Hi,
I've now enabled by default the fix for the low-frequency problems and
so far it seems to be an improvement in terms of CPU too. Please try it
out and report and success/failure.
Cheers,
Jean-Marc
Jean-Marc Valin a ?crit :> Hi,
>
> We've just found an issue when encoding signals that have very powerful
> bass (e.g. a loud 40 Hz tone). Turns out that was a combination of the
> MDCT leakage with some inefficiency in the VQ search. There's a
> tentative fix for that in git, which I'm encouraging everyone to test.
> To enable the fix, look for these lines in alg_quant() (vq.c):
>
> #if 0
> if (K > (N>>1))
>
> Just replace with #if 1 and the fix will be enabled. I'm not yet 100%
> sure it's safe, so please report and problem (or success). The good
> thing is that this should not only solve the original, but it should
> also reduce the CPU usage.
>
> Cheers,
>
> Jean-Marc
> _______________________________________________
> celt-dev mailing list
> celt-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/celt-dev
>
>