[Asterisk-Dev] GPU backending for codecs

[Nick Bachmann]

Howdy all-

There's been some recent discussion here on -dev and on IRC about adding 
support to Asterisk for offloading codec processing onto GPUs, which 
have the potential to do coding much more efficiently than CPUs.  
Recently, I've been getting more interested in this; however, I've never 
done any coding work with either GPUs or codecs, so I'm posting this 
request for help here:

Does anybody out there have the know-how and time to lead (or at least 
help get some of us started) the implementation of this?  If anybody is 
willing to take up the challenge, I'd be willing to donate time coding, 
testing, documentation, or whatever kind of grunt work they want :).  
I've also talked to a few people who, like me, don't know how to get 
started but would be willing to help once they knew what to do. I'm also 
sure that there would be quite a few people interested in putting up a 
bounty for this.  Although I doubt any bounty would cover what this 
project would cost to hire somebody to implement, it's at least better 
than just the undying gratitude of users everywhere...

So far, the best tool I've come across for this project is libsh, which 
has the advantage of working with several GPU architectures while still 
being able to compile for CPUs, so it seems possible to use the exact 
same code no matter what the processor backend.

Please, if you're interested in helping with any facet of this, contact 
me.  Email or IRC work best, but I'm open to Morse code and smoke 
signals if you can help :)!

Nick
(IRC hermie)