Performance evaluation of magnetohydrodynamics simulation for magnetosphere on K computer

Keiichiro Fukazawa, Takeshi Nanri, Takayuki Umeda

研究成果: 著書/レポートタイプへの貢献会議での発言

2 引用 (Scopus)

抄録

To use the massively parallel scalar type supercomputer for the large scale numerical magnetospheric simulation, we made performance evaluation of a three-dimensional Magnetohydrodynamic (MHD) simulation code for space plasma simulations of planetary magnetosphere on the K computer with 262,144 cores at RIKEN. For parallelization of the MHD code, we use four different methods, i.e., regular one-dimensional, two-dimensional, three-dimensional domain decomposition methods and a cache-hit type of three-dimensional domain decomposition method. We found that the onedimensional decomposition and cache-hit type of three-dimensional decomposition achieves the best performance and our MHD code can scale over 200,000 processes on the K computer.

元の言語英語
ホスト出版物のタイトルAsiaSim 2013 - 13th International Conference on Systems Simulation, Proceedings
出版者Springer Verlag
ページ570-576
ページ数7
ISBN(印刷物)9783642450365
DOI
出版物ステータス出版済み - 1 1 2013
イベント13th International Conference on Systems Simulation, AsiaSim 2013 - , シンガポール
継続期間: 11 6 201311 8 2013

出版物シリーズ

名前Communications in Computer and Information Science
402
ISSN(印刷物)1865-0929

その他

その他13th International Conference on Systems Simulation, AsiaSim 2013
シンガポール
期間11/6/1311/8/13

Fingerprint

Magnetosphere
Magnetohydrodynamics
Domain decomposition methods
Plasma simulation
Decomposition
Supercomputers
Computer simulation

All Science Journal Classification (ASJC) codes

  • Computer Science(all)
  • Mathematics(all)

これを引用

Fukazawa, K., Nanri, T., & Umeda, T. (2013). Performance evaluation of magnetohydrodynamics simulation for magnetosphere on K computer. : AsiaSim 2013 - 13th International Conference on Systems Simulation, Proceedings (pp. 570-576). (Communications in Computer and Information Science; 巻数 402). Springer Verlag. https://doi.org/10.1007/978-3-642-45037-2_61

Performance evaluation of magnetohydrodynamics simulation for magnetosphere on K computer. / Fukazawa, Keiichiro; Nanri, Takeshi; Umeda, Takayuki.

AsiaSim 2013 - 13th International Conference on Systems Simulation, Proceedings. Springer Verlag, 2013. p. 570-576 (Communications in Computer and Information Science; 巻 402).

研究成果: 著書/レポートタイプへの貢献会議での発言

Fukazawa, K, Nanri, T & Umeda, T 2013, Performance evaluation of magnetohydrodynamics simulation for magnetosphere on K computer. : AsiaSim 2013 - 13th International Conference on Systems Simulation, Proceedings. Communications in Computer and Information Science, 巻. 402, Springer Verlag, pp. 570-576, 13th International Conference on Systems Simulation, AsiaSim 2013, シンガポール, 11/6/13. https://doi.org/10.1007/978-3-642-45037-2_61
Fukazawa K, Nanri T, Umeda T. Performance evaluation of magnetohydrodynamics simulation for magnetosphere on K computer. : AsiaSim 2013 - 13th International Conference on Systems Simulation, Proceedings. Springer Verlag. 2013. p. 570-576. (Communications in Computer and Information Science). https://doi.org/10.1007/978-3-642-45037-2_61
Fukazawa, Keiichiro ; Nanri, Takeshi ; Umeda, Takayuki. / Performance evaluation of magnetohydrodynamics simulation for magnetosphere on K computer. AsiaSim 2013 - 13th International Conference on Systems Simulation, Proceedings. Springer Verlag, 2013. pp. 570-576 (Communications in Computer and Information Science).
@inproceedings{ba37e9f387db4a22bf1042f5c431e538,
title = "Performance evaluation of magnetohydrodynamics simulation for magnetosphere on K computer",
abstract = "To use the massively parallel scalar type supercomputer for the large scale numerical magnetospheric simulation, we made performance evaluation of a three-dimensional Magnetohydrodynamic (MHD) simulation code for space plasma simulations of planetary magnetosphere on the K computer with 262,144 cores at RIKEN. For parallelization of the MHD code, we use four different methods, i.e., regular one-dimensional, two-dimensional, three-dimensional domain decomposition methods and a cache-hit type of three-dimensional domain decomposition method. We found that the onedimensional decomposition and cache-hit type of three-dimensional decomposition achieves the best performance and our MHD code can scale over 200,000 processes on the K computer.",
author = "Keiichiro Fukazawa and Takeshi Nanri and Takayuki Umeda",
year = "2013",
month = "1",
day = "1",
doi = "10.1007/978-3-642-45037-2_61",
language = "English",
isbn = "9783642450365",
series = "Communications in Computer and Information Science",
publisher = "Springer Verlag",
pages = "570--576",
booktitle = "AsiaSim 2013 - 13th International Conference on Systems Simulation, Proceedings",
address = "Germany",

}

TY - GEN

T1 - Performance evaluation of magnetohydrodynamics simulation for magnetosphere on K computer

AU - Fukazawa, Keiichiro

AU - Nanri, Takeshi

AU - Umeda, Takayuki

PY - 2013/1/1

Y1 - 2013/1/1

N2 - To use the massively parallel scalar type supercomputer for the large scale numerical magnetospheric simulation, we made performance evaluation of a three-dimensional Magnetohydrodynamic (MHD) simulation code for space plasma simulations of planetary magnetosphere on the K computer with 262,144 cores at RIKEN. For parallelization of the MHD code, we use four different methods, i.e., regular one-dimensional, two-dimensional, three-dimensional domain decomposition methods and a cache-hit type of three-dimensional domain decomposition method. We found that the onedimensional decomposition and cache-hit type of three-dimensional decomposition achieves the best performance and our MHD code can scale over 200,000 processes on the K computer.

AB - To use the massively parallel scalar type supercomputer for the large scale numerical magnetospheric simulation, we made performance evaluation of a three-dimensional Magnetohydrodynamic (MHD) simulation code for space plasma simulations of planetary magnetosphere on the K computer with 262,144 cores at RIKEN. For parallelization of the MHD code, we use four different methods, i.e., regular one-dimensional, two-dimensional, three-dimensional domain decomposition methods and a cache-hit type of three-dimensional domain decomposition method. We found that the onedimensional decomposition and cache-hit type of three-dimensional decomposition achieves the best performance and our MHD code can scale over 200,000 processes on the K computer.

UR - http://www.scopus.com/inward/record.url?scp=84893877070&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84893877070&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-45037-2_61

DO - 10.1007/978-3-642-45037-2_61

M3 - Conference contribution

AN - SCOPUS:84893877070

SN - 9783642450365

T3 - Communications in Computer and Information Science

SP - 570

EP - 576

BT - AsiaSim 2013 - 13th International Conference on Systems Simulation, Proceedings

PB - Springer Verlag

ER -