Code Optimization and Performance Analysis of Oceanographic Software Package NEMO for GPGPU Systems
Computer Systems Department, Technical University of Sofia, 1000 Sofia, Bulgaria
The 2014 International Conference on Circuits, Systems, Signal Processing, Communications and Computers (CSSCC’14), 2014
@article{borovska2014code,
title={Code Optimization and Performance Analysis of Oceanographic Software Package NEMO for GPGPU Systems},
author={Borovska, Plamenka and Ivanova, Desislava},
year={2014}
}
The paper presents our experience in code optimization and performance analysis of software package NEMO on hybrid parallel computer systems with accelerators. NEMO Ocean is a software package for oceanology, simulating ocean gyres and sea-ice models. Code optimization and performance analysis are performed for the case study of NEMO ORCA2_LIM configuration. All experiments are conducted on a heterogeneous IBM Blade Center comprising Blade Servers and NVidia Tesla M2075 accelerators. The parallel software for oceanology NEMO has been optimized for GPGPU systems. For this purpose an optimization has been implemented and verified. The experimental results show that the optimized NEMO code provides considerably better performance than the original code.
May 12, 2014 by hgpu
