Chaotic Iterations for Radiation Transport

A new splitting method for discrete ordinates, radiation transport problems will be introduced. The problem we focus on requires solving a large linear system in parallel using an iterative solver. To increase parallel efficiency, it is common to split the linear operator and lag data at the price of slower convergence. For the chaotic splitting method introduced in this talk, a possibly different splitting is used at every iteration depending on the asynchronous communication between processes. Numerical results will be presented showing the efficacy of the new method, and a proof will be presented showing the chaotic iterations method converges in many regimes..

BIOGRAPHY: Kris Garrett received a PhD in mathematics at the University of Texas at Arlington. Afterwards, he did a postdoc at Oak Ridge National Laboratory followed by a postdoc at Los Alamos National Laboratory. He is currently a staff scientist at Los Alamos National Laboratory working on computational performance of discrete ordinates radiation transport codes. He is also a co-lead for the Parallel Computing Summer Research Internship at Los Alamos National Laboratory.

Event website