[Vorbis] encoding vorbis in windows using on multi-core cpu

[Oscar Sundbom]

Patrick Byrne skrev:
> > You may be interested in these threads:
> > http://osdir.com/ml/video.thoggen.devel/2006-11/msg00008.html
> > http://lists.xiph.org/pipermail/theora/2006-October/001298.html
> >
> > It sounds as though they have done this for Dirac, but I have
> > struggled to find real enthusiasm for it in Theora, where there are
> > really serious potential gains.
> >
> > Do you have one really big mono file to encode at a time? If you have
> > more than one, couldn't you just do two at once?
>
> Thanks very much!
>
> We are encoding gigabytes of mono speech - using multiple cores could 
> cut quite a lot of time off our asset builds.
>
> tbh, it wouldn't be that much work to change our audio compiler so it 
> runs 2 compiles at the same time, but it would save doing the work if 
> the encode dll took care of all that.
>
> ..of course, the audio compiler would need modifying _again_ for a 
> quad core machine - we need a new paradigm, dammit!
>
>
> Cheers
> Patrick Byrne
>


Hmm. Apparently, this seems like a quite difficult task (see to Monty's 
post).
I've two ideas when it comes to your problem.
If you're encoding gigabyte sized files, you could always slice them up 
(say in two halves, for a dual core machine) and encode them in 
parallel. Then you could tack the two vorbis streams together into a 
single file.
If the files themselves aren't that huge, you could always encode more 
than one at a time, utilising as many cores/processors as you've got.

Not as sexy, I know, but should work right now. :)

Regards,
Oscar