Hi, I am looking for some help regarding non-convex radiation computing. I have heard that there are some issues running these models on parallel processors but I can't find any information on it! I'm very new to the field so any help at all would be much appreciated. Many thanks, Sile

Sile,

It very much depends on your solution method.

For ray-trace (Monte-Carlo or DTM) parallelization is trivial, since each ray is independent of other rays. Consequently we reached nearly 100% parallel efficiency with 16 CPUs (Cray and SGI Origin).

With DOM, FVM etc., the parallel efficiency depends a lot on your base method and the parallelization method. Please see our paper (and references therein)

J. Tal, R. Ben-Zvi, A. Kribus, A High-Efficiency Parallel Solution of the Radiative Transfer Equation, Numerical Heat Transfer, Part B, Vol. 44 No. 3, pp. 295-308, 2003.

BTW: 1. What do you mean by "non-convex radiation computing"? 2. Are you considering inter-surface radiation or radiation in participating medium?

Rami,

Thank you very much for your reply - it is much appreciated. I'll check out that reference.

I am considering inter-surface radiation. The term non-convex was used in the project description I was given. My understanding of it in the context of the project(although I may be wrong) is that it refers to inter-surface radiation on concave surfaces, i.e. non-flat or convex surfaces. Thanks again! Sile