[CELT-dev] Implications of reducing MAX_PERIOD

Jean-Marc Valin jean-marc.valin at usherbrooke.ca
Sun Oct 25 18:20:32 PDT 2009

I just added a SMALL_FOOTPRINT option in git. This should reduce the
code size a bit more at the cost of a bit more CPU. Note that I've only
done a minimal amount of testing. This should be released as 0.7.0 soon.


Chen-Po Sun a écrit :
> That is great to hear!
> We don't need to worry about packet loss or anything like that as we are
> using CELT as a sound bank format, and as such don't need to worry about
> other encoders either since we are building into our own bank format.
> This will be a great memory saving for us.
> *Chen-Po Sun* | Programmer
> Firelight Technologies Pty Ltd.
> FMOD Sound System | www.fmod.org
> PH: +61 3 96635947 Fax: +61 3 96635951
> Gregory Maxwell wrote:
>> On Fri, Oct 23, 2009 at 7:00 AM, Jean-Marc Valin
>> <jean-marc.valin at usherbrooke.ca> wrote:
>>> If you don't need packet loss concealment *and* you don't need the
>>> encoder, then you can simply make it equal to zero.
>> Reducing/eliminating MAX_PERIOD will also make your decoder
>> incompatible with any encoder not changed in the same way unless you
>> disable the long term predictor (or run the encoder in complexity 1,
>> which disables the LTP).  Were it not for this compatibility issue
>> we'd probably make this a configure option.
>> But it's a perfectly fine and reasonable thing to do.  Disabling the
>> long term predictor / reducing MAX_PERIOD will lower quality/bitrate
>> only at fairly low bitrates, but it will save a fair amount of memory
>> and lower complexity.

More information about the celt-dev mailing list