comparison vp8.c @ 12342:b4c63ffd959b libavcodec

VP8: much faster DC transform handling A lot of the time the DC block is empty: don't do the WHT in this case. A lot of the rest of the time, there's only one coefficient: make a special DC-only transform for that case. When the block is empty, don't incorrectly mark luma DCT blocks as having DC coefficients.

12341:ad24cca213ae

                      uint8_t t_nnz[9], uint8_t l_nnz[9])
{
    int i, x, y, luma_start = 0, luma_ctx = 3;
    int nnz_pred, nnz, nnz_total = 0;
    int segment = s->segment;

    if (mb->mode != MODE_I4x4 && mb->mode != VP8_MVMODE_SPLIT) {
        nnz_pred = t_nnz[8] + l_nnz[8];

        // decode DC values and do hadamard
        nnz = decode_block_coeffs(c, s->block_dc, s->prob->token[1], 0, nnz_pred,
                                  s->qmat[segment].luma_dc_qmul);
        l_nnz[8] = t_nnz[8] = !!nnz;
        nnz_total += nnz;
        s->vp8dsp.vp8_luma_dc_wht(s->block, s->block_dc);
        luma_start = 1;
        luma_ctx = 0;
    }

    // luma blocks
    for (y = 0; y < 4; y++)
        for (x = 0; x < 4; x++) {
            nnz_pred = l_nnz[y] + t_nnz[x];
            nnz = decode_block_coeffs(c, s->block[y][x], s->prob->token[luma_ctx], luma_start,
                                      nnz_pred, s->qmat[segment].luma_qmul);
            // nnz+luma_start may be one more than the actual last index, but we don't care
            s->non_zero_count_cache[y][x] = nnz + luma_start;
            t_nnz[x] = l_nnz[y] = !!nnz;
            nnz_total += nnz;
        }

    // chroma blocks

12342:b4c63ffd959b

                      uint8_t t_nnz[9], uint8_t l_nnz[9])
{
    int i, x, y, luma_start = 0, luma_ctx = 3;
    int nnz_pred, nnz, nnz_total = 0;
    int segment = s->segment;
    int block_dc = 0;

    if (mb->mode != MODE_I4x4 && mb->mode != VP8_MVMODE_SPLIT) {
        nnz_pred = t_nnz[8] + l_nnz[8];

        // decode DC values and do hadamard
        nnz = decode_block_coeffs(c, s->block_dc, s->prob->token[1], 0, nnz_pred,
                                  s->qmat[segment].luma_dc_qmul);
        l_nnz[8] = t_nnz[8] = !!nnz;
        if (nnz) {
            nnz_total += nnz;
            block_dc = 1;
            if (nnz == 1)
                s->vp8dsp.vp8_luma_dc_wht_dc(s->block, s->block_dc);
            else
                s->vp8dsp.vp8_luma_dc_wht(s->block, s->block_dc);
        }
        luma_start = 1;
        luma_ctx = 0;
    }

    // luma blocks
    for (y = 0; y < 4; y++)
        for (x = 0; x < 4; x++) {
            nnz_pred = l_nnz[y] + t_nnz[x];
            nnz = decode_block_coeffs(c, s->block[y][x], s->prob->token[luma_ctx], luma_start,
                                      nnz_pred, s->qmat[segment].luma_qmul);
            // nnz+block_dc may be one more than the actual last index, but we don't care
            s->non_zero_count_cache[y][x] = nnz + block_dc;
            t_nnz[x] = l_nnz[y] = !!nnz;
            nnz_total += nnz;
        }

    // chroma blocks