Predictive Modeling and Analysis of OP2 on Distributed Memory GPU Clusters
Oxford e-Research Centre, University of Oxford
2nd International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computing Systems (PMBS 11), 2011
@article{mudalige2011predictive,
title={Predictive Modeling and Analysis of OP2 on Distributed Memory GPU Clusters},
author={Mudalige, GR and Giles, MB and Bertolli, C. and Kelly, PH},
journal={SIGMETRICS Performance Evaluation Review},
volume={40},
number={2},
year={2011}
}
OP2 is an "active" library framework for the development and solution of unstructured mesh-based applications. It aims to decouple the scientific specification of an application from its parallel implementation to achieve code longevity and near-optimal performance through re-targeting the backend to different multi-core/many-core hardware. This paper presents a summary of a predictive performance analysis and benchmarking study of OP2 on heterogeneous cluster systems. In this work, an industrial representative CFD application written using the OP2 framework is benchmarked during the solution of an unstructured mesh of 1.5M and 26M edges. Benchmark systems include a large-scale Cray XE6 system and an Intel Westmere/InfiniBand cluster. Performance modeling is then used to predict the application’s performance on an NVIDIA Tesla C2070-based GPU cluster, enabling the comparison of OP2’s performance capabilities on emerging distributed memory heterogeneous systems. Results illustrate the performance benefits that can be gained through many-core solutions both on single-node and heterogeneous configurations in comparison to traditional homogeneous cluster systems for this class of application.
November 19, 2011 by hgpu
