Ray decomposition radiation transport for high performance computing

Abstract

Radiation transport is essential in many high-performance computing problems. However, its complexity presents computational challenges. This study presents a novel algorithm, the ray decomposition method for long characteristics transport, designed to address communication challenges specific to distributed memory computing. Reordering of ray property calculations reduces communication cost associated with sequential ray integration. Verification studies demonstrate solution convergence. Performance modeling of the ray decomposition method predicts the compute time from first principles. Consistency of experimentally measured performance with analytical predictions validates the performance scaling model. This work represents a step towards more scalable and efficient radiation transport simulations.