[asterisk-dev] Using the Video Hardware as a DSP?

[Eric "ManxPower" Wieling]

Haider Raza wrote:
> Hi,
> 
>     I have only recently started using Asterisk and after doing a couple 
> of stress tests I noticed choppy phone calls. Anyway, back in the days 
> of 386 PCs I remember that using DMA to do memory transfers to and from 
> the sound card was a big thing. Every game did it. This method reduced 
> the CPU load which translated into higher frame count. I also know that 
> nowadays Video Hardware has become quite powerfull. There are cards with 
> 256 bit Buses and high speed memory in the ranges of 256MB and 512MB. I 
> was wondering how hard would it be to code (either using OpenGL or 
> directly using assembler) a routine to compress / decompress audio using 
> the video hardware. Somehow placing the data in a 3d World Matrix and 
> having all the data displaced or "translated" into a compressed / 
> decompressed "frame" of audio, which could later be played back by to 
> appropriate user (remote or local). If this could be done even the 
> sloppiest video hardware could work as a powerful DSP.
> 
>    Or not just the Video hardware but also the Audio hardware. There are 
> some pretty good audio cards out there that could do the work of 
> expensive DSPs.
> 
> Any thoughts?

There are MANY things that can cause choppy phone calls, only one of 
these issues is related to CPU usage.

Other things that can cause choppy audio:
   VAD
   IRQ Sharing
   Network Jitter
   Poor Jitter Buffer Implementation


-- 
Now accepting new clients in Birmingham, Atlanta, Huntsville, 
Chattanooga, and Montgomery.