frame.c.svn-base

X264最新源码

                int bS[4];  /* filtering strength */

                if( b_8x8_transform && (i_edge&1) )
                    continue;

                mbn_xy  = i_edge > 0 ? mb_xy  : ( i_dir == 0 ? mb_xy  - 1 : mb_xy - h->mb.i_mb_stride );
                mbn_8x8 = i_edge > 0 ? mb_8x8 : ( i_dir == 0 ? mb_8x8 - 2 : mb_8x8 - 2 * s8x8 );
                mbn_4x4 = i_edge > 0 ? mb_4x4 : ( i_dir == 0 ? mb_4x4 - 4 : mb_4x4 - 4 * s4x4 );

                if( b_interlaced && i_edge == 0 && i_dir == 1 )
                {
                    mbn_xy -= h->mb.i_mb_stride;
                    mbn_8x8 -= 2 * s8x8;
                    mbn_4x4 -= 4 * s4x4;
                }

                /* *** Get bS for each 4px for the current edge *** */
                if( IS_INTRA( h->mb.type[mb_xy] ) || IS_INTRA( h->mb.type[mbn_xy] ) )
                {
                    bS[0] = bS[1] = bS[2] = bS[3] = ( i_edge == 0 && !(b_interlaced && i_dir) ? 4 : 3 );
                }
                else
                {
                    int i;
                    for( i = 0; i < 4; i++ )
                    {
                        int x  = i_dir == 0 ? i_edge : i;
                        int y  = i_dir == 0 ? i      : i_edge;
                        int xn = (x - (i_dir == 0 ? 1 : 0 ))&0x03;
                        int yn = (y - (i_dir == 0 ? 0 : 1 ))&0x03;

                        if( h->mb.non_zero_count[mb_xy][block_idx_xy[x][y]] != 0 ||
                            h->mb.non_zero_count[mbn_xy][block_idx_xy[xn][yn]] != 0 )
                        {
                            bS[i] = 2;
                        }
                        else
                        {
                            /* FIXME: A given frame may occupy more than one position in
                             * the reference list. So we should compare the frame numbers,
                             * not the indices in the ref list.
                             * No harm yet, as we don't generate that case.*/

                            int i8p= mb_8x8+(x/2)+(y/2)*s8x8;
                            int i8q= mbn_8x8+(xn/2)+(yn/2)*s8x8;
                            int i4p= mb_4x4+x+y*s4x4;
                            int i4q= mbn_4x4+xn+yn*s4x4;
                            int l;

                            bS[i] = 0;

                            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 ||
                                    abs( h->mb.mv[l][i4p][1] - h->mb.mv[l][i4q][1] ) >= mvy_limit )
                                {
                                    bS[i] = 1;
                                    break;
                                }
                            }
                        }
                    }
                }

                /* *** filter *** */
                /* Y plane */
                i_qp = h->mb.qp[mb_xy];
                i_qpn= h->mb.qp[mbn_xy];

                if( i_dir == 0 )
                {
                    /* vertical edge */
                    deblock_edge( h, &h->fdec->plane[0][i_pix_y[0] + 4*i_edge],
                                  i_stride2[0], bS, (i_qp+i_qpn+1) >> 1, 0,
                                  h->loopf.deblock_h_luma, h->loopf.deblock_h_luma_intra );
                    if( !(i_edge & 1) )
                    {
                        /* U/V planes */
                        int i_qpc = ( i_chroma_qp_table[x264_clip3( i_qp + h->pps->i_chroma_qp_index_offset, 0, 51 )] +
                                      i_chroma_qp_table[x264_clip3( i_qpn + h->pps->i_chroma_qp_index_offset, 0, 51 )] + 1 ) >> 1;
                        deblock_edge( h, &h->fdec->plane[1][i_pix_y[1] + 2*i_edge],
                                      i_stride2[1], bS, i_qpc, 1,
                                      h->loopf.deblock_h_chroma, h->loopf.deblock_h_chroma_intra );
                        deblock_edge( h, &h->fdec->plane[2][i_pix_y[2] + 2*i_edge],
                                      i_stride2[2], bS, i_qpc, 1,
                                      h->loopf.deblock_h_chroma, h->loopf.deblock_h_chroma_intra );
                    }
                }
                else
                {
                    /* horizontal edge */
                    deblock_edge( h, &h->fdec->plane[0][i_pix_y[0] + 4*i_edge*i_stride2[0]],
                                  i_stride2[0], bS, (i_qp+i_qpn+1) >> 1, 0,
                                  h->loopf.deblock_v_luma, h->loopf.deblock_v_luma_intra );
                    /* U/V planes */
                    if( !(i_edge & 1) )
                    {
                        int i_qpc = ( i_chroma_qp_table[x264_clip3( i_qp + h->pps->i_chroma_qp_index_offset, 0, 51 )] +
                                      i_chroma_qp_table[x264_clip3( i_qpn + h->pps->i_chroma_qp_index_offset, 0, 51 )] + 1 ) >> 1;
                        deblock_edge( h, &h->fdec->plane[1][i_pix_y[1] + 2*i_edge*i_stride2[1]],
                                      i_stride2[1], bS, i_qpc, 1,
                                      h->loopf.deblock_v_chroma, h->loopf.deblock_v_chroma_intra );
                        deblock_edge( h, &h->fdec->plane[2][i_pix_y[2] + 2*i_edge*i_stride2[2]],
                                      i_stride2[2], bS, i_qpc, 1,
                                      h->loopf.deblock_v_chroma, h->loopf.deblock_v_chroma_intra );
                    }
                }
            }
        }

        /* next mb */
        if( !b_interlaced || (mb_y&1) )
            mb_x++;
        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 );
}

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 );
void x264_deblock_h_chroma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );

#ifdef ARCH_X86_64
void x264_deblock_v_luma_sse2( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_h_luma_sse2( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
#else
void x264_deblock_h_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_v8_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );

void x264_deblock_v_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
{
    x264_deblock_v8_luma_mmxext( pix,   stride, alpha, beta, tc0   );
    x264_deblock_v8_luma_mmxext( pix+8, stride, alpha, beta, tc0+2 );
}
#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_h_luma = deblock_h_luma_c;
    pf->deblock_v_chroma = deblock_v_chroma_c;
    pf->deblock_h_chroma = deblock_h_chroma_c;
    pf->deblock_v_luma_intra = deblock_v_luma_intra_c;
    pf->deblock_h_luma_intra = deblock_h_luma_intra_c;
    pf->deblock_v_chroma_intra = deblock_v_chroma_intra_c;
    pf->deblock_h_chroma_intra = deblock_h_chroma_intra_c;

#ifdef HAVE_MMX
    if( cpu&X264_CPU_MMXEXT )
    {
        pf->deblock_v_chroma = x264_deblock_v_chroma_mmxext;
        pf->deblock_h_chroma = x264_deblock_h_chroma_mmxext;
        pf->deblock_v_chroma_intra = x264_deblock_v_chroma_intra_mmxext;
        pf->deblock_h_chroma_intra = x264_deblock_h_chroma_intra_mmxext;

#ifdef ARCH_X86_64
        if( cpu&X264_CPU_SSE2 )
        {
            pf->deblock_v_luma = x264_deblock_v_luma_sse2;
            pf->deblock_h_luma = x264_deblock_h_luma_sse2;
        }
#else
        pf->deblock_v_luma = x264_deblock_v_luma_mmxext;
        pf->deblock_h_luma = x264_deblock_h_luma_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 )
{
    pthread_mutex_lock( &frame->mutex );
    frame->i_lines_completed = i_lines_completed;
    pthread_cond_broadcast( &frame->cv );
    pthread_mutex_unlock( &frame->mutex );
}

void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
{
    pthread_mutex_lock( &frame->mutex );
    while( frame->i_lines_completed < i_lines_completed )
        pthread_cond_wait( &frame->cv, &frame->mutex );
    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 );
}