fld[p] + i*luma_plane_size + frame->i_stride[p] * i_padv + PADH;
168 frame->plane[p] = frame->filtered[p][0];
169 frame->plane_fld[p] = frame->filtered_fld[p][0];
173 CHECKED_MALLOC( frame->buffer[p], luma_plane_size * sizeof(pixel) );
174 if( PARAM_INTERLACED )
175 CHECKED_MALLOC( frame->buffer_fld[p], luma_plane_size * sizeof(pixel) );
176 frame->filtered[p][0] = frame->plane[p] = frame->buffer[p] + frame->i_stride[p] * i_padv + PADH;
177 frame->filtered_fld[p][0] = frame->plane_fld[p] = frame->buffer_fld[p] + frame->i_stride[p] * i_padv + PADH;
181 frame->b_duplicate = 0;
183 if( b_fdec ) /* fdec frame */
185 CHECKED_MALLOC( frame->mb_type, i_mb_count * sizeof(int8_t));
186 CHECKED_MALLOC( frame->mb_partition, i_mb_count * sizeof(uint8_t));
187 CHECKED_MALLOC( frame->mv[0], 2*16 * i_mb_count * sizeof(int16_t) );
188 CHECKED_MALLOC( frame->mv16x16, 2*(i_mb_count+1) * sizeof(int16_t) );
189 M32( frame->mv16x16[0] ) = 0;
191 CHECKED_MALLOC( frame->ref[0], 4 * i_mb_count * sizeof(int8_t) );
192 if( h->param.i_bframe )
194 CHECKED_MALLOC( frame->mv[1], 2*16 * i_mb_count * sizeof(int16_t) );
195 CHECKED_MALLOC( frame->ref[1], 4 * i_mb_count * sizeof(int8_t) );
200 frame->ref[1] = NULL;
202 CHECKED_MALLOC( frame->i_row_bits, i_lines/16 * sizeof(int) );
203 CHECKED_MALLOC( frame->f_row_qp, i_lines/16 * sizeof(float) );
204 CHECKED_MALLOC( frame->f_row_qscale, i_lines/16 * sizeof(float) );
205 if( h->param.analyse.i_me_method >= X264_ME_ESA )
207 CHECKED_MALLOC( frame->buffer[3],
208 frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) << h->frames.b_have_sub8x8_esa );
209 frame->integral = (uint16_t*)frame->buffer[3] + frame->i_stride[0] * i_padv + PADH;
211 if( PARAM_INTERLACED )
212 CHECKED_MALLOC( frame->field, i_mb_count * sizeof(uint8_t) );
214 else /* fenc frame */
216 if( h->frames.b_have_lowres )
218 int luma_plane_size = align_plane_size( frame->i_stride_lowres * (frame->i_lines[0]/2 + 2*PADV), disalign );
220 CHECKED_MALLOC( frame->buffer_lowres[0], 4 * luma_plane_size * sizeof(pixel) );
221 for( int i = 0; i < 4; i++ )
222 frame->lowres[i] = frame->buffer_lowres[0] + (frame->i_stride_lowres * PADV + PADH) + i * luma_plane_size;
224 for( int j = 0; j <= !!h->param.i_bframe; j++ )
225 for( int i = 0; i <= h->param.i_bframe; i++ )
227 CHECKED_MALLOCZERO( frame->lowres_mvs[j][i], 2*h->mb.i_mb_count*sizeof(int16_t) );
228 CHECKED_MALLOC( frame->lowres_mv_costs[j][i], h->mb.i_mb_count*sizeof(int) );
230 CHECKED_MALLOC( frame->i_propagate_cost, (i_mb_count+7) * sizeof(uint16_t) );
231 for( int j = 0; j <= h->param.i_bframe+1; j++ )
232 for( int i = 0; i <= h->param.i_bframe+1; i++ )
233 CHECKED_MALLOC( frame->lowres_costs[j][i], (i_mb_count+3) * sizeof(uint16_t) );
234 frame->i_intra_cost = frame->lowres_costs[0][0];
235 memset( frame->i_intra_cost, -1, (i_mb_count+3) * sizeof(uint16_t) );
237 if( h->param.rc.i_aq_mode )
239 CHECKED_MALLOC( frame->f_qp_offset, h->mb.i_mb_count * sizeof(float) );
240 CHECKED_MALLOC( frame->f_qp_offset_aq, h->mb.i_mb_count * sizeof(float) );
241 if( h->frames.b_have_lowres )
242 /* shouldn't really be initialized, just silences a valgrind false-positive in x264_mbtree_propagate_cost_sse2 */
243 CHECKED_MALLOCZERO( frame->i_inv_qscale_factor, (h->mb.i_mb_count+3) * sizeof(uint16_t) );
247 if( x264_pthread_mutex_init( &frame->mutex, NULL ) )
249 if( x264_pthread_cond_init( &frame->cv, NULL ) )
259 void x264_frame_delete( x264_frame_t *frame )
261 /* Duplicate frames are blank copies of real frames (including pointers),
262 * so freeing those pointers would cause a double free later. */
263 if( !frame->b_duplicate )
265 for( int i = 0; i < 4; i++ )
267 x264_free( frame->buffer[i] );
268 x264_free( frame->buffer_fld[i] );
270 for( int i = 0; i < 4; i++ )
271 x264_free( frame->buffer_lowres[i] );
272 for( int i = 0; i < X264_BFRAME_MAX+2; i++ )
273 for( int j = 0; j < X264_BFRAME_MAX+2; j++ )
274 x264_free( frame->i_row_satds[i][j] );
275 for( int j = 0; j < 2; j++ )
276 for( int i = 0; i <= X264_BFRAME_MAX; i++ )
278 x264_free( frame->lowres_mvs[j][i] );
279 x264_free( frame->lowres_mv_costs[j][i] );
281 x264_free( frame->i_propagate_cost );
282 for( int j = 0; j <= X264_BFRAME_MAX+1; j++ )
283 for( int i = 0; i <= X264_BFRAME_MAX+1; i++ )
284 x264_free( frame->lowres_costs[j][i] );
285 x264_free( frame->f_qp_offset );
286 x264_free( frame->f_qp_offset_aq );
287 x264_free( frame->i_inv_qscale_factor );
288 x264_free( frame->i_row_bits );
289 x264_free( frame->f_row_qp );
290 x264_free( frame->f_row_qscale );
291 x264_free( frame->field );
292 x264_free( frame->mb_type );
293 x264_free( frame->mb_partition );
294 x264_free( frame->mv[0] );
295 x264_free( frame->mv[1] );
297 x264_free( frame->mv16x16-1 );
298 x264_free( frame->ref[0] );
299 x264_free( frame->ref[1] );
300 x264_pthread_mutex_destroy( &frame->mutex );
301 x264_pthread_cond_destroy( &frame->cv );
306 static int get_plane_ptr( x264_t *h, x264_picture_t *src, uint8_t **pix, int *stride, int plane, int xshift, int yshift )
308 int width = h->param.i_width >> xshift;
309 int height = h->param.i_height >> yshift;
310 *pix = src->img.plane[plane];
311 *stride = src->img.i_stride[plane];
312 if( src->img.i_csp & X264_CSP_VFLIP )
314 *pix += (height-1) * *stride;
317 if( width > abs(*stride) )
319 x264_log( h, X264_LOG_ERROR, "Input picture width (%d) is greater than stride (%d)\n", width, *stride );
325 #define get_plane_ptr(...) do{ if( get_plane_ptr(__VA_ARGS__) < 0 ) return -1; }while(0)
327 int x264_frame_copy_picture( x264_t *h, x264_frame_t *dst, x264_picture_t *src )
329 int i_csp = src->img.i_csp & X264_CSP_MASK;
330 if( dst->i_csp != x264_frame_internal_csp( i_csp ) )
332 x264_log( h, X264_LOG_ERROR, "Invalid input colorspace\n" );
337 if( !(src->img.i_csp & X264_CSP_HIGH_DEPTH) )
339 x264_log( h, X264_LOG_ERROR, "This build of x264 requires high depth input. Rebuild to support 8-bit input.\n" );
343 if( src->img.i_csp & X264_CSP_HIGH_DEPTH )
345 x264_log( h, X264_LOG_ERROR, "This build of x264 requires 8-bit input. Rebuild to support high depth input.\n" );
350 dst->i_type = src->i_type;
351 dst->i_qpplus1 = src->i_qpplus1;
352 dst->i_pts = dst->i_reordered_pts = src->i_pts;
353 dst->param = src->param;
354 dst->i_pic_struct = src->i_pic_struct;
355 dst->extra_sei = src->extra_sei;
356 dst->opaque = src->opaque;
360 if ( i_csp >= X264_CSP_BGR )
362 stride[0] = src->img.i_stride[0];
363 pix[0] = src->img.plane[0];
364 if( src->img.i_csp & X264_CSP_VFLIP )
366 pix[0] += (h->param.i_height-1) * stride[0];
367 stride[0] = -stride[0];
369 int b = i_csp==X264_CSP_RGB;
370 h->mc.plane_copy_deinterleave_rgb( dst->plane[1+b], dst->i_stride[1+b],
371 dst->plane[0], dst->i_stride[0],
372 dst->plane[2-b], dst->i_stride[2-b],
373 (pixel*)pix[0], stride[0]/sizeof(pixel), i_csp==X264_CSP_BGRA ? 4 : 3, h->param.i_width, h->param.i_height );
377 int v_shift = CHROMA_V_SHIFT;
378 get_plane_ptr( h, src, &pix[0], &stride[0], 0, 0, 0 );
379 h->mc.plane_copy( dst->plane[0], dst->i_stride[0], (pixel*)pix[0],
380 stride[0]/sizeof(pixel), h->param.i_width, h->param.i_height );
381 if( i_csp == X264_CSP_NV12 || i_csp == X264_CSP_NV16 )
383 get_plane_ptr( h, src, &pix[1], &stride[1], 1, 0, v_shift );
384 h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
385 stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height>>v_shift );
387 else if( i_csp == X264_CSP_I420 || i_csp == X264_CSP_I422 || i_csp == X264_CSP_YV12 || i_csp == X264_CSP_YV16 )
389 int uv_swap = i_csp == X264_CSP_YV12 || i_csp == X264_CSP_YV16;
390 get_plane_ptr( h, src, &pix[1], &stride[1], uv_swap ? 2 : 1, 1, v_shift );
391 get_plane_ptr( h, src, &pix[2], &stride[2], uv_swap ? 1 : 2, 1, v_shift );
392 h->mc.plane_copy_interleave( dst->plane[1], dst->i_stride[1],
393 (pixel*)pix[1], stride[1]/sizeof(pixel),
394 (pixel*)pix[2], stride[2]/sizeof(pixel),
395 h->param.i_width>>1, h->param.i_height>>v_shift );
397 else //if( i_csp == X264_CSP_I444 || i_csp == X264_CSP_YV24 )
399 get_plane_ptr( h, src, &pix[1], &stride[1], i_csp==X264_CSP_I444 ? 1 : 2, 0, 0 );
400 get_plane_ptr( h, src, &pix[2], &stride[2], i_csp==X264_CSP_I444 ? 2 : 1, 0, 0 );
401 h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
402 stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height );
403 h->mc.plane_copy( dst->plane[2], dst->i_stride[2], (pixel*)pix[2],
404 stride[2]/sizeof(pixel), h->param.i_width, h->param.i_height );
410 static void ALWAYS_INLINE pixel_memset( pixel *dst, pixel *src, int len, int size )
412 uint8_t *dstp = (uint8_t*)dst;
414 uint32_t v2 = size == 1 ? v1 + (v1 << 8) : M16( src );
415 uint32_t v4 = size <= 2 ? v2 + (v2 << 16) : M32( src );
419 /* Align the input pointer if it isn't already */
420 if( (intptr_t)dstp & (WORD_SIZE - 1) )
422 if( size <= 2 && ((intptr_t)dstp & 3) )
424 if( size == 1 && ((intptr_t)dstp & 1) )
426 if( (intptr_t)dstp & 2 )
432 if( WORD_SIZE == 8 && (intptr_t)dstp & 4 )
442 uint64_t v8 = v4 + ((uint64_t)v4<<32);
443 for( ; i < len - 7; i+=8 )
446 for( ; i < len - 3; i+=4 )
449 /* Finish up the last few bytes */
457 if( size == 1 && i != len )
462 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 )
464 #define PPIXEL(x, y) ( pix + (x) + (y)*i_stride )
465 for( int y = 0; y < i_height; y++ )
468 pixel_memset( PPIXEL(-i_padh, y), PPIXEL(0, y), i_padh>>b_chroma, sizeof(pixel)<<b_chroma );
470 pixel_memset( PPIXEL(i_width, y), PPIXEL(i_width-1-b_chroma, y), i_padh>>b_chroma, sizeof(pixel)<<b_chroma );
474 for( int y = 0; y < i_padv; y++ )
475 memcpy( PPIXEL(-i_padh, -y-1), PPIXEL(-i_padh, 0), (i_width+2*i_padh) * sizeof(pixel) );
478 for( int y = 0; y < i_padv; y++ )
479 memcpy( PPIXEL(-i_padh, i_height+y), PPIXEL(-i_padh, i_height-1), (i_width+2*i_padh) * sizeof(pixel) );
483 void x264_frame_expand_border( x264_t *h, x264_frame_t *frame, int mb_y )
485 int pad_top = mb_y == 0;
486 int pad_bot = mb_y == h->mb.i_mb_height - (1 << SLICE_MBAFF);
487 int b_start = mb_y == h->i_threadslice_start;
488 int b_end = mb_y == h->i_threadslice_end - (1 << SLICE_MBAFF);
489 if( mb_y & SLICE_MBAFF )
491 for( int i = 0; i < frame->i_plane; i++ )
493 int h_shift = i && CHROMA_H_SHIFT;
494 int v_shift = i && CHROMA_V_SHIFT;
495 int stride = frame->i_stride[i];
496 int width = 16*h->mb.i_mb_width;
497 int height = (pad_bot ? 16*(h->mb.i_mb_height - mb_y) >> SLICE_MBAFF : 16) >> v_shift;
499 int padv = PADV >> v_shift;
500 // buffer: 2 chroma, 3 luma (rounded to 4) because deblocking goes beyond the top of the mb
501 if( b_end && !b_start )
502 height += 4 >> (v_shift + SLICE_MBAFF);
504 int starty = 16*mb_y - 4*!b_start;
507 // border samples for each field are extended separately
508 pix = frame->plane_fld[i] + (starty*stride >> v_shift);
509 plane_expand_border( pix, stride*2, width, height, padh, padv, pad_top, pad_bot, h_shift );
510 plane_expand_border( pix+stride, stride*2, width, height, padh, padv, pad_top, pad_bot, h_shift );
512 height = (pad_bot ? 16*(h->mb.i_mb_height - mb_y) : 32) >> v_shift;
513 if( b_end && !b_start )
514 height += 4 >> v_shift;
515 pix = frame->plane[i] + (starty*stride >> v_shift);
516 plane_expand_border( pix, stride, width, height, padh, padv, pad_top, pad_bot, h_shift );
520 pix = frame->plane[i] + (starty*stride >> v_shift);
521 plane_expand_border( pix, stride, width, height, padh, padv, pad_top, pad_bot, h_shift );
526 void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
528 /* during filtering, 8 extra pixels were filtered on each edge,
529 * but up to 3 of the horizontal ones may be wrong.
530 we want to expand border from the last filtered pixel */
532 int width = 16*h->mb.i_mb_width + 8;
533 int height = b_end ? (16*(h->mb.i_mb_height - mb_y) >> SLICE_MBAFF) + 16 : 16;
536 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
537 for( int i = 1; i < 4; i++ )
539 int stride = frame->i_stride[p];
540 // buffer: 8 luma, to match the hpel filter
544 pix = frame->filtered_fld[p][i] + (16*mb_y - 16) * stride - 4;
545 plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end, 0 );
546 plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end, 0 );
549 pix = frame->filtered[p][i] + (16*mb_y - 8) * stride - 4;
550 plane_expand_border( pix, stride, width, height << SLICE_MBAFF, padh, padv, b_start, b_end, 0 );
554 void x264_frame_expand_border_lowres( x264_frame_t *frame )
556 for( int i = 0; i < 4; i++ )
557 plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_width_lowres, frame->i_lines_lowres, PADH, PADV, 1, 1, 0 );
560 void x264_frame_expand_border_chroma( x264_t *h, x264_frame_t *frame, int plane )
562 int v_shift = CHROMA_V_SHIFT;
563 plane_expand_border( frame->plane[plane], frame->i_stride[plane], 16*h->mb.i_mb_width, 16*h->mb.i_mb_height>>v_shift,
564 PADH, PADV>>v_shift, 1, 1, CHROMA_H_SHIFT );
567 void x264_frame_expand_border_mod16( x264_t *h, x264_frame_t *frame )
569 for( int i = 0; i < frame->i_plane; i++ )
571 int i_width = h->param.i_width;
572 int h_shift = i && CHROMA_H_SHIFT;
573 int v_shift = i && CHROMA_V_SHIFT;
574 int i_height = h->param.i_height >> v_shift;
575 int i_padx = (h->mb.i_mb_width * 16 - h->param.i_width);
576 int i_pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> v_shift;
580 for( int y = 0; y < i_height; y++ )
581 pixel_memset( &frame->plane[i][y*frame->i_stride[i] + i_width],
582 &frame->plane[i][y*frame->i_stride[i] + i_width - 1-h_shift],
583 i_padx>>h_shift, sizeof(pixel)<<h_shift );
587 for( int y = i_height; y < i_height + i_pady; y++ )
588 memcpy( &frame->plane[i][y*frame->i_stride[i]],
589 &frame->plane[i][(i_height-(~y&PARAM_INTERLACED)-1)*frame->i_stride[i]],
590 (i_width + i_padx) * sizeof(pixel) );
595 void x264_expand_border_mbpair( x264_t *h, int mb_x, int mb_y )
597 for( int i = 0; i < h->fenc->i_plane; i++ )
599 int v_shift = i && CHROMA_V_SHIFT;
600 int stride = h->fenc->i_stride[i];
601 int height = h->param.i_height >> v_shift;
602 int pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> v_shift;
603 pixel *fenc = h->fenc->plane[i] + 16*mb_x;
604 for( int y = height; y < height + pady; y++ )
605 memcpy( fenc + y*stride, fenc + (height-1)*stride, 16*sizeof(pixel) );
610 void x264_frame_cond_broadcast( x264_frame_t *frame, int i_lines_completed )
612 x264_pthread_mutex_lock( &frame->mutex );
613 frame->i_lines_completed = i_lines_completed;
614 x264_pthread_cond_broadcast( &frame->cv );
615 x264_pthread_mutex_unlock( &frame->mutex );
618 void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
620 x264_pthread_mutex_lock( &frame->mutex );
621 while( frame->i_lines_completed < i_lines_completed )
622 x264_pthread_cond_wait( &frame->cv, &frame->mutex );
623 x264_pthread_mutex_unlock( &frame->mutex );
626 void x264_threadslice_cond_broadcast( x264_t *h, int pass )
628 x264_pthread_mutex_lock( &h->mutex );
629 h->i_threadslice_pass = pass;
631 x264_pthread_cond_broadcast( &h->cv );
632 x264_pthread_mutex_unlock( &h->mutex );
635 void x264_threadslice_cond_wait( x264_t *h, int pass )
637 x264_pthread_mutex_lock( &h->mutex );
638 while( h->i_threadslice_pass < pass )
639 x264_pthread_cond_wait( &h->cv, &h->mutex );
640 x264_pthread_mutex_unlock( &h->mutex );
645 void x264_frame_push( x264_frame_t **list, x264_frame_t *frame )
648 while( list[i] ) i++;
652 x264_frame_t *x264_frame_pop( x264_frame_t **list )
657 while( list[i+1] ) i++;
663 void x264_frame_unshift( x264_frame_t **list, x264_frame_t *frame )
666 while( list[i] ) i++;
672 x264_frame_t *x264_frame_shift( x264_frame_t **list )
674 x264_frame_t *frame = list[0];
676 for( i = 0; list[i]; i++ )
682 void x264_frame_push_unused( x264_t *h, x264_frame_t *frame )
684 assert( frame->i_reference_count > 0 );
685 frame->i_reference_count--;
686 if( frame->i_reference_count == 0 )
687 x264_frame_push( h->frames.unused[frame->b_fdec], frame );
690 x264_frame_t *x264_frame_pop_unused( x264_t *h, int b_fdec )
693 if( h->frames.unused[b_fdec][0] )
694 frame = x264_frame_pop( h->frames.unused[b_fdec] );
696 frame = x264_frame_new( h, b_fdec );
699 frame->b_last_minigop_bframe = 0;
700 frame->i_reference_count = 1;
701 frame->b_intra_calculated = 0;
702 frame->b_scenecut = 1;
703 frame->b_keyframe = 0;
704 frame->b_corrupt = 0;
706 memset( frame->weight, 0, sizeof(frame->weight) );
707 memset( frame->f_weighted_cost_delta, 0, sizeof(frame->f_weighted_cost_delta) );
712 void x264_frame_push_blank_unused( x264_t *h, x264_frame_t *frame )
714 assert( frame->i_reference_count > 0 );
715 frame->i_reference_count--;
716 if( frame->i_reference_count == 0 )
717 x264_frame_push( h->frames.blank_unused, frame );
720 x264_frame_t *x264_frame_pop_blank_unused( x264_t *h )
723 if( h->frames.blank_unused[0] )
724 frame = x264_frame_pop( h->frames.blank_unused );
726 frame = x264_malloc( sizeof(x264_frame_t) );
729 frame->b_duplicate = 1;
730 frame->i_reference_count = 1;
734 void x264_weight_scale_plane( x264_t *h, pixel *dst, intptr_t i_dst_stride, pixel *src, intptr_t i_src_stride,
735 int i_width, int i_height, x264_weight_t *w )
737 /* Weight horizontal strips of height 16. This was found to be the optimal height
738 * in terms of the cache loads. */
739 while( i_height > 0 )
742 for( x = 0; x < i_width-8; x += 16 )
743 w->weightfn[16>>2]( dst+x, i_dst_stride, src+x, i_src_stride, w, X264_MIN( i_height, 16 ) );
745 w->weightfn[ 8>>2]( dst+x, i_dst_stride, src+x, i_src_stride, w, X264_MIN( i_height, 16 ) );
747 dst += 16 * i_dst_stride;
748 src += 16 * i_src_stride;
752 void x264_frame_delete_list( x264_frame_t **list )
758 x264_frame_delete( list[i++] );
762 int x264_sync_frame_list_init( x264_sync_frame_list_t *slist, int max_size )
766 slist->i_max_size = max_size;
768 CHECKED_MALLOCZERO( slist->list, (max_size+1) * sizeof(x264_frame_t*) );
769 if( x264_pthread_mutex_init( &slist->mutex, NULL ) ||
770 x264_pthread_cond_init( &slist->cv_fill, NULL ) ||
771 x264_pthread_cond_init( &slist->cv_empty, NULL ) )
778 void x264_sync_frame_list_delete( x264_sync_frame_list_t *slist )
780 x264_pthread_mutex_destroy( &slist->mutex );
781 x264_pthread_cond_destroy( &slist->cv_fill );
782 x264_pthread_cond_destroy( &slist->cv_empty );
783 x264_frame_delete_list( slist->list );
786 void x264_sync_frame_list_push( x264_sync_frame_list_t *slist, x264_frame_t *frame )
788 x264_pthread_mutex_lock( &slist->mutex );
789 while( slist->i_size == slist->i_max_size )
790 x264_pthread_cond_wait( &slist->cv_empty, &slist->mutex );
791 slist->list[ slist->i_size++ ] = frame;
792 x264_pthread_mutex_unlock( &slist->mutex );
793 x264_pthread_cond_broadcast( &slist->cv_fill );
796 x264_frame_t *x264_sync_frame_list_pop( x264_sync_frame_list_t *slist )
799 x264_pthread_mutex_lock( &slist->mutex );
800 while( !slist->i_size )
801 x264_pthread_cond_wait( &slist->cv_fill, &slist->mutex );
802 frame = slist->list[ --slist->i_size ];
803 slist->list[ slist->i_size ] = NULL;
804 x264_pthread_cond_broadcast( &slist->cv_empty );
805 x264_pthread_mutex_unlock( &slist->mutex );