Am 26.10.2012 01:14, schrieb Chris Cason:
> The Parallela platform seems to have potential in the future for use with raytracing
> as Adapteva are coming out with a double-precision version early in 2013. Given
their
> kickstarter is approaching its end date and they are still a little bit short of
their
> target we thought we'd mention this in case any POV users are interested in it.
>
> http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone
>
> Once the appropriate hardware is available the POV team will take a close look at it
> to see if it's a suitable target for a port of our software.

 From what I've glimpsed from the specs, the individual "crunching 
cores" can't directly access any memory worth mentioning - IIRC it's 
less than a MByte per core. Other cores' memory appears to be 
addressable as well, but with some latency, and it still gives you no 
more than a few MByte.

To me that doesn't sound like an ideal architecture for POV-Ray (or 
raytracing in general) - future improvements to CUDA or OpenCL sound 
like a more promising thing.

Post a reply to this message