[Radiance-general] rtcontrib problems
Greg Ward
gregoryjward at gmail.com
Tue Jun 12 10:23:47 PDT 2012
Hi David,
Thanks for testing this out! It's great to get the feedback. I've run my own (similar) tests and things look OK, but I haven't checked the data thoroughly, and I need to check corner cases, other operation modes, recovery, etc.
I just checked in some more changes that improve efficiency when you have a lot of processes. I tried it with 24 processes on my Mac Pro, and it went from 85% utilization to about 99% with these new changes. I don't think I ever got better than 20% utilization with the original code on as many cores.
Cheers,
-Greg
> From: David Geisler-Moroder <david.moroder at gmail.com>
> Date: June 12, 2012 9:10:41 AM PDT
>
> Hi Greg,
>
> thanks a lot for your efforts in improving rtcontrib -- this makes a great tool even more valuable
>
> I just ran genBSDF for a specular lamella system generating a BSDF with full Klems representation. To compare rcontrib to rtcontrib I just replaced the
> respective calls in the genBSDF perl script. I compared the resulting BSDFs just by quick simulations of diffuse-diffuse transmissions and reflections
> as well as diffuse transmissions and reflections from a CIE overcast sky, a diffuse sky and a diffuse ground. Modulo MC noise the result seem to be
> perfectly the same.
>
> Using 8 cores the simulation times were appoximately 185min for the existing version with rtcontrib and 104 min with the new rcontrib tool!!
>
> Cheers,
> David
>
>
>
> 2012/6/11 Greg Ward <gregoryjward at gmail.com>
> Coincidentally, I'm working now to improve the speed & efficiency of rtcontrib, especially for multi-processors. If any Unix users are interested in trying it out, I've checked a working version into the HEAD. The program is (temporarily) named rcontrib, to segregate it from its predecessor, but the options and operation should be identical. Once it's fully tested, I plan to replace the existing rtcontrib with this new version, which has been substantially redesigned.
>
> Cheers,
> -Greg
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.radiance-online.org/pipermail/radiance-general/attachments/20120612/57a5b966/attachment.html>
More information about the Radiance-general
mailing list