1 /*****************************************************************************
2 * frame.c: frame handling
3 *****************************************************************************
4 * Copyright (C) 2003-2016 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 )
70 static x264_frame_t *x264_frame_new( x264_t *h, int b_fdec )
73 int i_csp = x264_frame_internal_csp( h->param.i_csp );
74 int i_mb_count = h->mb.i_mb_count;
75 int i_stride, i_width, i_lines, luma_plane_count;
76 int i_padv = PADV << PARAM_INTERLACED;
78 #if ARCH_X86 || ARCH_X86_64
79 if( h->param.cpu&X264_CPU_CACHELINE_64 )
81 else if( h->param.cpu&X264_CPU_CACHELINE_32 || h->param.cpu&X264_CPU_AVX )
90 CHECKED_MALLOCZERO( frame, sizeof(x264_frame_t) );
93 /* allocate frame data (+64 for extra data for me) */
94 i_width = h->mb.i_mb_width*16;
95 i_lines = h->mb.i_mb_height*16;
96 i_stride = align_stride( i_width + 2*PADH, align, disalign );
98 if( i_csp == X264_CSP_NV12 || i_csp == X264_CSP_NV16 )
100 luma_plane_count = 1;
102 for( int i = 0; i < 2; i++ )
104 frame->i_width[i] = i_width >> i;
105 frame->i_lines[i] = i_lines >> (i && i_csp == X264_CSP_NV12);
106 frame->i_stride[i] = i_stride;
109 else if( i_csp == X264_CSP_I444 )
111 luma_plane_count = 3;
113 for( int i = 0; i < 3; i++ )
115 frame->i_width[i] = i_width;
116 frame->i_lines[i] = i_lines;
117 frame->i_stride[i] = i_stride;
123 frame->i_csp = i_csp;
124 frame->i_width_lowres = frame->i_width[0]/2;
125 frame->i_lines_lowres = frame->i_lines[0]/2;
126 frame->i_stride_lowres = align_stride( frame->i_width_lowres + 2*PADH, align, disalign<<1 );
128 for( int i = 0; i < h->param.i_bframe + 2; i++ )
129 for( int j = 0; j < h->param.i_bframe + 2; j++ )
130 PREALLOC( frame->i_row_satds[i][j], i_lines/16 * sizeof(int) );
133 frame->i_type = X264_TYPE_AUTO;
134 frame->i_qpplus1 = X264_QP_AUTO;
137 frame->i_frame_num = -1;
138 frame->i_lines_completed = -1;
139 frame->b_fdec = b_fdec;
140 frame->i_pic_struct = PIC_STRUCT_AUTO;
141 frame->i_field_cnt = -1;
143 frame->i_cpb_duration =
144 frame->i_dpb_output_delay =
145 frame->i_cpb_delay = 0;
146 frame->i_coded_fields_lookahead =
147 frame->i_cpb_delay_lookahead = -1;
151 if( i_csp == X264_CSP_NV12 || i_csp == X264_CSP_NV16 )
153 int chroma_padv = i_padv >> (i_csp == X264_CSP_NV12);
154 int chroma_plane_size = (frame->i_stride[1] * (frame->i_lines[1] + 2*chroma_padv));
155 PREALLOC( frame->buffer[1], chroma_plane_size * sizeof(pixel) );
156 if( PARAM_INTERLACED )
157 PREALLOC( frame->buffer_fld[1], chroma_plane_size * sizeof(pixel) );
160 /* all 4 luma planes allocated together, since the cacheline split code
161 * requires them to be in-phase wrt cacheline alignment. */
163 for( int p = 0; p < luma_plane_count; p++ )
165 int luma_plane_size = align_plane_size( frame->i_stride[p] * (frame->i_lines[p] + 2*i_padv), disalign );
166 if( h->param.analyse.i_subpel_refine && b_fdec )
168 /* FIXME: Don't allocate both buffers in non-adaptive MBAFF. */
169 PREALLOC( frame->buffer[p], 4*luma_plane_size * sizeof(pixel) );
170 if( PARAM_INTERLACED )
171 PREALLOC( frame->buffer_fld[p], 4*luma_plane_size * sizeof(pixel) );
175 PREALLOC( frame->buffer[p], luma_plane_size * sizeof(pixel) );
176 if( PARAM_INTERLACED )
177 PREALLOC( frame->buffer_fld[p], luma_plane_size * sizeof(pixel) );
181 frame->b_duplicate = 0;
183 if( b_fdec ) /* fdec frame */
185 PREALLOC( frame->mb_type, i_mb_count * sizeof(int8_t) );
186 PREALLOC( frame->mb_partition, i_mb_count * sizeof(uint8_t) );
187 PREALLOC( frame->mv[0], 2*16 * i_mb_count * sizeof(int16_t) );
188 PREALLOC( frame->mv16x16, 2*(i_mb_count+1) * sizeof(int16_t) );
189 PREALLOC( frame->ref[0], 4 * i_mb_count * sizeof(int8_t) );
190 if( h->param.i_bframe )
192 PREALLOC( frame->mv[1], 2*16 * i_mb_count * sizeof(int16_t) );
193 PREALLOC( frame->ref[1], 4 * i_mb_count * sizeof(int8_t) );
198 frame->ref[1] = NULL;
200 PREALLOC( frame->i_row_bits, i_lines/16 * sizeof(int) );
201 PREALLOC( frame->f_row_qp, i_lines/16 * sizeof(float) );
202 PREALLOC( frame->f_row_qscale, i_lines/16 * sizeof(float) );
203 if( h->param.analyse.i_me_method >= X264_ME_ESA )
204 PREALLOC( frame->buffer[3], frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) << h->frames.b_have_sub8x8_esa );
205 if( PARAM_INTERLACED )
206 PREALLOC( frame->field, i_mb_count * sizeof(uint8_t) );
207 if( h->param.analyse.b_mb_info )
208 PREALLOC( frame->effective_qp, i_mb_count * sizeof(uint8_t) );
210 else /* fenc frame */
212 if( h->frames.b_have_lowres )
214 int luma_plane_size = align_plane_size( frame->i_stride_lowres * (frame->i_lines[0]/2 + 2*PADV), disalign );
216 PREALLOC( frame->buffer_lowres[0], 4 * luma_plane_size * sizeof(pixel) );
218 for( int j = 0; j <= !!h->param.i_bframe; j++ )
219 for( int i = 0; i <= h->param.i_bframe; i++ )
221 PREALLOC( frame->lowres_mvs[j][i], 2*h->mb.i_mb_count*sizeof(int16_t) );
222 PREALLOC( frame->lowres_mv_costs[j][i], h->mb.i_mb_count*sizeof(int) );
224 PREALLOC( 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 PREALLOC( frame->lowres_costs[j][i], (i_mb_count+3) * sizeof(uint16_t) );
230 if( h->param.rc.i_aq_mode )
232 PREALLOC( frame->f_qp_offset, h->mb.i_mb_count * sizeof(float) );
233 PREALLOC( frame->f_qp_offset_aq, h->mb.i_mb_count * sizeof(float) );
234 if( h->frames.b_have_lowres )
235 PREALLOC( frame->i_inv_qscale_factor, (h->mb.i_mb_count+3) * sizeof(uint16_t) );
239 PREALLOC_END( frame->base );
241 if( i_csp == X264_CSP_NV12 || i_csp == X264_CSP_NV16 )
243 int chroma_padv = i_padv >> (i_csp == X264_CSP_NV12);
244 frame->plane[1] = frame->buffer[1] + frame->i_stride[1] * chroma_padv + PADH;
245 if( PARAM_INTERLACED )
246 frame->plane_fld[1] = frame->buffer_fld[1] + frame->i_stride[1] * chroma_padv + PADH;
249 for( int p = 0; p < luma_plane_count; p++ )
251 int luma_plane_size = align_plane_size( frame->i_stride[p] * (frame->i_lines[p] + 2*i_padv), disalign );
252 if( h->param.analyse.i_subpel_refine && b_fdec )
254 for( int i = 0; i < 4; i++ )
256 frame->filtered[p][i] = frame->buffer[p] + i*luma_plane_size + frame->i_stride[p] * i_padv + PADH;
257 frame->filtered_fld[p][i] = frame->buffer_fld[p] + i*luma_plane_size + frame->i_stride[p] * i_padv + PADH;
259 frame->plane[p] = frame->filtered[p][0];
260 frame->plane_fld[p] = frame->filtered_fld[p][0];
264 frame->filtered[p][0] = frame->plane[p] = frame->buffer[p] + frame->i_stride[p] * i_padv + PADH;
265 frame->filtered_fld[p][0] = frame->plane_fld[p] = frame->buffer_fld[p] + frame->i_stride[p] * i_padv + PADH;
271 M32( frame->mv16x16[0] ) = 0;
274 if( h->param.analyse.i_me_method >= X264_ME_ESA )
275 frame->integral = (uint16_t*)frame->buffer[3] + frame->i_stride[0] * i_padv + PADH;
279 if( h->frames.b_have_lowres )
281 int luma_plane_size = align_plane_size( frame->i_stride_lowres * (frame->i_lines[0]/2 + 2*PADV), disalign );
282 for( int i = 0; i < 4; i++ )
283 frame->lowres[i] = frame->buffer_lowres[0] + (frame->i_stride_lowres * PADV + PADH) + i * luma_plane_size;
285 for( int j = 0; j <= !!h->param.i_bframe; j++ )
286 for( int i = 0; i <= h->param.i_bframe; i++ )
287 memset( frame->lowres_mvs[j][i], 0, 2*h->mb.i_mb_count*sizeof(int16_t) );
289 frame->i_intra_cost = frame->lowres_costs[0][0];
290 memset( frame->i_intra_cost, -1, (i_mb_count+3) * sizeof(uint16_t) );
292 if( h->param.rc.i_aq_mode )
293 /* shouldn't really be initialized, just silences a valgrind false-positive in x264_mbtree_propagate_cost_sse2 */
294 memset( frame->i_inv_qscale_factor, 0, (h->mb.i_mb_count+3) * sizeof(uint16_t) );
298 if( x264_pthread_mutex_init( &frame->mutex, NULL ) )
300 if( x264_pthread_cond_init( &frame->cv, NULL ) )
304 frame->opencl.ocl = h->opencl.ocl;
314 void x264_frame_delete( x264_frame_t *frame )
316 /* Duplicate frames are blank copies of real frames (including pointers),
317 * so freeing those pointers would cause a double free later. */
318 if( !frame->b_duplicate )
320 x264_free( frame->base );
322 if( frame->param && frame->param->param_free )
323 frame->param->param_free( frame->param );
324 if( frame->mb_info_free )
325 frame->mb_info_free( frame->mb_info );
326 if( frame->extra_sei.sei_free )
328 for( int i = 0; i < frame->extra_sei.num_payloads; i++ )
329 frame->extra_sei.sei_free( frame->extra_sei.payloads[i].payload );
330 frame->extra_sei.sei_free( frame->extra_sei.payloads );
332 x264_pthread_mutex_destroy( &frame->mutex );
333 x264_pthread_cond_destroy( &frame->cv );
335 x264_opencl_frame_delete( frame );
341 static int get_plane_ptr( x264_t *h, x264_picture_t *src, uint8_t **pix, int *stride, int plane, int xshift, int yshift )
343 int width = h->param.i_width >> xshift;
344 int height = h->param.i_height >> yshift;
345 *pix = src->img.plane[plane];
346 *stride = src->img.i_stride[plane];
347 if( src->img.i_csp & X264_CSP_VFLIP )
349 *pix += (height-1) * *stride;
352 if( width > abs(*stride) )
354 x264_log( h, X264_LOG_ERROR, "Input picture width (%d) is greater than stride (%d)\n", width, *stride );
360 #define get_plane_ptr(...) do{ if( get_plane_ptr(__VA_ARGS__) < 0 ) return -1; }while(0)
362 int x264_frame_copy_picture( x264_t *h, x264_frame_t *dst, x264_picture_t *src )
364 int i_csp = src->img.i_csp & X264_CSP_MASK;
365 if( dst->i_csp != x264_frame_internal_csp( i_csp ) )
367 x264_log( h, X264_LOG_ERROR, "Invalid input colorspace\n" );
372 if( !(src->img.i_csp & X264_CSP_HIGH_DEPTH) )
374 x264_log( h, X264_LOG_ERROR, "This build of x264 requires high depth input. Rebuild to support 8-bit input.\n" );
378 if( src->img.i_csp & X264_CSP_HIGH_DEPTH )
380 x264_log( h, X264_LOG_ERROR, "This build of x264 requires 8-bit input. Rebuild to support high depth input.\n" );
385 if( BIT_DEPTH != 10 && i_csp == X264_CSP_V210 )
387 x264_log( h, X264_LOG_ERROR, "v210 input is only compatible with bit-depth of 10 bits\n" );
391 if( src->i_type < X264_TYPE_AUTO || src->i_type > X264_TYPE_KEYFRAME )
393 x264_log( h, X264_LOG_WARNING, "forced frame type (%d) at %d is unknown\n", src->i_type, h->frames.i_input );
394 dst->i_forced_type = X264_TYPE_AUTO;
397 dst->i_forced_type = src->i_type;
399 dst->i_type = dst->i_forced_type;
400 dst->i_qpplus1 = src->i_qpplus1;
401 dst->i_pts = dst->i_reordered_pts = src->i_pts;
402 dst->param = src->param;
403 dst->i_pic_struct = src->i_pic_struct;
404 dst->extra_sei = src->extra_sei;
405 dst->opaque = src->opaque;
406 dst->mb_info = h->param.analyse.b_mb_info ? src->prop.mb_info : NULL;
407 dst->mb_info_free = h->param.analyse.b_mb_info ? src->prop.mb_info_free : NULL;
411 if( i_csp == X264_CSP_V210 )
413 stride[0] = src->img.i_stride[0];
414 pix[0] = src->img.plane[0];
416 h->mc.plane_copy_deinterleave_v210( dst->plane[0], dst->i_stride[0],
417 dst->plane[1], dst->i_stride[1],
418 (uint32_t *)pix[0], stride[0]/sizeof(uint32_t), h->param.i_width, h->param.i_height );
420 else if( i_csp >= X264_CSP_BGR )
422 stride[0] = src->img.i_stride[0];
423 pix[0] = src->img.plane[0];
424 if( src->img.i_csp & X264_CSP_VFLIP )
426 pix[0] += (h->param.i_height-1) * stride[0];
427 stride[0] = -stride[0];
429 int b = i_csp==X264_CSP_RGB;
430 h->mc.plane_copy_deinterleave_rgb( dst->plane[1+b], dst->i_stride[1+b],
431 dst->plane[0], dst->i_stride[0],
432 dst->plane[2-b], dst->i_stride[2-b],
433 (pixel*)pix[0], stride[0]/sizeof(pixel), i_csp==X264_CSP_BGRA ? 4 : 3, h->param.i_width, h->param.i_height );
437 int v_shift = CHROMA_V_SHIFT;
438 get_plane_ptr( h, src, &pix[0], &stride[0], 0, 0, 0 );
439 h->mc.plane_copy( dst->plane[0], dst->i_stride[0], (pixel*)pix[0],
440 stride[0]/sizeof(pixel), h->param.i_width, h->param.i_height );
441 if( i_csp == X264_CSP_NV12 || i_csp == X264_CSP_NV16 )
443 get_plane_ptr( h, src, &pix[1], &stride[1], 1, 0, v_shift );
444 h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
445 stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height>>v_shift );
447 else if( i_csp == X264_CSP_NV21 )
449 get_plane_ptr( h, src, &pix[1], &stride[1], 1, 0, v_shift );
450 h->mc.plane_copy_swap( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
451 stride[1]/sizeof(pixel), h->param.i_width>>1, h->param.i_height>>v_shift );
453 else if( i_csp == X264_CSP_I420 || i_csp == X264_CSP_I422 || i_csp == X264_CSP_YV12 || i_csp == X264_CSP_YV16 )
455 int uv_swap = i_csp == X264_CSP_YV12 || i_csp == X264_CSP_YV16;
456 get_plane_ptr( h, src, &pix[1], &stride[1], uv_swap ? 2 : 1, 1, v_shift );
457 get_plane_ptr( h, src, &pix[2], &stride[2], uv_swap ? 1 : 2, 1, v_shift );
458 h->mc.plane_copy_interleave( dst->plane[1], dst->i_stride[1],
459 (pixel*)pix[1], stride[1]/sizeof(pixel),
460 (pixel*)pix[2], stride[2]/sizeof(pixel),
461 h->param.i_width>>1, h->param.i_height>>v_shift );
463 else //if( i_csp == X264_CSP_I444 || i_csp == X264_CSP_YV24 )
465 get_plane_ptr( h, src, &pix[1], &stride[1], i_csp==X264_CSP_I444 ? 1 : 2, 0, 0 );
466 get_plane_ptr( h, src, &pix[2], &stride[2], i_csp==X264_CSP_I444 ? 2 : 1, 0, 0 );
467 h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1],
468 stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height );
469 h->mc.plane_copy( dst->plane[2], dst->i_stride[2], (pixel*)pix[2],
470 stride[2]/sizeof(pixel), h->param.i_width, h->param.i_height );
476 static void ALWAYS_INLINE pixel_memset( pixel *dst, pixel *src, int len, int size )
478 uint8_t *dstp = (uint8_t*)dst;
480 uint32_t v2 = size == 1 ? v1 + (v1 << 8) : M16( src );
481 uint32_t v4 = size <= 2 ? v2 + (v2 << 16) : M32( src );
485 /* Align the input pointer if it isn't already */
486 if( (intptr_t)dstp & (WORD_SIZE - 1) )
488 if( size <= 2 && ((intptr_t)dstp & 3) )
490 if( size == 1 && ((intptr_t)dstp & 1) )
492 if( (intptr_t)dstp & 2 )
498 if( WORD_SIZE == 8 && (intptr_t)dstp & 4 )
508 uint64_t v8 = v4 + ((uint64_t)v4<<32);
509 for( ; i < len - 7; i+=8 )
512 for( ; i < len - 3; i+=4 )
515 /* Finish up the last few bytes */
523 if( size == 1 && i != len )
528 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 )
530 #define PPIXEL(x, y) ( pix + (x) + (y)*i_stride )
531 for( int y = 0; y < i_height; y++ )
534 pixel_memset( PPIXEL(-i_padh, y), PPIXEL(0, y), i_padh>>b_chroma, sizeof(pixel)<<b_chroma );
536 pixel_memset( PPIXEL(i_width, y), PPIXEL(i_width-1-b_chroma, y), i_padh>>b_chroma, sizeof(pixel)<<b_chroma );
540 for( int y = 0; y < i_padv; y++ )
541 memcpy( PPIXEL(-i_padh, -y-1), PPIXEL(-i_padh, 0), (i_width+2*i_padh) * sizeof(pixel) );
544 for( int y = 0; y < i_padv; y++ )
545 memcpy( PPIXEL(-i_padh, i_height+y), PPIXEL(-i_padh, i_height-1), (i_width+2*i_padh) * sizeof(pixel) );
549 void x264_frame_expand_border( x264_t *h, x264_frame_t *frame, int mb_y )
551 int pad_top = mb_y == 0;
552 int pad_bot = mb_y == h->mb.i_mb_height - (1 << SLICE_MBAFF);
553 int b_start = mb_y == h->i_threadslice_start;
554 int b_end = mb_y == h->i_threadslice_end - (1 << SLICE_MBAFF);
555 if( mb_y & SLICE_MBAFF )
557 for( int i = 0; i < frame->i_plane; i++ )
559 int h_shift = i && CHROMA_H_SHIFT;
560 int v_shift = i && CHROMA_V_SHIFT;
561 int stride = frame->i_stride[i];
562 int width = 16*h->mb.i_mb_width;
563 int height = (pad_bot ? 16*(h->mb.i_mb_height - mb_y) >> SLICE_MBAFF : 16) >> v_shift;
565 int padv = PADV >> v_shift;
566 // buffer: 2 chroma, 3 luma (rounded to 4) because deblocking goes beyond the top of the mb
567 if( b_end && !b_start )
568 height += 4 >> (v_shift + SLICE_MBAFF);
570 int starty = 16*mb_y - 4*!b_start;
573 // border samples for each field are extended separately
574 pix = frame->plane_fld[i] + (starty*stride >> v_shift);
575 plane_expand_border( pix, stride*2, width, height, padh, padv, pad_top, pad_bot, h_shift );
576 plane_expand_border( pix+stride, stride*2, width, height, padh, padv, pad_top, pad_bot, h_shift );
578 height = (pad_bot ? 16*(h->mb.i_mb_height - mb_y) : 32) >> v_shift;
579 if( b_end && !b_start )
580 height += 4 >> v_shift;
581 pix = frame->plane[i] + (starty*stride >> v_shift);
582 plane_expand_border( pix, stride, width, height, padh, padv, pad_top, pad_bot, h_shift );
586 pix = frame->plane[i] + (starty*stride >> v_shift);
587 plane_expand_border( pix, stride, width, height, padh, padv, pad_top, pad_bot, h_shift );
592 void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
594 /* during filtering, 8 extra pixels were filtered on each edge,
595 * but up to 3 of the horizontal ones may be wrong.
596 we want to expand border from the last filtered pixel */
598 int width = 16*h->mb.i_mb_width + 8;
599 int height = b_end ? (16*(h->mb.i_mb_height - mb_y) >> SLICE_MBAFF) + 16 : 16;
602 for( int p = 0; p < (CHROMA444 ? 3 : 1); p++ )
603 for( int i = 1; i < 4; i++ )
605 int stride = frame->i_stride[p];
606 // buffer: 8 luma, to match the hpel filter
610 pix = frame->filtered_fld[p][i] + (16*mb_y - 16) * stride - 4;
611 plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end, 0 );
612 plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end, 0 );
615 pix = frame->filtered[p][i] + (16*mb_y - 8) * stride - 4;
616 plane_expand_border( pix, stride, width, height << SLICE_MBAFF, padh, padv, b_start, b_end, 0 );
620 void x264_frame_expand_border_lowres( x264_frame_t *frame )
622 for( int i = 0; i < 4; i++ )
623 plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_width_lowres, frame->i_lines_lowres, PADH, PADV, 1, 1, 0 );
626 void x264_frame_expand_border_chroma( x264_t *h, x264_frame_t *frame, int plane )
628 int v_shift = CHROMA_V_SHIFT;
629 plane_expand_border( frame->plane[plane], frame->i_stride[plane], 16*h->mb.i_mb_width, 16*h->mb.i_mb_height>>v_shift,
630 PADH, PADV>>v_shift, 1, 1, CHROMA_H_SHIFT );
633 void x264_frame_expand_border_mod16( x264_t *h, x264_frame_t *frame )
635 for( int i = 0; i < frame->i_plane; i++ )
637 int i_width = h->param.i_width;
638 int h_shift = i && CHROMA_H_SHIFT;
639 int v_shift = i && CHROMA_V_SHIFT;
640 int i_height = h->param.i_height >> v_shift;
641 int i_padx = (h->mb.i_mb_width * 16 - h->param.i_width);
642 int i_pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> v_shift;
646 for( int y = 0; y < i_height; y++ )
647 pixel_memset( &frame->plane[i][y*frame->i_stride[i] + i_width],
648 &frame->plane[i][y*frame->i_stride[i] + i_width - 1-h_shift],
649 i_padx>>h_shift, sizeof(pixel)<<h_shift );
653 for( int y = i_height; y < i_height + i_pady; y++ )
654 memcpy( &frame->plane[i][y*frame->i_stride[i]],
655 &frame->plane[i][(i_height-(~y&PARAM_INTERLACED)-1)*frame->i_stride[i]],
656 (i_width + i_padx) * sizeof(pixel) );
661 void x264_expand_border_mbpair( x264_t *h, int mb_x, int mb_y )
663 for( int i = 0; i < h->fenc->i_plane; i++ )
665 int v_shift = i && CHROMA_V_SHIFT;
666 int stride = h->fenc->i_stride[i];
667 int height = h->param.i_height >> v_shift;
668 int pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> v_shift;
669 pixel *fenc = h->fenc->plane[i] + 16*mb_x;
670 for( int y = height; y < height + pady; y++ )
671 memcpy( fenc + y*stride, fenc + (height-1)*stride, 16*sizeof(pixel) );
676 void x264_frame_cond_broadcast( x264_frame_t *frame, int i_lines_completed )
678 x264_pthread_mutex_lock( &frame->mutex );
679 frame->i_lines_completed = i_lines_completed;
680 x264_pthread_cond_broadcast( &frame->cv );
681 x264_pthread_mutex_unlock( &frame->mutex );
684 void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
686 x264_pthread_mutex_lock( &frame->mutex );
687 while( frame->i_lines_completed < i_lines_completed )
688 x264_pthread_cond_wait( &frame->cv, &frame->mutex );
689 x264_pthread_mutex_unlock( &frame->mutex );
692 void x264_threadslice_cond_broadcast( x264_t *h, int pass )
694 x264_pthread_mutex_lock( &h->mutex );
695 h->i_threadslice_pass = pass;
697 x264_pthread_cond_broadcast( &h->cv );
698 x264_pthread_mutex_unlock( &h->mutex );
701 void x264_threadslice_cond_wait( x264_t *h, int pass )
703 x264_pthread_mutex_lock( &h->mutex );
704 while( h->i_threadslice_pass < pass )
705 x264_pthread_cond_wait( &h->cv, &h->mutex );
706 x264_pthread_mutex_unlock( &h->mutex );
709 int x264_frame_new_slice( x264_t *h, x264_frame_t *frame )
711 if( h->param.i_slice_count_max )
714 if( h->param.b_sliced_threads )
715 slice_count = x264_pthread_fetch_and_add( &frame->i_slice_count, 1, &frame->mutex );
717 slice_count = frame->i_slice_count++;
718 if( slice_count >= h->param.i_slice_count_max )
726 void x264_frame_push( x264_frame_t **list, x264_frame_t *frame )
729 while( list[i] ) i++;
733 x264_frame_t *x264_frame_pop( x264_frame_t **list )
738 while( list[i+1] ) i++;
744 void x264_frame_unshift( x264_frame_t **list, x264_frame_t *frame )
747 while( list[i] ) i++;
753 x264_frame_t *x264_frame_shift( x264_frame_t **list )
755 x264_frame_t *frame = list[0];
757 for( i = 0; list[i]; i++ )
763 void x264_frame_push_unused( x264_t *h, x264_frame_t *frame )
765 assert( frame->i_reference_count > 0 );
766 frame->i_reference_count--;
767 if( frame->i_reference_count == 0 )
768 x264_frame_push( h->frames.unused[frame->b_fdec], frame );
771 x264_frame_t *x264_frame_pop_unused( x264_t *h, int b_fdec )
774 if( h->frames.unused[b_fdec][0] )
775 frame = x264_frame_pop( h->frames.unused[b_fdec] );
777 frame = x264_frame_new( h, b_fdec );
780 frame->b_last_minigop_bframe = 0;
781 frame->i_reference_count = 1;
782 frame->b_intra_calculated = 0;
783 frame->b_scenecut = 1;
784 frame->b_keyframe = 0;
785 frame->b_corrupt = 0;
786 frame->i_slice_count = h->param.b_sliced_threads ? h->param.i_threads : 1;
788 memset( frame->weight, 0, sizeof(frame->weight) );
789 memset( frame->f_weighted_cost_delta, 0, sizeof(frame->f_weighted_cost_delta) );
794 void x264_frame_push_blank_unused( x264_t *h, x264_frame_t *frame )
796 assert( frame->i_reference_count > 0 );
797 frame->i_reference_count--;
798 if( frame->i_reference_count == 0 )
799 x264_frame_push( h->frames.blank_unused, frame );
802 x264_frame_t *x264_frame_pop_blank_unused( x264_t *h )
805 if( h->frames.blank_unused[0] )
806 frame = x264_frame_pop( h->frames.blank_unused );
808 frame = x264_malloc( sizeof(x264_frame_t) );
811 frame->b_duplicate = 1;
812 frame->i_reference_count = 1;
816 void x264_weight_scale_plane( x264_t *h, pixel *dst, intptr_t i_dst_stride, pixel *src, intptr_t i_src_stride,
817 int i_width, int i_height, x264_weight_t *w )
819 /* Weight horizontal strips of height 16. This was found to be the optimal height
820 * in terms of the cache loads. */
821 while( i_height > 0 )
824 for( x = 0; x < i_width-8; x += 16 )
825 w->weightfn[16>>2]( dst+x, i_dst_stride, src+x, i_src_stride, w, X264_MIN( i_height, 16 ) );
827 w->weightfn[ 8>>2]( dst+x, i_dst_stride, src+x, i_src_stride, w, X264_MIN( i_height, 16 ) );
829 dst += 16 * i_dst_stride;
830 src += 16 * i_src_stride;
834 void x264_frame_delete_list( x264_frame_t **list )
840 x264_frame_delete( list[i++] );
844 int x264_sync_frame_list_init( x264_sync_frame_list_t *slist, int max_size )
848 slist->i_max_size = max_size;
850 CHECKED_MALLOCZERO( slist->list, (max_size+1) * sizeof(x264_frame_t*) );
851 if( x264_pthread_mutex_init( &slist->mutex, NULL ) ||
852 x264_pthread_cond_init( &slist->cv_fill, NULL ) ||
853 x264_pthread_cond_init( &slist->cv_empty, NULL ) )
860 void x264_sync_frame_list_delete( x264_sync_frame_list_t *slist )
862 x264_pthread_mutex_destroy( &slist->mutex );
863 x264_pthread_cond_destroy( &slist->cv_fill );
864 x264_pthread_cond_destroy( &slist->cv_empty );
865 x264_frame_delete_list( slist->list );
868 void x264_sync_frame_list_push( x264_sync_frame_list_t *slist, x264_frame_t *frame )
870 x264_pthread_mutex_lock( &slist->mutex );
871 while( slist->i_size == slist->i_max_size )
872 x264_pthread_cond_wait( &slist->cv_empty, &slist->mutex );
873 slist->list[ slist->i_size++ ] = frame;
874 x264_pthread_mutex_unlock( &slist->mutex );
875 x264_pthread_cond_broadcast( &slist->cv_fill );
878 x264_frame_t *x264_sync_frame_list_pop( x264_sync_frame_list_t *slist )
881 x264_pthread_mutex_lock( &slist->mutex );
882 while( !slist->i_size )
883 x264_pthread_cond_wait( &slist->cv_fill, &slist->mutex );
884 frame = slist->list[ --slist->i_size ];
885 slist->list[ slist->i_size ] = NULL;
886 x264_pthread_cond_broadcast( &slist->cv_empty );
887 x264_pthread_mutex_unlock( &slist->mutex );