Speeding up a Lattice Boltzmann Kernel on nVIDIA GPUs

Habich J, Zeiser T, Hager G, Wellein G (2009)

Publication Type: Conference contribution

Publication year: 2009

Publisher: Civil-Comp Press

Edited Volumes: Civil-Comp Proceedings

City/Town: Kippen, Stirlingshire, United Kingdom

Pages Range: 17

Conference Proceedings Title: Proceedings of the First International Conference on Parallel, Distributed and Grid Computing for Engineering

Event location: Pécs, Hungary HU

ISBN: 978-1-905088-29-4

Abstract

This paper presents implementation strategies and optimization approaches for a lattice Boltzmann flow solver on nVIDIA graphics processing units (GPUs). We outline the GPU-specific implementation of the flow solver, potential problems that occur and discuss how to avoid or at least ameliorate them. It is shown that ten times the performance of a quad core based two socket server can be achieved with a single graphics card. Furthermore we employ STREAM benchmarks to evaluate achievable performance and to establish programming guidelines for applications. With regard to hybrid computing the attainable bandwidth of different interfaces involved is evaluated. © Civil-Comp Press, 2009.

Authors with CRIS profile

Johannes Habich Regionales Rechenzentrum Erlangen (RRZE) Thomas Zeiser Regionales Rechenzentrum Erlangen (RRZE) Georg Hager Regionales Rechenzentrum Erlangen (RRZE) Gerhard Wellein Professur für Höchstleistungsrechnen

How to cite

APA:

Habich, J., Zeiser, T., Hager, G., & Wellein, G. (2009). Speeding up a Lattice Boltzmann Kernel on nVIDIA GPUs. In Proceedings of the First International Conference on Parallel, Distributed and Grid Computing for Engineering (pp. 17). Pécs, Hungary, HU: Kippen, Stirlingshire, United Kingdom: Civil-Comp Press.

MLA:

Habich, Johannes, et al. "Speeding up a Lattice Boltzmann Kernel on nVIDIA GPUs." Proceedings of the PARENG2009, Pécs, Hungary Kippen, Stirlingshire, United Kingdom: Civil-Comp Press, 2009. 17.

BibTeX: Download