comparison ac3dec.c @ 10865:bcdc5343a577 libavcodec

Use MULH instead of 64x64 multiplication, around 5% overall speedup on Intel Atom. Patch by myself and Yuriy Kaminskiy [yumkam mail ru]

10864:e3f5eb016712

    for (band = 0; band < s->num_cpl_bands; band++) {
        int band_start = bin;
        int band_end = bin + s->cpl_band_sizes[band];
        for (ch = 1; ch <= s->fbw_channels; ch++) {
            if (s->channel_in_cpl[ch]) {
                int64_t cpl_coord = s->cpl_coords[ch][band];
                for (bin = band_start; bin < band_end; bin++) {
                    s->fixed_coeffs[ch][bin] = ((int64_t)s->fixed_coeffs[CPL_CH][bin] *
                                                cpl_coord) >> 23;
                }
                if (ch == 2 && s->phase_flags[band]) {
                    for (bin = band_start; bin < band_end; bin++)
                        s->fixed_coeffs[2][bin] = -s->fixed_coeffs[2][bin];
                }

10865:bcdc5343a577

    for (band = 0; band < s->num_cpl_bands; band++) {
        int band_start = bin;
        int band_end = bin + s->cpl_band_sizes[band];
        for (ch = 1; ch <= s->fbw_channels; ch++) {
            if (s->channel_in_cpl[ch]) {
                int cpl_coord = s->cpl_coords[ch][band] << 5;
                for (bin = band_start; bin < band_end; bin++) {
                    s->fixed_coeffs[ch][bin] = MULH(s->fixed_coeffs[CPL_CH][bin] << 4, cpl_coord);
                }
                if (ch == 2 && s->phase_flags[band]) {
                    for (bin = band_start; bin < band_end; bin++)
                        s->fixed_coeffs[2][bin] = -s->fixed_coeffs[2][bin];
                }