[CELT-dev] A few build problems
jmvalin at jmvalin.ca
Mon Aug 15 18:10:26 PDT 2011
On 11-08-14 11:07 AM, Chris Weiland wrote:
> First, this application is cross platform, and my primary development
> environment is Windows. However, there is no config.h included with
> CELT, so I seem to have to make my own. I managed to get it to build
> after a lot of trial and error, but I'm not positive if I've
> configured everything exactly properly. I guess this is more of a
> complaint/request that an "official" config.h for Windows be included
> in future releases.
Yeah, that's on the TODO list. In general, the main thing is that you'll
#define inline __inline
or something like this.
> Second, I started this project using Speex. I don't see any reason to
> get rid of Speex support (at least for now), so I'd like to keep that
> option. However, I can't link both Speex and CELT at the same time
> because the kiss_fft function is defined in both libraries, which
> causes the linker to be confused. I thought that I could simply
> remove one, but it looks like each implementation is very different
> and each library depends on its own implementation. Is there any way
> around this?
I just renamed the kiss-fft functions in the Opus code to avoid this.
Though it will not be updated in the CELT repo.
> Lastly, documentation is quite sparse for some of the parameters that
> I'd like to play around with. The biggest one is latency. The demo
> page shows a comparison of different latencies that CELT can run at,
> but I don't see anything in the API that has anything to do with that.
> Do I simply adjust the frame size? Also, what is input clipping?
> And what exactly does "complexity" control? Does it have any effect
> on latency or bit rate? How much effect on CPU usage does varying the
> complexity make?
The latency is simply (frame_size+2.5ms) for CELT and (frame_size+5ms)
for Opus (though there will be a +2.5 option). As for the complexity
parameter, I just suggest you play around with it.
More information about the celt-dev