* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
*****************************************************************************/
-#include <stdio.h>
-#include <string.h>
-
#include "common.h"
#define PADH 32
frame->i_pts = -1;
frame->i_frame = -1;
frame->i_frame_num = -1;
+ frame->i_lines_completed = -1;
CHECKED_MALLOC( frame->mb_type, i_mb_count * sizeof(int8_t));
CHECKED_MALLOC( frame->mv[0], 2*16 * i_mb_count * sizeof(int16_t) );
for( j = 0; j < h->param.i_bframe + 2; j++ )
CHECKED_MALLOC( frame->i_row_satds[i][j], i_lines/16 * sizeof(int) );
+ x264_pthread_mutex_init( &frame->mutex, NULL );
+ x264_pthread_cond_init( &frame->cv, NULL );
+
return frame;
fail:
x264_free( frame->mv[1] );
x264_free( frame->ref[0] );
x264_free( frame->ref[1] );
+ x264_pthread_mutex_destroy( &frame->mutex );
+ x264_pthread_cond_destroy( &frame->cv );
x264_free( frame );
}
void 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;
dst->i_type = src->i_type;
dst->i_qpplus1 = src->i_qpplus1;
dst->i_pts = src->i_pts;
- switch( src->img.i_csp & X264_CSP_MASK )
- {
- case X264_CSP_I420:
- h->csp.i420( dst, &src->img, h->param.i_width, h->param.i_height );
- break;
- case X264_CSP_YV12:
- h->csp.yv12( dst, &src->img, h->param.i_width, h->param.i_height );
- break;
- case X264_CSP_I422:
- h->csp.i422( dst, &src->img, h->param.i_width, h->param.i_height );
- break;
- case X264_CSP_I444:
- h->csp.i444( dst, &src->img, h->param.i_width, h->param.i_height );
- break;
- case X264_CSP_YUYV:
- h->csp.yuyv( dst, &src->img, h->param.i_width, h->param.i_height );
- break;
- case X264_CSP_RGB:
- h->csp.rgb( dst, &src->img, h->param.i_width, h->param.i_height );
- break;
- case X264_CSP_BGR:
- h->csp.bgr( dst, &src->img, h->param.i_width, h->param.i_height );
- break;
- case X264_CSP_BGRA:
- h->csp.bgra( dst, &src->img, h->param.i_width, h->param.i_height );
- break;
-
- default:
- x264_log( h, X264_LOG_ERROR, "Arg invalid CSP\n" );
- break;
- }
+ if( i_csp <= X264_CSP_NONE || i_csp >= X264_CSP_MAX )
+ x264_log( h, X264_LOG_ERROR, "Arg invalid CSP\n" );
+ else
+ h->csp.convert[i_csp]( &h->mc, dst, &src->img, h->param.i_width, h->param.i_height );
}
-static void plane_expand_border( uint8_t *pix, int i_stride, int i_width, int i_height, int i_padh, int i_padv )
+static void plane_expand_border( uint8_t *pix, int i_stride, int i_width, int i_height, int i_padh, int i_padv, int b_pad_top, int b_pad_bottom )
{
#define PPIXEL(x, y) ( pix + (x) + (y)*i_stride )
int y;
memset( PPIXEL(i_width, y), PPIXEL(i_width-1, y)[0], i_padh );
}
/* upper band */
+ if( b_pad_top )
for( y = 0; y < i_padv; y++ )
memcpy( PPIXEL(-i_padh, -y-1), PPIXEL(-i_padh, 0), i_width+2*i_padh );
/* lower band */
+ if( b_pad_bottom )
for( y = 0; y < i_padv; y++ )
memcpy( PPIXEL(-i_padh, i_height+y), PPIXEL(-i_padh, i_height-1), i_width+2*i_padh );
#undef PPIXEL
}
-void x264_frame_expand_border( x264_t *h, x264_frame_t *frame )
+void x264_frame_expand_border( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
{
int i;
+ int b_start = !mb_y;
+ if( mb_y & h->sh.b_mbaff )
+ return;
for( i = 0; i < frame->i_plane; i++ )
{
int stride = frame->i_stride[i];
int width = 16*h->sps->i_mb_width >> !!i;
- int height = 16*h->sps->i_mb_height >> !!i;
+ int height = (b_end ? 16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff : 16) >> !!i;
int padh = PADH >> !!i;
int padv = PADV >> !!i;
- if( h->param.b_interlaced )
+ // buffer: 2 chroma, 3 luma (rounded to 4) because deblocking goes beyond the top of the mb
+ uint8_t *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);
+ if( h->sh.b_mbaff )
{
- plane_expand_border( frame->plane[i], stride*2, width, height>>1, padh, padv );
- plane_expand_border( frame->plane[i]+stride, stride*2, width, height>>1, padh, padv );
+ 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 );
}
else
{
- plane_expand_border( frame->plane[i], stride, width, height, padh, padv );
+ plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end );
}
}
}
-void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame )
+void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
{
/* during filtering, 8 extra pixels were filtered on each edge.
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;
- int height = 16*h->sps->i_mb_height;
+ int width = 16*h->sps->i_mb_width + 16;
+ int height = b_end ? (16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff) + 16 : 16;
int padh = PADH - 8;
int padv = PADV - 8;
int i;
for( i = 1; i < 4; i++ )
{
- if( h->param.b_interlaced )
+ // buffer: 8 luma, to match the hpel filter
+ uint8_t *pix = frame->filtered[i] + (16*mb_y - (8 << h->sh.b_mbaff)) * stride - 8;
+ if( h->sh.b_mbaff )
{
- plane_expand_border( frame->filtered[i] - 16*stride - 8, stride*2, width+16, (height>>1)+16, padh, padv );
- plane_expand_border( frame->filtered[i] - 15*stride - 8, stride*2, width+16, (height>>1)+16, padh, padv );
+ 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 );
}
else
{
- plane_expand_border( frame->filtered[i] - 8*stride - 8, stride, width+16, height+16, padh, padv );
+ plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end );
}
}
}
{
int i;
for( i = 0; i < 4; i++ )
- plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_stride_lowres - 2*PADH, frame->i_lines_lowres, PADH, PADV );
+ plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_stride_lowres - 2*PADH, frame->i_lines_lowres, PADH, PADV, 1, 1 );
}
void x264_frame_expand_border_mod16( x264_t *h, x264_frame_t *frame )
}
+/* cavlc + 8x8 transform stores nnz per 16 coeffs for the purpose of
+ * entropy coding, but per 64 coeffs for the purpose of deblocking */
+void munge_cavlc_nnz_row( x264_t *h, int mb_y, uint8_t (*buf)[16] )
+{
+ uint32_t (*src)[6] = (uint32_t(*)[6])h->mb.non_zero_count + mb_y * h->sps->i_mb_width;
+ int8_t *transform = h->mb.mb_transform_size + mb_y * h->sps->i_mb_width;
+ int x;
+ for( x=0; x<h->sps->i_mb_width; x++ )
+ {
+ memcpy( buf+x, src+x, 16 );
+ if( transform[x] )
+ {
+ if( src[x][0] ) src[x][0] = 0x01010101;
+ if( src[x][1] ) src[x][1] = 0x01010101;
+ if( src[x][2] ) src[x][2] = 0x01010101;
+ if( src[x][3] ) src[x][3] = 0x01010101;
+ }
+ }
+}
+
+static void restore_cavlc_nnz_row( x264_t *h, int mb_y, uint8_t (*buf)[16] )
+{
+ uint8_t (*dst)[24] = h->mb.non_zero_count + mb_y * h->sps->i_mb_width;
+ int x;
+ for( x=0; x<h->sps->i_mb_width; x++ )
+ memcpy( dst+x, buf+x, 16 );
+}
+
+static void munge_cavlc_nnz( x264_t *h, int mb_y, uint8_t (*buf)[16], void (*func)(x264_t*, int, uint8_t (*)[16]) )
+{
+ func( h, mb_y, buf );
+ if( mb_y > 0 )
+ func( h, mb_y-1, buf + h->sps->i_mb_width );
+ if( h->sh.b_mbaff )
+ {
+ func( h, mb_y+1, buf + h->sps->i_mb_width * 2 );
+ if( mb_y > 0 )
+ func( h, mb_y-2, buf + h->sps->i_mb_width * 3 );
+ }
+}
+
+
/* Deblocking filter */
static const int i_alpha_table[52] =
}
}
-void x264_frame_deblocking_filter( x264_t *h, int i_slice_type )
+void x264_frame_deblock_row( x264_t *h, int mb_y )
{
const int s8x8 = 2 * h->mb.i_mb_stride;
const int s4x4 = 4 * h->mb.i_mb_stride;
- const int b_interlaced = h->param.b_interlaced;
+ const int b_interlaced = h->sh.b_mbaff;
const int mvy_limit = 4 >> b_interlaced;
- int mb_y, mb_x;
+ int mb_x;
int i_stride2[3] = { h->fdec->i_stride[0] << b_interlaced,
h->fdec->i_stride[1] << b_interlaced,
h->fdec->i_stride[2] << b_interlaced };
- for( mb_y = 0, mb_x = 0; mb_y < h->sps->i_mb_height; )
+ if( !h->pps->b_cabac && h->pps->b_transform_8x8_mode )
+ munge_cavlc_nnz( h, mb_y, h->mb.nnz_backup, munge_cavlc_nnz_row );
+
+ for( mb_x = 0; mb_x < h->sps->i_mb_width; )
{
const int mb_xy = mb_y * h->mb.i_mb_stride + mb_x;
const int mb_8x8 = 2 * s8x8 * mb_y + 2 * mb_x;
i_pix_y[2] -= 7*h->fdec->i_stride[2];
}
- /* cavlc + 8x8 transform stores nnz per 16 coeffs for the purpose of
- * entropy coding, but per 64 coeffs for the purpose of deblocking */
- if( !h->param.b_cabac && b_8x8_transform )
- {
- uint32_t *nnz = (uint32_t*)h->mb.non_zero_count[mb_xy];
- if( nnz[0] ) nnz[0] = 0x01010101;
- if( nnz[1] ) nnz[1] = 0x01010101;
- if( nnz[2] ) nnz[2] = 0x01010101;
- if( nnz[3] ) nnz[3] = 0x01010101;
- }
+ x264_prefetch_fenc( h, h->fdec, mb_x, mb_y );
/* i_dir == 0 -> vertical edge
* i_dir == 1 -> horizontal edge */
bS[i] = 0;
- for( l = 0; l < 1 + (i_slice_type == SLICE_TYPE_B); l++ )
+ for( l = 0; l < 1 + (h->sh.i_type == SLICE_TYPE_B); l++ )
{
if( h->mb.ref[l][i8p] != h->mb.ref[l][i8q] ||
abs( h->mb.mv[l][i4p][0] - h->mb.mv[l][i4q][0] ) >= 4 ||
/* next mb */
if( !b_interlaced || (mb_y&1) )
mb_x++;
- if( mb_x >= h->sps->i_mb_width )
- {
- mb_x = 0;
- mb_y++;
- }
- else
- mb_y ^= b_interlaced;
+ mb_y ^= b_interlaced;
}
+
+ if( !h->pps->b_cabac && h->pps->b_transform_8x8_mode )
+ munge_cavlc_nnz( h, mb_y, h->mb.nnz_backup, restore_cavlc_nnz_row );
}
-#ifdef HAVE_MMXEXT
+void x264_frame_deblock( x264_t *h )
+{
+ int mb_y;
+ for( mb_y = 0; mb_y < h->sps->i_mb_height; mb_y += 1 + h->sh.b_mbaff )
+ x264_frame_deblock_row( h, mb_y );
+}
+
+#ifdef HAVE_MMX
void x264_deblock_v_chroma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_h_chroma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_v_chroma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );
#endif
#endif
+#ifdef ARCH_PPC
+void x264_deblock_v_luma_altivec( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
+void x264_deblock_h_luma_altivec( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
+#endif // ARCH_PPC
+
void x264_deblock_init( int cpu, x264_deblock_function_t *pf )
{
pf->deblock_v_luma = deblock_v_luma_c;
pf->deblock_v_chroma_intra = deblock_v_chroma_intra_c;
pf->deblock_h_chroma_intra = deblock_h_chroma_intra_c;
-#ifdef HAVE_MMXEXT
+#ifdef HAVE_MMX
if( cpu&X264_CPU_MMXEXT )
{
pf->deblock_v_chroma = x264_deblock_v_chroma_mmxext;
#endif
}
#endif
+
+#ifdef ARCH_PPC
+ if( cpu&X264_CPU_ALTIVEC )
+ {
+ pf->deblock_v_luma = x264_deblock_v_luma_altivec;
+ pf->deblock_h_luma = x264_deblock_h_luma_altivec;
+ }
+#endif // ARCH_PPC
+}
+
+
+/* threading */
+
+#ifdef HAVE_PTHREAD
+void x264_frame_cond_broadcast( x264_frame_t *frame, int i_lines_completed )
+{
+ x264_pthread_mutex_lock( &frame->mutex );
+ frame->i_lines_completed = i_lines_completed;
+ x264_pthread_cond_broadcast( &frame->cv );
+ x264_pthread_mutex_unlock( &frame->mutex );
+}
+
+void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
+{
+ x264_pthread_mutex_lock( &frame->mutex );
+ while( frame->i_lines_completed < i_lines_completed )
+ x264_pthread_cond_wait( &frame->cv, &frame->mutex );
+ x264_pthread_mutex_unlock( &frame->mutex );
+}
+
+#else
+void x264_frame_cond_broadcast( x264_frame_t *frame, int i_lines_completed )
+{}
+void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
+{}
+#endif
+
+
+/* list operators */
+
+void x264_frame_push( x264_frame_t **list, x264_frame_t *frame )
+{
+ int i = 0;
+ while( list[i] ) i++;
+ list[i] = frame;
+}
+
+x264_frame_t *x264_frame_pop( x264_frame_t **list )
+{
+ x264_frame_t *frame;
+ int i = 0;
+ assert( list[0] );
+ while( list[i+1] ) i++;
+ frame = list[i];
+ list[i] = NULL;
+ return frame;
+}
+
+void x264_frame_unshift( x264_frame_t **list, x264_frame_t *frame )
+{
+ int i = 0;
+ while( list[i] ) i++;
+ while( i-- )
+ list[i+1] = list[i];
+ list[0] = frame;
}
+x264_frame_t *x264_frame_shift( x264_frame_t **list )
+{
+ x264_frame_t *frame = list[0];
+ int i;
+ for( i = 0; list[i]; i++ )
+ list[i] = list[i+1];
+ assert(frame);
+ return frame;
+}
+
+void x264_frame_push_unused( x264_t *h, x264_frame_t *frame )
+{
+ assert( frame->i_reference_count > 0 );
+ frame->i_reference_count--;
+ if( frame->i_reference_count == 0 )
+ x264_frame_push( h->frames.unused, frame );
+ assert( h->frames.unused[ sizeof(h->frames.unused) / sizeof(*h->frames.unused) - 1 ] == NULL );
+}
+
+x264_frame_t *x264_frame_pop_unused( x264_t *h )
+{
+ x264_frame_t *frame;
+ if( h->frames.unused[0] )
+ frame = x264_frame_pop( h->frames.unused );
+ else
+ frame = x264_frame_new( h );
+ assert( frame->i_reference_count == 0 );
+ frame->i_reference_count = 1;
+ return frame;
+}
+
+void x264_frame_sort( x264_frame_t **list, int b_dts )
+{
+ int i, b_ok;
+ do {
+ b_ok = 1;
+ for( i = 0; list[i+1]; i++ )
+ {
+ int dtype = list[i]->i_type - list[i+1]->i_type;
+ int dtime = list[i]->i_frame - list[i+1]->i_frame;
+ int swap = b_dts ? dtype > 0 || ( dtype == 0 && dtime > 0 )
+ : dtime > 0;
+ if( swap )
+ {
+ XCHG( x264_frame_t*, list[i], list[i+1] );
+ b_ok = 0;
+ }
+ }
+ } while( !b_ok );
+}