macroblock.c

法国人的264代码 大家看看啊,里面有很多可以借鉴的东西啊

                int oe = block_idx_x[i]*4 + block_idx_y[i]*4*FENC_STRIDE;
                int od = block_idx_x[i]*4 + block_idx_y[i]*4*FDEC_STRIDE;
                sub_zigzag_4x4( h->dct.block[16+i+ch*4].residual_ac, p_src+oe, p_dst+od );
                h->dct.chroma_dc[ch][i] = p_src[oe] - p_dst[od];
                p_dst[od] = p_src[oe];
            }
            continue;
        }
            
        h->dctf.sub8x8_dct( dct4x4, p_src, p_dst );
        /* calculate dct coeffs */
        for( i = 0; i < 4; i++ )
        {
            /* copy dc coeff */
            dct2x2[block_idx_y[i]][block_idx_x[i]] = dct4x4[i][0][0];

            /* no trellis; it doesn't seem to help chroma noticeably */
            quant_4x4( h, dct4x4[i], h->quant4_mf[CQM_4IC + b_inter], i_qscale, !b_inter );
            scan_zigzag_4x4( h->dct.block[16+i+ch*4].residual_ac, dct4x4[i] );

            if( b_decimate )
            {
                i_decimate_score += x264_mb_decimate_score( h->dct.block[16+i+ch*4].residual_ac, 15 );
            }
        }

        h->dctf.dct2x2dc( dct2x2 );
        quant_2x2_dc( h, dct2x2, h->quant4_mf[CQM_4IC + b_inter], i_qscale, !b_inter );
        scan_zigzag_2x2_dc( h->dct.chroma_dc[ch], dct2x2 );

        /* output samples to fdec */
        h->dctf.idct2x2dc( dct2x2 );
        x264_mb_dequant_2x2_dc( dct2x2, h->dequant4_mf[CQM_4IC + b_inter], i_qscale );  /* XXX not inversed */

        if( b_decimate && i_decimate_score < 7 )
        {
            /* Near null chroma 8x8 block so make it null (bits saving) */
            memset( &h->dct.block[16+ch*4], 0, 4 * sizeof( *h->dct.block ) );
            if( !array_non_zero( (int*)dct2x2, sizeof(dct2x2)/sizeof(int) ) )
                continue;
            memset( dct4x4, 0, sizeof( dct4x4 ) );
        }
        else
        {
            for( i = 0; i < 4; i++ )
                h->quantf.dequant_4x4( dct4x4[i], h->dequant4_mf[CQM_4IC + b_inter], i_qscale );
        }

        for( i = 0; i < 4; i++ )
            dct4x4[i][0][0] = dct2x2[0][i];
        h->dctf.add8x8_idct( p_dst, dct4x4 );
    }
}

static void x264_macroblock_encode_skip( x264_t *h )
{
    int i;
    h->mb.i_cbp_luma = 0x00;
    h->mb.i_cbp_chroma = 0x00;

    for( i = 0; i < 16+8; i++ )
    {
        h->mb.cache.non_zero_count[x264_scan8[i]] = 0;
    }

    /* store cbp */
    h->mb.cbp[h->mb.i_mb_xy] = 0;
}

/*****************************************************************************
 * x264_macroblock_encode_pskip:
 *  Encode an already marked skip block
 *****************************************************************************/
void x264_macroblock_encode_pskip( x264_t *h )
{
    const int mvx = x264_clip3( h->mb.cache.mv[0][x264_scan8[0]][0],
                                h->mb.mv_min[0], h->mb.mv_max[0] );
    const int mvy = x264_clip3( h->mb.cache.mv[0][x264_scan8[0]][1],
                                h->mb.mv_min[1], h->mb.mv_max[1] );

    /* Motion compensation XXX probably unneeded */
    h->mc.mc_luma( h->mb.pic.p_fref[0][0], h->mb.pic.i_stride[0],
                   h->mb.pic.p_fdec[0],    FDEC_STRIDE,
                   mvx, mvy, 16, 16 );

    /* Chroma MC */
    h->mc.mc_chroma( h->mb.pic.p_fref[0][0][4], h->mb.pic.i_stride[1],
                     h->mb.pic.p_fdec[1],       FDEC_STRIDE,
                     mvx, mvy, 8, 8 );

    h->mc.mc_chroma( h->mb.pic.p_fref[0][0][5], h->mb.pic.i_stride[2],
                     h->mb.pic.p_fdec[2],       FDEC_STRIDE,
                     mvx, mvy, 8, 8 );

    x264_macroblock_encode_skip( h );
}

/*****************************************************************************
 * x264_macroblock_encode:
 *****************************************************************************/
void x264_macroblock_encode( x264_t *h )
{
    int i_cbp_dc = 0;
    int i_qp = h->mb.i_qp;
    int b_decimate = h->sh.i_type == SLICE_TYPE_B || h->param.analyse.b_dct_decimate;
    int i;

    if( h->mb.i_type == P_SKIP )
    {
        /* A bit special */
        x264_macroblock_encode_pskip( h );
        return;
    }
    if( h->mb.i_type == B_SKIP )
    {
        /* XXX motion compensation is probably unneeded */
        x264_mb_mc( h );
        x264_macroblock_encode_skip( h );
        return;
    }

    if( h->mb.i_type == I_16x16 )
    {
        const int i_mode = h->mb.i_intra16x16_pred_mode;
        h->mb.b_transform_8x8 = 0;
        /* do the right prediction */
        h->predict_16x16[i_mode]( h->mb.pic.p_fdec[0] );

        /* encode the 16x16 macroblock */
        x264_mb_encode_i16x16( h, i_qp );
    }
    else if( h->mb.i_type == I_8x8 )
    {
        DECLARE_ALIGNED( uint8_t, edge[33], 8 );
        h->mb.b_transform_8x8 = 1;
        for( i = 0; i < 4; i++ )
        {
            uint8_t  *p_dst = &h->mb.pic.p_fdec[0][8 * (i&1) + 8 * (i>>1) * FDEC_STRIDE];
            int      i_mode = h->mb.cache.intra4x4_pred_mode[x264_scan8[4*i]];

            x264_predict_8x8_filter( p_dst, edge, h->mb.i_neighbour8[i], x264_pred_i4x4_neighbors[i_mode] );
            h->predict_8x8[i_mode]( p_dst, edge );
            x264_mb_encode_i8x8( h, i, i_qp );
        }
    }
    else if( h->mb.i_type == I_4x4 )
    {
        h->mb.b_transform_8x8 = 0;
        for( i = 0; i < 16; i++ )
        {
            uint8_t  *p_dst = &h->mb.pic.p_fdec[0][4 * block_idx_x[i] + 4 * block_idx_y[i] * FDEC_STRIDE];
            int      i_mode = h->mb.cache.intra4x4_pred_mode[x264_scan8[i]];

            if( (h->mb.i_neighbour4[i] & (MB_TOPRIGHT|MB_TOP)) == MB_TOP )
                /* emulate missing topright samples */
                *(uint32_t*) &p_dst[4-FDEC_STRIDE] = p_dst[3-FDEC_STRIDE] * 0x01010101U;

            h->predict_4x4[i_mode]( p_dst );
            x264_mb_encode_i4x4( h, i, i_qp );
        }
    }
    else    /* Inter MB */
    {
        int i8x8, i4x4, idx;
        int i_decimate_mb = 0;

        /* Motion compensation */
        x264_mb_mc( h );

        if( h->mb.b_lossless )
        {
            for( i4x4 = 0; i4x4 < 16; i4x4++ )
            {
                int x = 4*block_idx_x[i4x4];
                int y = 4*block_idx_y[i4x4];
                sub_zigzag_4x4full( h->dct.block[i4x4].luma4x4,
                                    h->mb.pic.p_fenc[0]+x+y*FENC_STRIDE,
                                    h->mb.pic.p_fdec[0]+x+y*FDEC_STRIDE );
            }
        }
        else if( h->mb.b_transform_8x8 )
        {
            int16_t dct8x8[4][8][8];
            int nnz8x8[4] = {1,1,1,1};
            b_decimate &= !h->mb.b_trellis; // 8x8 trellis is inherently optimal decimation
            h->dctf.sub16x16_dct8( dct8x8, h->mb.pic.p_fenc[0], h->mb.pic.p_fdec[0] );

            for( idx = 0; idx < 4; idx++ )
            {
                if( h->mb.b_noise_reduction )
                    x264_denoise_dct( h, (int16_t*)dct8x8[idx] );
                if( h->mb.b_trellis )
                    x264_quant_8x8_trellis( h, dct8x8[idx], CQM_8PY, i_qp, 0 );
                else
                    quant_8x8( h, dct8x8[idx], h->quant8_mf[CQM_8PY], i_qp, 0 );

                scan_zigzag_8x8full( h->dct.luma8x8[idx], dct8x8[idx] );

                if( b_decimate )
                {
                    int i_decimate_8x8 = x264_mb_decimate_score( h->dct.luma8x8[idx], 64 );
                    i_decimate_mb += i_decimate_8x8;
                    if( i_decimate_8x8 < 4 )
                    {
                        memset( h->dct.luma8x8[idx], 0, sizeof( h->dct.luma8x8[idx] ) );
                        memset( dct8x8[idx], 0, sizeof( dct8x8[idx] ) );
                        nnz8x8[idx] = 0;
                    }
                }
                else
                    nnz8x8[idx] = array_non_zero( (int*)dct8x8[idx], sizeof(*dct8x8)/sizeof(int) );
            }

            if( i_decimate_mb < 6 && b_decimate )
                memset( h->dct.luma8x8, 0, sizeof( h->dct.luma8x8 ) );
            else
            {
                for( idx = 0; idx < 4; idx++ )
                    if( nnz8x8[idx] )
                    {
                        h->quantf.dequant_8x8( dct8x8[idx], h->dequant8_mf[CQM_8PY], i_qp );
                        h->dctf.add8x8_idct8( &h->mb.pic.p_fdec[0][(idx&1)*8 + (idx>>1)*8*FDEC_STRIDE], dct8x8[idx] );
                    }
            }
        }
        else
        {
            int16_t dct4x4[16][4][4];
            int nnz8x8[4] = {1,1,1,1};
            h->dctf.sub16x16_dct( dct4x4, h->mb.pic.p_fenc[0], h->mb.pic.p_fdec[0] );

            for( i8x8 = 0; i8x8 < 4; i8x8++ )
            {
                int i_decimate_8x8;

                /* encode one 4x4 block */
                i_decimate_8x8 = 0;
                for( i4x4 = 0; i4x4 < 4; i4x4++ )
                {
                    idx = i8x8 * 4 + i4x4;

                    if( h->mb.b_noise_reduction )
                        x264_denoise_dct( h, (int16_t*)dct4x4[idx] );
                    if( h->mb.b_trellis )
                        x264_quant_4x4_trellis( h, dct4x4[idx], CQM_4PY, i_qp, DCT_LUMA_4x4, 0 );
                    else
                        quant_4x4( h, dct4x4[idx], h->quant4_mf[CQM_4PY], i_qp, 0 );

                    scan_zigzag_4x4full( h->dct.block[idx].luma4x4, dct4x4[idx] );
                    
                    if( b_decimate )
                        i_decimate_8x8 += x264_mb_decimate_score( h->dct.block[idx].luma4x4, 16 );
                }

                /* decimate this 8x8 block */
                i_decimate_mb += i_decimate_8x8;
                if( i_decimate_8x8 < 4 && b_decimate )
                {
                    memset( &dct4x4[i8x8*4], 0, 4 * sizeof( *dct4x4 ) );
                    memset( &h->dct.block[i8x8*4], 0, 4 * sizeof( *h->dct.block ) );
                    nnz8x8[i8x8] = 0;
                }
            }

            if( i_decimate_mb < 6 && b_decimate )
                memset( h->dct.block, 0, 16 * sizeof( *h->dct.block ) );
            else
            {
                for( i8x8 = 0; i8x8 < 4; i8x8++ )
                    if( nnz8x8[i8x8] )
                    {
                        for( i = 0; i < 4; i++ )
                            h->quantf.dequant_4x4( dct4x4[i8x8*4+i], h->dequant4_mf[CQM_4PY], i_qp );
                        h->dctf.add8x8_idct( &h->mb.pic.p_fdec[0][(i8x8&1)*8 + (i8x8>>1)*8*FDEC_STRIDE], &dct4x4[i8x8*4] );
                    }
            }
        }
    }

    /* encode chroma */
    i_qp = i_chroma_qp_table[x264_clip3( i_qp + h->pps->i_chroma_qp_index_offset, 0, 51 )];
    if( IS_INTRA( h->mb.i_type ) )
    {
        const int i_mode = h->mb.i_chroma_pred_mode;
        h->predict_8x8c[i_mode]( h->mb.pic.p_fdec[1] );
        h->predict_8x8c[i_mode]( h->mb.pic.p_fdec[2] );
    }

    /* encode the 8x8 blocks */
    x264_mb_encode_8x8_chroma( h, !IS_INTRA( h->mb.i_type ), i_qp );

    /* Calculate the Luma/Chroma patern and non_zero_count */
    h->mb.i_cbp_luma = 0x00;
    if( h->mb.i_type == I_16x16 )
    {
        for( i = 0; i < 16; i++ )
        {
            const int nz = array_non_zero_count( h->dct.block[i].residual_ac, 15 );
            h->mb.cache.non_zero_count[x264_scan8[i]] = nz;
            if( nz > 0 )
                h->mb.i_cbp_luma = 0x0f;
        }
    }
    else if( h->mb.b_transform_8x8 )
    {
        /* coded_block_flag is enough for CABAC.
         * the full non_zero_count is done only in CAVLC. */
        for( i = 0; i < 4; i++ )
        {
            const int nz = array_non_zero( h->dct.luma8x8[i], 64 );
            int j;
            for( j = 0; j < 4; j++ )
                h->mb.cache.non_zero_count[x264_scan8[4*i+j]] = nz;
            if( nz > 0 )