Evaluating Novel Load Balancing Algorithms for Particle Advection Toward Integration in VisIt's Infrastructure

Robert Sisneros, University of Illinois at Urbana-Champaign

Usage Details

Foundational visualization algorithms are central to the production of visualization tools running at computing centers around the world but consume tremendous amounts of limited resources. We believe that understanding the performance characteristics of these algorithms is critical in being good stewards of those unique resources. In this project, we investigate the foundational algorithm for parallel particle advection by focusing on a parameter sweep of the de facto standard algorithm "Parallelize Over Data." Our work has led to several discoveries: a glaring and resolvable issue in poor default values, identification of the best choices of parameters for future tests, and the existence of viable testing alternatives that save time and resources. Our work sets forth a framework for applying solutions to a practical and often overlooked area in computer visualization.