analyse.c

H.264 source codes

            x264_me_search( h, m, &lX->me16x16.mv, 1 );

            x264_macroblock_cache_mv( h, 2*x8, 2*y8, 2, 2, l, m->mv[0], m->mv[1] );

            /* BI mode */
            h->mc.mc_luma( m->p_fref, m->i_stride[0], pix[l], 8,
                            m->mv[0], m->mv[1], 8, 8 );
            i_part_cost_bi += m->cost_mv;
            /* FIXME: ref cost */
        }

        WEIGHTED_AVG( PIXEL_8x8, pix[0], 8, pix[1], 8 );
        i_part_cost_bi += h->pixf.satd[PIXEL_8x8]( a->l0.me8x8[i].p_fenc[0], h->mb.pic.i_stride[0], pix[0], 8 )
                        + a->i_lambda * i_sub_mb_b_cost_table[D_BI_8x8];
        a->l0.me8x8[i].cost += a->i_lambda * i_sub_mb_b_cost_table[D_L0_8x8];
        a->l1.me8x8[i].cost += a->i_lambda * i_sub_mb_b_cost_table[D_L1_8x8];

        i_part_cost = a->l0.me8x8[i].cost;
        h->mb.i_sub_partition[i] = D_L0_8x8;
        if( a->l1.me8x8[i].cost < i_part_cost )
        {
            i_part_cost = a->l1.me8x8[i].cost;
            h->mb.i_sub_partition[i] = D_L1_8x8;
        }
        if( i_part_cost_bi < i_part_cost )
        {
            i_part_cost = i_part_cost_bi;
            h->mb.i_sub_partition[i] = D_BI_8x8;
        }
        if( a->i_cost8x8direct[i] < i_part_cost )
        {
            i_part_cost = a->i_cost8x8direct[i];
            h->mb.i_sub_partition[i] = D_DIRECT_8x8;
        }
        a->i_cost8x8bi += i_part_cost;

        /* XXX Needed for x264_mb_predict_mv */
        x264_mb_cache_mv_b8x8( h, a, i, 0 );
    }

    /* mb type cost */
    a->i_cost8x8bi += a->i_lambda * i_mb_b_cost_table[B_8x8];
}

static void x264_mb_analyse_inter_b16x8( x264_t *h, x264_mb_analysis_t *a )
{
    uint8_t **p_fref[2] =
        { h->mb.pic.p_fref[0][a->l0.i_ref],
          h->mb.pic.p_fref[1][a->l1.i_ref] };
    uint8_t pix[2][16*8];
    int mvc[2][2];
    int i, l;

    h->mb.i_partition = D_16x8;
    a->i_cost16x8bi = 0;

    for( i = 0; i < 2; i++ )
    {
        int i_part_cost;
        int i_part_cost_bi = 0;

        /* TODO: check only the list(s) that were used in b8x8? */
        for( l = 0; l < 2; l++ )
        {
            x264_mb_analysis_list_t *lX = l ? &a->l1 : &a->l0;
            x264_me_t *m = &lX->me16x8[i];

            m->i_pixel = PIXEL_16x8;
            m->p_cost_mv = a->p_cost_mv;

            LOAD_FENC( m, h->mb.pic.p_fenc, 0, 8*i );
            LOAD_HPELS( m, p_fref[l], 0, 8*i );

            mvc[0][0] = lX->me8x8[2*i].mv[0];
            mvc[0][1] = lX->me8x8[2*i].mv[1];
            mvc[1][0] = lX->me8x8[2*i+1].mv[0];
            mvc[1][1] = lX->me8x8[2*i+1].mv[1];

            x264_mb_predict_mv( h, 0, 8*i, 2, m->mvp );
            x264_me_search( h, m, mvc, 2 );

            /* BI mode */
            h->mc.mc_luma( m->p_fref, m->i_stride[0], pix[l], 16,
                            m->mv[0], m->mv[1], 16, 8 );
            /* FIXME: ref cost */
            i_part_cost_bi += m->cost_mv;
        }

        WEIGHTED_AVG( PIXEL_16x8, pix[0], 16, pix[1], 16 );
        i_part_cost_bi += h->pixf.satd[PIXEL_16x8]( a->l0.me16x8[i].p_fenc[0], h->mb.pic.i_stride[0], pix[0], 16 );

        i_part_cost = a->l0.me16x8[i].cost;
        a->i_mb_partition16x8[i] = D_L0_8x8; /* not actually 8x8, only the L0 matters */
        if( a->l1.me16x8[i].cost < i_part_cost )
        {
            i_part_cost = a->l1.me16x8[i].cost;
            a->i_mb_partition16x8[i] = D_L1_8x8;
        }
        if( i_part_cost_bi + a->i_lambda * 1 < i_part_cost )
        {
            i_part_cost = i_part_cost_bi;
            a->i_mb_partition16x8[i] = D_BI_8x8;
        }
        a->i_cost16x8bi += i_part_cost;

        if( i == 0 )
            x264_mb_cache_mv_b16x8( h, a, i, 0 );
    }

    /* mb type cost */
    a->i_mb_type16x8 = B_L0_L0
        + (a->i_mb_partition16x8[0]>>2) * 3
        + (a->i_mb_partition16x8[1]>>2);
    a->i_cost16x8bi += a->i_lambda * i_mb_b16x8_cost_table[a->i_mb_type16x8];
}
static void x264_mb_analyse_inter_b8x16( x264_t *h, x264_mb_analysis_t *a )
{
    uint8_t **p_fref[2] =
        { h->mb.pic.p_fref[0][a->l0.i_ref],
          h->mb.pic.p_fref[1][a->l1.i_ref] };
    uint8_t pix[2][8*16];
    int mvc[2][2];
    int i, l;

    h->mb.i_partition = D_8x16;
    a->i_cost8x16bi = 0;

    for( i = 0; i < 2; i++ )
    {
        int i_part_cost;
        int i_part_cost_bi = 0;

        for( l = 0; l < 2; l++ )
        {
            x264_mb_analysis_list_t *lX = l ? &a->l1 : &a->l0;
            x264_me_t *m = &lX->me8x16[i];

            m->i_pixel = PIXEL_8x16;
            m->p_cost_mv = a->p_cost_mv;

            LOAD_FENC( m, h->mb.pic.p_fenc, 8*i, 0 );
            LOAD_HPELS( m, p_fref[l], 8*i, 0 );

            mvc[0][0] = lX->me8x8[i].mv[0];
            mvc[0][1] = lX->me8x8[i].mv[1];
            mvc[1][0] = lX->me8x8[i+2].mv[0];
            mvc[1][1] = lX->me8x8[i+2].mv[1];

            x264_mb_predict_mv( h, 0, 4*i, 2, m->mvp );
            x264_me_search( h, m, mvc, 2 );

            /* BI mode */
            h->mc.mc_luma( m->p_fref, m->i_stride[0], pix[l], 8,
                            m->mv[0], m->mv[1], 8, 16 );
            /* FIXME: ref cost */
            i_part_cost_bi += m->cost_mv;
        }

        WEIGHTED_AVG( PIXEL_8x16, pix[0], 8, pix[1], 8 );
        i_part_cost_bi += h->pixf.satd[PIXEL_8x16]( a->l0.me8x16[i].p_fenc[0], h->mb.pic.i_stride[0], pix[0], 8 );

        i_part_cost = a->l0.me8x16[i].cost;
        a->i_mb_partition8x16[i] = D_L0_8x8;
        if( a->l1.me8x16[i].cost < i_part_cost )
        {
            i_part_cost = a->l1.me8x16[i].cost;
            a->i_mb_partition8x16[i] = D_L1_8x8;
        }
        if( i_part_cost_bi + a->i_lambda * 1 < i_part_cost )
        {
            i_part_cost = i_part_cost_bi;
            a->i_mb_partition8x16[i] = D_BI_8x8;
        }
        a->i_cost8x16bi += i_part_cost;

        if( i == 0 )
            x264_mb_cache_mv_b8x16( h, a, i, 0 );
    }

    /* mb type cost */
    a->i_mb_type8x16 = B_L0_L0
        + (a->i_mb_partition8x16[0]>>2) * 3
        + (a->i_mb_partition8x16[1]>>2);
    a->i_cost8x16bi += a->i_lambda * i_mb_b16x8_cost_table[a->i_mb_type8x16];
}

/*****************************************************************************
 * x264_macroblock_analyse:
 *****************************************************************************/
void x264_macroblock_analyse( x264_t *h )
{
    x264_mb_analysis_t analysis;
    int i;

    h->mb.qp[h->mb.i_mb_xy] = x264_ratecontrol_qp(h);

    /* prevent QP from varying too fast. FIXME what's a sane limit? */
    h->mb.qp[h->mb.i_mb_xy] = x264_clip3( h->mb.qp[h->mb.i_mb_xy],
                                          h->mb.i_last_qp - 12, h->mb.i_last_qp + 12 );

    /* init analysis */
    x264_mb_analyse_init( h, &analysis, h->mb.qp[h->mb.i_mb_xy] );

    /*--------------------------- Do the analysis ---------------------------*/
    if( h->sh.i_type == SLICE_TYPE_I )
    {
        x264_mb_analyse_intra( h, &analysis, COST_MAX );

        if( analysis.i_sad_i4x4 < analysis.i_sad_i16x16 )
            h->mb.i_type = I_4x4;
        else
            h->mb.i_type = I_16x16;
    }
    else if( h->sh.i_type == SLICE_TYPE_P )
    {
        const unsigned int i_neighbour = h->mb.i_neighbour;

        int b_skip = 0;
        int i_cost;
        int i_intra_cost, i_intra_type;

        /* Fast P_SKIP detection */
        if( ( (i_neighbour&MB_LEFT) && h->mb.type[h->mb.i_mb_xy - 1] == P_SKIP ) ||
            ( (i_neighbour&MB_TOP) && h->mb.type[h->mb.i_mb_xy - h->mb.i_mb_stride] == P_SKIP ) ||
            ( ((i_neighbour&(MB_TOP|MB_LEFT)) == (MB_TOP|MB_LEFT) ) && h->mb.type[h->mb.i_mb_xy - h->mb.i_mb_stride-1 ] == P_SKIP ) ||
            ( (i_neighbour&MB_TOPRIGHT) && h->mb.type[h->mb.i_mb_xy - h->mb.i_mb_stride+1 ] == P_SKIP ) )
        {
            b_skip = x264_macroblock_probe_pskip( h );
        }

        if( b_skip )
        {
            h->mb.i_type = P_SKIP;
            h->mb.i_partition = D_16x16;
        }
        else
        {
            const unsigned int flags = h->param.analyse.inter;
            int i_type;
            int i_partition;

            x264_mb_analyse_load_costs( h, &analysis );

            x264_mb_analyse_inter_p16x16( h, &analysis );
            if( flags & X264_ANALYSE_PSUB16x16 )
                x264_mb_analyse_inter_p8x8( h, &analysis );

            /* Select best inter mode */
            i_type = P_L0;
            i_partition = D_16x16;
            i_cost = analysis.l0.me16x16.cost;

            if( ( flags & X264_ANALYSE_PSUB16x16 ) &&
                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost )
            {
                int i;

                i_type = P_8x8;
                i_partition = D_8x8;
                h->mb.i_sub_partition[0] = D_L0_8x8;
                h->mb.i_sub_partition[1] = D_L0_8x8;
                h->mb.i_sub_partition[2] = D_L0_8x8;
                h->mb.i_sub_partition[3] = D_L0_8x8;

                i_cost = analysis.l0.i_cost8x8;

                /* Do sub 8x8 */
                if( flags & X264_ANALYSE_PSUB8x8 )
                {
                    for( i = 0; i < 4; i++ )
                    {
                        x264_mb_analyse_inter_p4x4( h, &analysis, i );
                        if( analysis.l0.i_cost4x4[i] < analysis.l0.me8x8[i].cost )
                        {
                            int i_cost8x8;

                            h->mb.i_sub_partition[i] = D_L0_4x4;
                            i_cost8x8 = analysis.l0.i_cost4x4[i];

                            x264_mb_analyse_inter_p8x4( h, &analysis, i );
                            if( analysis.l0.i_cost8x4[i] < analysis.l0.i_cost4x4[i] )
                            {
                                h->mb.i_sub_partition[i] = D_L0_8x4;
                                i_cost8x8 = analysis.l0.i_cost8x4[i];
                            }

                            x264_mb_analyse_inter_p4x8( h, &analysis, i );
                            if( analysis.l0.i_cost4x8[i] < analysis.l0.i_cost4x4[i] )
                            {
                                h->mb.i_sub_partition[i] = D_L0_4x8;
                                i_cost8x8 = analysis.l0.i_cost4x8[i];
                            }

                            i_cost += i_cost8x8 - analysis.l0.me8x8[i].cost;
                        }
                    }
                }

                /* Now do sub 16x8/8x16 */
                x264_mb_analyse_inter_p16x8( h, &analysis );
                if( analysis.l0.i_cost16x8 < i_cost )
                {
                    i_type = P_L0;
                    i_partition = D_16x8;
                    i_cost = analysis.l0.i_cost16x8;
                }

                x264_mb_analyse_inter_p8x16( h, &analysis );
                if( analysis.l0.i_cost8x16 < i_cost )
                {
                    i_type = P_L0;
                    i_partition = D_8x16;
                    i_cost = analysis.l0.i_cost8x16;
                }
            }

            h->mb.i_type = i_type;
            h->mb.i_partition = i_partition;

            /* refine qpel */
            if( h->mb.i_partition == D_16x16 )
            {
                x264_me_refine_qpel( h, &analysis.l0.me16x16 );
                i_cost = analysis.l0.me16x16.cost;
            }
            else if( h->mb.i_partition == D_16x8 )
            {
                x264_me_refine_qpel( h, &analysis.l0.me16x8[0] );
                x264_me_refine_qpel( h, &analysis.l0.me16x8[1] );
                i_cost = analysis.l0.me16x8[0].cost + analysis.l0.me16x8[1].cost;
            }
            else if( h->mb.i_partition == D_8x16 )
            {
                x264_me_refine_qpel( h, &analysis.l0.me8x16[0] );
                x264_me_refine_qpel( h, &analysis.l0.me8x16[1] );
                i_cost = analysis.l0.me8x16[0].cost + analysis.l0.me8x16[1].cost;
            }
            else if( h->mb.i_partition == D_8x8 )
            {
                int i8x8;
                i_cost = 0;
                for( i8x8 = 0; i8x8 < 4; i8x8++ )
                {
                    switch( h->mb.i_sub_partition[i8x8] )
                    {
                        case D_L0_8x8:
                            x264_me_refine_qpel( h, &analysis.l0.me8x8[i8x8] );
                            i_cost += analysis.l0.me8x8[i8x8].cost;
                            break;
                        case D_L0_8x4:
                            x264_me_refine_qpel( h, &analysis.l0.me8x4[i8x8][0] );
                            x264_me_refine_qpel( h, &analysis.l0.me8x4[i8x8][1] );
                            i_cost += analysis.l0.me8x4[i8x8][0].cost +
                                      analysis.l0.me8x4[i8x8][1].cost;
                            break;
                        case D_L0_4x8:
                            x264_me_refine_qpel( h, &analysis.l0.me4x8[i8x8][0] );
                            x264_me_refine_qpel( h, &analysis.l0.me4x8[i8x8][1] );
                            i_cost += analysis.l0.me4x8[i8x8][0].cost +
                                      analysis.l0.me4x8[i8x8][1].cost;
                            break;

                        case D_L0_4x4:
                            x264_me_refine_qpel( h, &analysis.l0.me4x4[i8x8][0] );
                            x264_me_refine_qpel( h, &analysis.l0.me4x4[i8x8][1] );
                            x264_me_refine_qpel( h, &analysis.l0.me4x4[i8x8][2] );
                            x264_me_refine_qpel( h, &analysis.l0.me4x4[i8x8][3] );
                            i_cost += analysis.l0.me4x4[i8x8][0].cost +
                                      analysis.l0.me4x4[i8x8][1].cost +
                                      analysis.l0.me4x4[i8x8][2].cost +
                                      analysis.l0.me4x4[i8x8][3].cost;