Scaling Results for a Discontinuous Galerkin Finite-Element Wave Solver on Multi-GPU Systems
University of Basel, Department of Mathematics and Computer Science, Klingelbergstrasse 50, CH-4056 Basel, Switzerland
Workshop on Parallel Systems and Algorithms (PARS), 2011
@article{rietmann2011scaling,
title={Scaling Results for a Discontinuous Galerkin Finite-Element Wave Solver on Multi-GPU Systems},
author={Rietmann, M. and Schenk, O. and Burkhart, H.},
year={2011}
}
Among the various techniques for solving hyperbolic partial differential equations with inhomogeneous, irregularly-shaped domains, a relatively new type of finite element method has grown in popularity because of its flexibility and scalability across many parallel cores. Discont inuous Galerkin (DG) methods have shown themselves to be an effective scheme for the simulation of wave-propagation problems commonly arising in computational seismology. This paper presents performance results of a DG solver using one and more CUDA-based graphics processors (GPUs) and challenges associated with scaling the simulat ion across multiple GPUs.
November 26, 2011 by hgpu
