Nuclear fusion simulation code optimization and performance evaluation on GPU cluster

Fujita, Norihisa*; Nuga, Hideo*; Boku, Taisuke*; Idomura, Yasuhiro

GT5D is a nuclear fusion simulation program which aims to analyze the turbulence phenomena in tokamak plasma. In this research, we optimize it for GPU clusters with multiple GPUs on a node. Based on the profile result of GT5D on a CPU node, we decide to offload the whole of the time development part of the program to GPUs except MPI communication. Four NVIDIA M2090 GPUs get 3.35 times faster performance in maximum in function level evaluation, and 1.91 times faster performance in total than two Intel Xeon E5-2670 (SandyBridge) CPU with 8 cores. It includes 63% performance gain with overlapping the communication between MPI processes with GPU calculation.

Language	:	English
Journal	:	Proceedings of 2014 IEEE 28th International Parallel & Distributed Processing Symposium Workshops (Internet)
Volume	:
Number	:
Pages	:	p.1266 - 1274
Publication Year/Month	:	2014/12
Meeting title	:	15th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2014) / 28th IEEE International Parallel & Distributed Processing Symposium (IPDPS 2014)
Held date	:	2014/05
Location (city)	:	Phoenix
Location (country)	:	U.S.A.
Patent information	:
PDF	:
Paper URL	:	https://doi.org/10.1109/IPDPSW.2014.142
DOI for research data	:	Link to research data related to this paper/report.
Keywords	:	no keyword
Research Facility	:
Press Release	:
JAEA R&D Navigator	:
Cooperating Institute	:

Accesses	:	- Accesses
Web of Science® Times Cited Count	:	Times Cited Count： If you would like to get the latest times cited, please access the Web of Science®. http://www.webofknowledge.com/wos
InCites™	:	Percentile:27.78 Category:Computer Science, Theory & Methods
Altmetrics	:

Registration No. : BB20133394
JAEA Abstracts No. : 43000457
Paper Submission No. : 15595

[CLARIVATE ANALYTICS], [WEB OF SCIENCE], [HIGHLY CITED PAPER & CUP LOGO] and [HOT PAPER & FIRE LOGO] are trademarks of Clarivate Analytics, and/or its affiliated company or companies, and used herein by permission and/or license.