parallel Radiance

Hi,
I wanted to get some information on how parallel rendering is implemented in Radiance. For example, does it support OpenMP, MPI or both? Will it take advantage of SIMD if available? any pointers to design or architecture documentation would be very helpful.

Bruce

Hi Bruce,

There is the parallel rendering chapter in "Rendering with Radiance," and the following short write-up:

  http://radsite.lbl.gov/radiance/refer/Notes/parallel.html

As with most of the rest of Radiance, add-on libraries are avoided, so no OpenMP, MPI, etc. is used. Also, making ray-tracing work with SIMD is notoriously difficult, so parallelization is course-grained.

I can e-mail you the parallel rendering chapter if you don't have the book.

Cheers,
-Greg

···

From: "Bruce D'Amora" <[email protected]>
Date: January 17, 2011 1:11:17 PM PST

Hi,
I wanted to get some information on how parallel rendering is implemented in Radiance. For example, does it support OpenMP, MPI or both? Will it take advantage of SIMD if available? any pointers to design or architecture documentation would be very helpful.

Bruce

Hi!

There used to be a modified implementation of Radiance using MPI. I never used it, and it was based on a release that is pretty much outdated by now. The code modifications together with a paper titled "MPI-parallelized Radiance on SGI CoW and SMP" were published by Roland Koholka, Heinz Mayer and Alois Goller in 1999.

Most critical tools in Radiance supports coordinated processing on multiple cores by now. This is clearly aiming at platforms providing several CPUs/cores per node, not at SIMD. I am not aware of any attempt to port the code to GPUs or other specialized cores.

Cheers, Lars.