comparison apedec.c @ 10644:5da7180afadf libavcodec

refactor and optimize scalarproduct 29-105% faster apply_filter, 6-90% faster ape decoding on core2 (Any x86 other than core2 probably gets much less, since this is mostly due to ssse3 cachesplit avoidance and I haven't written the full gamut of other cachesplit modes.) 9-123% faster ape decoding on G4.

10643:7f6911429cdc

{
    do_init_filter(&f[0], buf, order);
    do_init_filter(&f[1], buf + order * 3 + HISTORY_SIZE, order);
}

static inline void do_apply_filter(APEContext * ctx, int version, APEFilter *f, int32_t *data, int count, int order, int fracbits)
{
    int res;
    int absres;

    while (count--) {
        /* round fixedpoint scalar product */
        res = (ctx->dsp.scalarproduct_int16(f->delay - order, f->coeffs, order, 0) + (1 << (fracbits - 1))) >> fracbits;

        if (*data < 0)
            ctx->dsp.add_int16(f->coeffs, f->adaptcoeffs - order, order);
        else if (*data > 0)
            ctx->dsp.sub_int16(f->coeffs, f->adaptcoeffs - order, order);

        res += *data;

        *data++ = res;

        /* Update the output history */
        *f->delay++ = av_clip_int16(res);

10644:5da7180afadf

{
    do_init_filter(&f[0], buf, order);
    do_init_filter(&f[1], buf + order * 3 + HISTORY_SIZE, order);
}

static void do_apply_filter(APEContext * ctx, int version, APEFilter *f, int32_t *data, int count, int order, int fracbits)
{
    int res;
    int absres;

    while (count--) {
        /* round fixedpoint scalar product */
        res = ctx->dsp.scalarproduct_and_madd_int16(f->coeffs, f->delay - order, f->adaptcoeffs - order, order, APESIGN(*data));
        res = (res + (1 << (fracbits - 1))) >> fracbits;
        res += *data;
        *data++ = res;

        /* Update the output history */
        *f->delay++ = av_clip_int16(res);