hobold (hobold@vectorizer.org) on 8/23/10 wrote:
---------------------------
>Steve Underwood (steveu@coppice.org) on 8/23/10 wrote:
>---------------------------
>[...]
>>If they start showing worthwhile gains for problems involving numerous
>>short vectors (e.g. the endless 40 element vectors in the G.729 speech codec), I
>>might start to get really interested in them.
>>
>Performance in terms of low latency is somewhat at odds with the general idea behind
>massively data parallel processors. I doubt that GPU designers will shift the balance
>towards lower latency. Their primary target is still graphics, with all its embarrassing
>parallelism. Real time audio codecs just don't have large chunks of data available
>to work with at any given point in time.

Latency is not necessarily a killer. If you are using a GPU to process numerous concurrent media streams, you should be able to keep many small vectors in flight. Last time I looked it seemed that didn't work out too well with the nVidia chips.

Steve