Towards unification of accelerated computing and interconnection for extreme-scale computing

Toshihiro Hanawa; Yuetsu Kodama; Taisuke Boku; Hideharu Amano; Hitoshi Murai; Masayuki Umemura; Mitsuhisa Sato

doi:10.1007/978-3-319-16214-0_43

Towards unification of accelerated computing and interconnection for extreme-scale computing

Toshihiro Hanawa, Yuetsu Kodama, Taisuke Boku, Hideharu Amano, Hitoshi Murai, Masayuki Umemura, Mitsuhisa Sato

Department of Information and Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Heterogeneous clusters using accelerators are widely used for high-performance computing system. In such systems, the inter-node communication among accelerators becomes bottleneck due to the data transfer between the accelerator and the host. To eliminate this overhead, we have been developing a novel communication system realizing direct communication among accelerators over computation nodes under the HA-PACS (Highly Accelerated Parallel Advanced system for Computational Sciences) project. Also we are investigating high-level parallel programming language, and several practical application programs on our concept, as well as studying the enhancement of TCA and developing system software stack in the CREST project.

Original language	English
Title of host publication	Applied Reconfigurable Computing - 11th International Symposium, ARC 2015, Proceedings
Editors	Pedro C. Diniz, Kentaro Sano, Michael Hübner, Dimitrios Soudris
Publisher	Springer Verlag
Pages	463-474
Number of pages	12
ISBN (Electronic)	9783319162133
DOIs	https://doi.org/10.1007/978-3-319-16214-0_43
Publication status	Published - 2015 Jan 1
Event	11th International Symposium on Applied Reconfigurable Computing, ARC 2015 - Bochum, Germany Duration: 2015 Apr 13 → 2015 Apr 17

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	9040
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	11th International Symposium on Applied Reconfigurable Computing, ARC 2015
Country	Germany
City	Bochum
Period	15/4/13 → 15/4/17

Fingerprint

Keywords

Accelerator computing
GPU direct
Interconnection network
PCI express
Remote DMA

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Cite this

Hanawa, T., Kodama, Y., Boku, T., Amano, H., Murai, H., Umemura, M., & Sato, M. (2015). Towards unification of accelerated computing and interconnection for extreme-scale computing. In P. C. Diniz, K. Sano, M. Hübner, & D. Soudris (Eds.), Applied Reconfigurable Computing - 11th International Symposium, ARC 2015, Proceedings (pp. 463-474). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9040). Springer Verlag. https://doi.org/10.1007/978-3-319-16214-0_43

Towards unification of accelerated computing and interconnection for extreme-scale computing. / Hanawa, Toshihiro; Kodama, Yuetsu; Boku, Taisuke; Amano, Hideharu; Murai, Hitoshi; Umemura, Masayuki; Sato, Mitsuhisa.

Applied Reconfigurable Computing - 11th International Symposium, ARC 2015, Proceedings. ed. / Pedro C. Diniz; Kentaro Sano; Michael Hübner; Dimitrios Soudris. Springer Verlag, 2015. p. 463-474 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9040).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Hanawa, T, Kodama, Y, Boku, T, Amano, H, Murai, H, Umemura, M & Sato, M 2015, Towards unification of accelerated computing and interconnection for extreme-scale computing. in PC Diniz, K Sano, M Hübner & D Soudris (eds), Applied Reconfigurable Computing - 11th International Symposium, ARC 2015, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9040, Springer Verlag, pp. 463-474, 11th International Symposium on Applied Reconfigurable Computing, ARC 2015, Bochum, Germany, 15/4/13. https://doi.org/10.1007/978-3-319-16214-0_43

Hanawa T, Kodama Y, Boku T, Amano H, Murai H, Umemura M et al. Towards unification of accelerated computing and interconnection for extreme-scale computing. In Diniz PC, Sano K, Hübner M, Soudris D, editors, Applied Reconfigurable Computing - 11th International Symposium, ARC 2015, Proceedings. Springer Verlag. 2015. p. 463-474. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-16214-0_43

Hanawa, Toshihiro ; Kodama, Yuetsu ; Boku, Taisuke ; Amano, Hideharu ; Murai, Hitoshi ; Umemura, Masayuki ; Sato, Mitsuhisa. / Towards unification of accelerated computing and interconnection for extreme-scale computing. Applied Reconfigurable Computing - 11th International Symposium, ARC 2015, Proceedings. editor / Pedro C. Diniz ; Kentaro Sano ; Michael Hübner ; Dimitrios Soudris. Springer Verlag, 2015. pp. 463-474 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{bf1bd27c55034d3cb86d1e4864ccf8b3,

title = "Towards unification of accelerated computing and interconnection for extreme-scale computing",

abstract = "Heterogeneous clusters using accelerators are widely used for high-performance computing system. In such systems, the inter-node communication among accelerators becomes bottleneck due to the data transfer between the accelerator and the host. To eliminate this overhead, we have been developing a novel communication system realizing direct communication among accelerators over computation nodes under the HA-PACS (Highly Accelerated Parallel Advanced system for Computational Sciences) project. Also we are investigating high-level parallel programming language, and several practical application programs on our concept, as well as studying the enhancement of TCA and developing system software stack in the CREST project.",

keywords = "Accelerator computing, GPU direct, Interconnection network, PCI express, Remote DMA",

author = "Toshihiro Hanawa and Yuetsu Kodama and Taisuke Boku and Hideharu Amano and Hitoshi Murai and Masayuki Umemura and Mitsuhisa Sato",

year = "2015",

month = jan

day = "1",

doi = "10.1007/978-3-319-16214-0_43",

language = "English",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "463--474",

editor = "Diniz, {Pedro C.} and Kentaro Sano and Michael H{\"u}bner and Dimitrios Soudris",

booktitle = "Applied Reconfigurable Computing - 11th International Symposium, ARC 2015, Proceedings",

note = "11th International Symposium on Applied Reconfigurable Computing, ARC 2015 ; Conference date: 13-04-2015 Through 17-04-2015",

}

TY - GEN

T1 - Towards unification of accelerated computing and interconnection for extreme-scale computing

AU - Hanawa, Toshihiro

AU - Kodama, Yuetsu

AU - Boku, Taisuke

AU - Amano, Hideharu

AU - Murai, Hitoshi

AU - Umemura, Masayuki

AU - Sato, Mitsuhisa

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Heterogeneous clusters using accelerators are widely used for high-performance computing system. In such systems, the inter-node communication among accelerators becomes bottleneck due to the data transfer between the accelerator and the host. To eliminate this overhead, we have been developing a novel communication system realizing direct communication among accelerators over computation nodes under the HA-PACS (Highly Accelerated Parallel Advanced system for Computational Sciences) project. Also we are investigating high-level parallel programming language, and several practical application programs on our concept, as well as studying the enhancement of TCA and developing system software stack in the CREST project.

AB - Heterogeneous clusters using accelerators are widely used for high-performance computing system. In such systems, the inter-node communication among accelerators becomes bottleneck due to the data transfer between the accelerator and the host. To eliminate this overhead, we have been developing a novel communication system realizing direct communication among accelerators over computation nodes under the HA-PACS (Highly Accelerated Parallel Advanced system for Computational Sciences) project. Also we are investigating high-level parallel programming language, and several practical application programs on our concept, as well as studying the enhancement of TCA and developing system software stack in the CREST project.

KW - Accelerator computing

KW - GPU direct

KW - Interconnection network

KW - PCI express

KW - Remote DMA

UR - http://www.scopus.com/inward/record.url?scp=84926661910&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84926661910&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-16214-0_43

DO - 10.1007/978-3-319-16214-0_43

M3 - Conference contribution

AN - SCOPUS:84926661910

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 463

EP - 474

BT - Applied Reconfigurable Computing - 11th International Symposium, ARC 2015, Proceedings

A2 - Diniz, Pedro C.

A2 - Sano, Kentaro

A2 - Hübner, Michael

A2 - Soudris, Dimitrios

PB - Springer Verlag

T2 - 11th International Symposium on Applied Reconfigurable Computing, ARC 2015

Y2 - 13 April 2015 through 17 April 2015

ER -

Access to Document

10.1007/978-3-319-16214-0_43