X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=common%2Fframe.c;h=c62d94e01d234cc0e088a5a4ee0a2da713428acd;hb=5eccc7d2263ffd63921b95b9e95152ffb6b3645c;hp=b589266f307ce369af56409425ab47edc1f64d54;hpb=4f5d9bcea757f049c1a14dd902c4af76dee231c1;p=x264 diff --git a/common/frame.c b/common/frame.c index b589266f..c62d94e0 100644 --- a/common/frame.c +++ b/common/frame.c @@ -1,7 +1,7 @@ /***************************************************************************** - * frame.c: h264 encoder library + * frame.c: frame handling ***************************************************************************** - * Copyright (C) 2003-2008 x264 project + * Copyright (C) 2003-2011 x264 project * * Authors: Laurent Aimar * Loren Merritt @@ -20,10 +20,28 @@ * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA. + * + * This program is also available under a commercial proprietary license. + * For more information, contact us at licensing@x264.com. *****************************************************************************/ #include "common.h" +static int align_stride( int x, int align, int disalign ) +{ + x = ALIGN( x, align ); + if( !(x&(disalign-1)) ) + x += align; + return x; +} + +static int align_plane_size( int x, int disalign ) +{ + if( !(x&(disalign-1)) ) + x += 128; + return x; +} + x264_frame_t *x264_frame_new( x264_t *h, int b_fdec ) { x264_frame_t *frame; @@ -31,32 +49,28 @@ x264_frame_t *x264_frame_new( x264_t *h, int b_fdec ) int i_mb_count = h->mb.i_mb_count; int i_stride, i_width, i_lines; int i_padv = PADV << h->param.b_interlaced; - int luma_plane_size; - int chroma_plane_size; + int luma_plane_size, chroma_plane_size; int align = h->param.cpu&X264_CPU_CACHELINE_64 ? 64 : h->param.cpu&X264_CPU_CACHELINE_32 ? 32 : 16; + int disalign = h->param.cpu&X264_CPU_ALTIVEC ? 1<<9 : 1<<10; CHECKED_MALLOCZERO( frame, sizeof(x264_frame_t) ); /* allocate frame data (+64 for extra data for me) */ - i_width = h->sps->i_mb_width*16; - i_stride = ALIGN( i_width + 2*PADH, align ); - i_lines = h->sps->i_mb_height*16; + i_width = h->mb.i_mb_width*16; + i_lines = h->mb.i_mb_height*16; + i_stride = align_stride( i_width + 2*PADH, align, disalign ); - frame->i_plane = 3; - for( int i = 0; i < 3; i++ ) + frame->i_plane = 2; + for( int i = 0; i < 2; i++ ) { - frame->i_stride[i] = ALIGN( i_stride >> !!i, align ); - frame->i_width[i] = i_width >> !!i; - frame->i_lines[i] = i_lines >> !!i; + frame->i_width[i] = i_width >> i; + frame->i_lines[i] = i_lines >> i; + frame->i_stride[i] = i_stride; } - luma_plane_size = (frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv)); - chroma_plane_size = (frame->i_stride[1] * (frame->i_lines[1] + 2*i_padv)); - for( int i = 1; i < 3; i++ ) - { - CHECKED_MALLOC( frame->buffer[i], chroma_plane_size * sizeof(pixel) ); - frame->plane[i] = frame->buffer[i] + (frame->i_stride[i] * i_padv + PADH)/2; - } + frame->i_width_lowres = frame->i_width[0]/2; + frame->i_lines_lowres = frame->i_lines[0]/2; + frame->i_stride_lowres = align_stride( frame->i_width_lowres + 2*PADH, align, disalign<<1 ); for( int i = 0; i < h->param.i_bframe + 2; i++ ) for( int j = 0; j < h->param.i_bframe + 2; j++ ) @@ -64,7 +78,7 @@ x264_frame_t *x264_frame_new( x264_t *h, int b_fdec ) frame->i_poc = -1; frame->i_type = X264_TYPE_AUTO; - frame->i_qpplus1 = 0; + frame->i_qpplus1 = X264_QP_AUTO; frame->i_pts = -1; frame->i_frame = -1; frame->i_frame_num = -1; @@ -81,19 +95,40 @@ x264_frame_t *x264_frame_new( x264_t *h, int b_fdec ) frame->orig = frame; + luma_plane_size = align_plane_size( frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv), disalign ); + chroma_plane_size = (frame->i_stride[1] * (frame->i_lines[1] + i_padv)); + + CHECKED_MALLOC( frame->buffer[1], chroma_plane_size * sizeof(pixel) ); + frame->plane[1] = frame->buffer[1] + frame->i_stride[1] * i_padv/2 + PADH; + if( h->param.b_interlaced ) + { + CHECKED_MALLOC( frame->buffer_fld[1], chroma_plane_size * sizeof(pixel) ); + frame->plane_fld[1] = frame->buffer_fld[1] + frame->i_stride[1] * i_padv/2 + PADH; + } + /* all 4 luma planes allocated together, since the cacheline split code * requires them to be in-phase wrt cacheline alignment. */ if( h->param.analyse.i_subpel_refine && b_fdec ) { + /* FIXME: Don't allocate both buffers in non-adaptive MBAFF. */ CHECKED_MALLOC( frame->buffer[0], 4*luma_plane_size * sizeof(pixel) ); + if( h->param.b_interlaced ) + CHECKED_MALLOC( frame->buffer_fld[0], 4*luma_plane_size * sizeof(pixel) ); for( int i = 0; i < 4; i++ ) + { frame->filtered[i] = frame->buffer[0] + i*luma_plane_size + frame->i_stride[0] * i_padv + PADH; + frame->filtered_fld[i] = frame->buffer_fld[0] + i*luma_plane_size + frame->i_stride[0] * i_padv + PADH; + } frame->plane[0] = frame->filtered[0]; + frame->plane_fld[0] = frame->filtered_fld[0]; } else { CHECKED_MALLOC( frame->buffer[0], luma_plane_size * sizeof(pixel) ); + if( h->param.b_interlaced ) + CHECKED_MALLOC( frame->buffer_fld[0], luma_plane_size * sizeof(pixel) ); frame->filtered[0] = frame->plane[0] = frame->buffer[0] + frame->i_stride[0] * i_padv + PADH; + frame->filtered_fld[0] = frame->plane_fld[0] = frame->buffer_fld[0] + frame->i_stride[0] * i_padv + PADH; } frame->b_duplicate = 0; @@ -103,7 +138,9 @@ x264_frame_t *x264_frame_new( x264_t *h, int b_fdec ) CHECKED_MALLOC( frame->mb_type, i_mb_count * sizeof(int8_t)); CHECKED_MALLOC( frame->mb_partition, i_mb_count * sizeof(uint8_t)); CHECKED_MALLOC( frame->mv[0], 2*16 * i_mb_count * sizeof(int16_t) ); - CHECKED_MALLOC( frame->mv16x16, 2*i_mb_count * sizeof(int16_t) ); + CHECKED_MALLOC( frame->mv16x16, 2*(i_mb_count+1) * sizeof(int16_t) ); + M32( frame->mv16x16[0] ) = 0; + frame->mv16x16++; CHECKED_MALLOC( frame->ref[0], 4 * i_mb_count * sizeof(int8_t) ); if( h->param.i_bframe ) { @@ -123,16 +160,14 @@ x264_frame_t *x264_frame_new( x264_t *h, int b_fdec ) frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) << h->frames.b_have_sub8x8_esa ); frame->integral = (uint16_t*)frame->buffer[3] + frame->i_stride[0] * i_padv + PADH; } + if( h->param.b_interlaced ) + CHECKED_MALLOC( frame->field, i_mb_count * sizeof(uint8_t) ); } else /* fenc frame */ { if( h->frames.b_have_lowres ) { - frame->i_width_lowres = frame->i_width[0]/2; - frame->i_stride_lowres = ALIGN( frame->i_width_lowres + 2*PADH, align ); - frame->i_lines_lowres = frame->i_lines[0]/2; - - luma_plane_size = frame->i_stride_lowres * (frame->i_lines[0]/2 + 2*PADV); + luma_plane_size = align_plane_size( frame->i_stride_lowres * (frame->i_lines[0]/2 + 2*PADV), disalign ); CHECKED_MALLOC( frame->buffer_lowres[0], 4 * luma_plane_size * sizeof(pixel) ); for( int i = 0; i < 4; i++ ) @@ -180,7 +215,10 @@ void x264_frame_delete( x264_frame_t *frame ) if( !frame->b_duplicate ) { for( int i = 0; i < 4; i++ ) + { x264_free( frame->buffer[i] ); + x264_free( frame->buffer_fld[i] ); + } for( int i = 0; i < 4; i++ ) x264_free( frame->buffer_lowres[i] ); for( int i = 0; i < X264_BFRAME_MAX+2; i++ ) @@ -201,11 +239,13 @@ void x264_frame_delete( x264_frame_t *frame ) x264_free( frame->i_inv_qscale_factor ); x264_free( frame->i_row_bits ); x264_free( frame->f_row_qp ); + x264_free( frame->field ); x264_free( frame->mb_type ); x264_free( frame->mb_partition ); x264_free( frame->mv[0] ); x264_free( frame->mv[1] ); - x264_free( frame->mv16x16 ); + if( frame->mv16x16 ) + x264_free( frame->mv16x16-1 ); x264_free( frame->ref[0] ); x264_free( frame->ref[1] ); x264_pthread_mutex_destroy( &frame->mutex ); @@ -214,58 +254,108 @@ void x264_frame_delete( x264_frame_t *frame ) x264_free( frame ); } +static int get_plane_ptr( x264_t *h, x264_picture_t *src, uint8_t **pix, int *stride, int plane, int xshift, int yshift ) +{ + int width = h->param.i_width >> xshift; + int height = h->param.i_height >> yshift; + *pix = src->img.plane[plane]; + *stride = src->img.i_stride[plane]; + if( src->img.i_csp & X264_CSP_VFLIP ) + { + *pix += (height-1) * *stride; + *stride = -*stride; + } + if( width > abs(*stride) ) + { + x264_log( h, X264_LOG_ERROR, "Input picture width (%d) is greater than stride (%d)\n", width, *stride ); + return -1; + } + return 0; +} + +#define get_plane_ptr(...) do{ if( get_plane_ptr(__VA_ARGS__) < 0 ) return -1; }while(0) + int x264_frame_copy_picture( x264_t *h, x264_frame_t *dst, x264_picture_t *src ) { int i_csp = src->img.i_csp & X264_CSP_MASK; - if( i_csp != X264_CSP_I420 && i_csp != X264_CSP_YV12 ) + if( i_csp <= X264_CSP_NONE || i_csp >= X264_CSP_MAX ) { x264_log( h, X264_LOG_ERROR, "Invalid input colorspace\n" ); return -1; } +#if HIGH_BIT_DEPTH + if( !(src->img.i_csp & X264_CSP_HIGH_DEPTH) ) + { + x264_log( h, X264_LOG_ERROR, "This build of x264 requires high depth input. Rebuild to support 8-bit input.\n" ); + return -1; + } +#else + if( src->img.i_csp & X264_CSP_HIGH_DEPTH ) + { + x264_log( h, X264_LOG_ERROR, "This build of x264 requires 8-bit input. Rebuild to support high depth input.\n" ); + return -1; + } +#endif + dst->i_type = src->i_type; dst->i_qpplus1 = src->i_qpplus1; dst->i_pts = dst->i_reordered_pts = src->i_pts; dst->param = src->param; dst->i_pic_struct = src->i_pic_struct; - - for( int i = 0; i < 3; i++ ) + dst->extra_sei = src->extra_sei; + + uint8_t *pix[3]; + int stride[3]; + get_plane_ptr( h, src, &pix[0], &stride[0], 0, 0, 0 ); + h->mc.plane_copy( dst->plane[0], dst->i_stride[0], (pixel*)pix[0], + stride[0]/sizeof(pixel), h->param.i_width, h->param.i_height ); + if( i_csp == X264_CSP_NV12 ) { - int s = (i_csp == X264_CSP_YV12 && i) ? i^3 : i; - uint8_t *plane = src->img.plane[s]; - int stride = src->img.i_stride[s]; - int width = h->param.i_width >> !!i; - int height = h->param.i_height >> !!i; - if( src->img.i_csp & X264_CSP_VFLIP ) - { - plane += (height-1)*stride; - stride = -stride; - } - if( width > abs(stride) ) - { - x264_log( h, X264_LOG_ERROR, "Input picture width is greater than stride\n" ); - return -1; - } - h->mc.plane_copy( dst->plane[i], dst->i_stride[i], plane, stride, width, height ); + get_plane_ptr( h, src, &pix[1], &stride[1], 1, 0, 1 ); + h->mc.plane_copy( dst->plane[1], dst->i_stride[1], (pixel*)pix[1], + stride[1]/sizeof(pixel), h->param.i_width, h->param.i_height>>1 ); + } + else + { + get_plane_ptr( h, src, &pix[1], &stride[1], i_csp==X264_CSP_I420 ? 1 : 2, 1, 1 ); + get_plane_ptr( h, src, &pix[2], &stride[2], i_csp==X264_CSP_I420 ? 2 : 1, 1, 1 ); + h->mc.plane_copy_interleave( dst->plane[1], dst->i_stride[1], + (pixel*)pix[1], stride[1]/sizeof(pixel), + (pixel*)pix[2], stride[2]/sizeof(pixel), + h->param.i_width>>1, h->param.i_height>>1 ); } return 0; } -static void ALWAYS_INLINE pixel_memset( pixel *dst, int value, int size ) +static void ALWAYS_INLINE pixel_memset( pixel *dst, pixel *src, int len, int size ) { - for( int i = 0; i < size; i++ ) - dst[i] = value; + uint8_t *dstp = (uint8_t*)dst; + if( size == 1 ) + memset(dst, *src, len); + else if( size == 2 ) + { + int v = M16( src ); + for( int i = 0; i < len; i++ ) + M16( dstp+i*2 ) = v; + } + else if( size == 4 ) + { + int v = M32( src ); + for( int i = 0; i < len; i++ ) + M32( dstp+i*4 ) = v; + } } -static void 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 ) +static void 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 ) { #define PPIXEL(x, y) ( pix + (x) + (y)*i_stride ) for( int y = 0; y < i_height; y++ ) { /* left band */ - pixel_memset( PPIXEL(-i_padh, y), PPIXEL(0, y)[0], i_padh ); + pixel_memset( PPIXEL(-i_padh, y), PPIXEL(0, y), i_padh>>b_chroma, sizeof(pixel)<>b_chroma, sizeof(pixel)<i_plane; i++ ) { int stride = frame->i_stride[i]; - int width = 16*h->sps->i_mb_width >> !!i; - int height = (b_end ? 16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff : 16) >> !!i; - int padh = PADH >> !!i; + int width = 16*h->sps->i_mb_width; + int height = (b_end ? 16*(h->mb.i_mb_height - mb_y) >> h->sh.b_mbaff : 16) >> !!i; + int padh = PADH; int padv = PADV >> !!i; // buffer: 2 chroma, 3 luma (rounded to 4) because deblocking goes beyond the top of the mb - pixel *pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> !!i); if( b_end && !b_start ) height += 4 >> (!!i + h->sh.b_mbaff); + pixel *pix; if( h->sh.b_mbaff ) { - plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end ); - plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end ); + // border samples for each field are extended separately + pix = frame->plane_fld[i] + X264_MAX(0, (16*mb_y-4)*stride >> !!i); + plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end, i ); + plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end, i ); + + height = (b_end ? 16*(h->mb.i_mb_height - mb_y) : 32) >> !!i; + if( b_end && !b_start ) + height += 4 >> (!!i); + pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> !!i); + plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end, i ); } else { - plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end ); + pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> !!i); + plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end, i ); } } } @@ -313,47 +412,47 @@ void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y we want to expand border from the last filtered pixel */ int b_start = !mb_y; int stride = frame->i_stride[0]; - int width = 16*h->sps->i_mb_width + 8; - int height = b_end ? (16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff) + 16 : 16; + int width = 16*h->mb.i_mb_width + 8; + int height = b_end ? (16*(h->mb.i_mb_height - mb_y) >> h->sh.b_mbaff) + 16 : 16; int padh = PADH - 4; int padv = PADV - 8; for( int i = 1; i < 4; i++ ) { // buffer: 8 luma, to match the hpel filter - pixel *pix = frame->filtered[i] + (16*mb_y - (8 << h->sh.b_mbaff)) * stride - 4; + pixel *pix; if( h->sh.b_mbaff ) { - plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end ); - plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end ); + pix = frame->filtered_fld[i] + (16*mb_y - 16) * stride - 4; + plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end, 0 ); + plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end, 0 ); } - else - plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end ); + + pix = frame->filtered[i] + (16*mb_y - 8) * stride - 4; + plane_expand_border( pix, stride, width, height << h->sh.b_mbaff, padh, padv, b_start, b_end, 0 ); } } void x264_frame_expand_border_lowres( x264_frame_t *frame ) { for( int i = 0; i < 4; i++ ) - plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_width_lowres, frame->i_lines_lowres, PADH, PADV, 1, 1 ); + plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_width_lowres, frame->i_lines_lowres, PADH, PADV, 1, 1, 0 ); } void x264_frame_expand_border_mod16( x264_t *h, x264_frame_t *frame ) { for( int i = 0; i < frame->i_plane; i++ ) { - int i_subsample = i ? 1 : 0; - int i_width = h->param.i_width >> i_subsample; - int i_height = h->param.i_height >> i_subsample; - int i_padx = (h->sps->i_mb_width * 16 - h->param.i_width) >> i_subsample; - int i_pady = (h->sps->i_mb_height * 16 - h->param.i_height) >> i_subsample; + int i_width = h->param.i_width; + int i_height = h->param.i_height >> !!i; + int i_padx = (h->mb.i_mb_width * 16 - h->param.i_width); + int i_pady = (h->mb.i_mb_height * 16 - h->param.i_height) >> !!i; if( i_padx ) { for( int y = 0; y < i_height; y++ ) - { - pixel value = frame->plane[i][y*frame->i_stride[i] + i_width - 1]; - pixel_memset( &frame->plane[i][y*frame->i_stride[i] + i_width], value, i_padx ); - } + pixel_memset( &frame->plane[i][y*frame->i_stride[i] + i_width], + &frame->plane[i][y*frame->i_stride[i] + i_width - 1-i], + i_padx>>i, sizeof(pixel)<b_intra_calculated = 0; frame->b_scenecut = 1; frame->b_keyframe = 0; + frame->b_corrupt = 0; memset( frame->weight, 0, sizeof(frame->weight) ); memset( frame->f_weighted_cost_delta, 0, sizeof(frame->f_weighted_cost_delta) ); @@ -517,7 +617,7 @@ void x264_frame_delete_list( x264_frame_t **list ) x264_free( list ); } -int x264_synch_frame_list_init( x264_synch_frame_list_t *slist, int max_size ) +int x264_sync_frame_list_init( x264_sync_frame_list_t *slist, int max_size ) { if( max_size < 0 ) return -1; @@ -533,7 +633,7 @@ fail: return -1; } -void x264_synch_frame_list_delete( x264_synch_frame_list_t *slist ) +void x264_sync_frame_list_delete( x264_sync_frame_list_t *slist ) { x264_pthread_mutex_destroy( &slist->mutex ); x264_pthread_cond_destroy( &slist->cv_fill ); @@ -541,7 +641,7 @@ void x264_synch_frame_list_delete( x264_synch_frame_list_t *slist ) x264_frame_delete_list( slist->list ); } -void x264_synch_frame_list_push( x264_synch_frame_list_t *slist, x264_frame_t *frame ) +void x264_sync_frame_list_push( x264_sync_frame_list_t *slist, x264_frame_t *frame ) { x264_pthread_mutex_lock( &slist->mutex ); while( slist->i_size == slist->i_max_size ) @@ -550,3 +650,16 @@ void x264_synch_frame_list_push( x264_synch_frame_list_t *slist, x264_frame_t *f x264_pthread_mutex_unlock( &slist->mutex ); x264_pthread_cond_broadcast( &slist->cv_fill ); } + +x264_frame_t *x264_sync_frame_list_pop( x264_sync_frame_list_t *slist ) +{ + x264_frame_t *frame; + x264_pthread_mutex_lock( &slist->mutex ); + while( !slist->i_size ) + x264_pthread_cond_wait( &slist->cv_fill, &slist->mutex ); + frame = slist->list[ --slist->i_size ]; + slist->list[ slist->i_size ] = NULL; + x264_pthread_cond_broadcast( &slist->cv_empty ); + x264_pthread_mutex_unlock( &slist->mutex ); + return frame; +}