Deep learning on high performance FPGA switching boards: Flow-in-cloud

Kazusa Musha; Tomohiro Kudoh; Hideharu Amano

doi:10.1007/978-3-319-78890-6_4

Deep learning on high performance FPGA switching boards: Flow-in-cloud

Kazusa Musha, Tomohiro Kudoh, Hideharu Amano

Department of Information and Computer Science

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

7 Citations (Scopus)

Abstract

FiC (Flow-in-Cloud)-SW is an FPGA-based switching node for an efficient AI computing system. It is equipped with a number of serial links directly connected to other nodes. Unlike other multi-FPGA systems, the circuit switching fabric with the STDM (Static Time Division Multiplexing) is implemented on the FPGA for predictable communication and cost-efficient data broadcasting. Parallel convolution modules for AlexNet are implemented on FiC-SW1 prototype boards consisting of Kintex Ultrascale FPGA, and evaluation results show that the parallel execution with 20 boards achieved 4.6 times better performance than the state of art implementation on a single Virtex 7 FPGA board.

Original language	English
Title of host publication	Applied Reconfigurable Computing
Subtitle of host publication	Architectures, Tools, and Applications - 14th International Symposium, ARC 2018, Proceedings
Editors	Nikolaos Voros, Georgios Keramidas, Christos Antonopoulos, Michael Huebner, Pedro C. Diniz, Diana Goehringer
Publisher	Springer Verlag
Pages	43-54
Number of pages	12
ISBN (Print)	9783319788890
DOIs	https://doi.org/10.1007/978-3-319-78890-6_4
Publication status	Published - 2018
Event	14th International Symposium on Applied Reconfigurable Computing, ARC 2018 - Santorini, Greece Duration: 2018 May 2 → 2018 May 4

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10824 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	14th International Symposium on Applied Reconfigurable Computing, ARC 2018
Country/Territory	Greece
City	Santorini
Period	18/5/2 → 18/5/4

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-319-78890-6_4

Cite this

Musha, K., Kudoh, T., & Amano, H. (2018). Deep learning on high performance FPGA switching boards: Flow-in-cloud. In N. Voros, G. Keramidas, C. Antonopoulos, M. Huebner, P. C. Diniz, & D. Goehringer (Eds.), Applied Reconfigurable Computing: Architectures, Tools, and Applications - 14th International Symposium, ARC 2018, Proceedings (pp. 43-54). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10824 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-78890-6_4

Deep learning on high performance FPGA switching boards: Flow-in-cloud. / Musha, Kazusa; Kudoh, Tomohiro; Amano, Hideharu.
Applied Reconfigurable Computing: Architectures, Tools, and Applications - 14th International Symposium, ARC 2018, Proceedings. ed. / Nikolaos Voros; Georgios Keramidas; Christos Antonopoulos; Michael Huebner; Pedro C. Diniz; Diana Goehringer. Springer Verlag, 2018. p. 43-54 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10824 LNCS).

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

Musha, K, Kudoh, T & Amano, H 2018, Deep learning on high performance FPGA switching boards: Flow-in-cloud. in N Voros, G Keramidas, C Antonopoulos, M Huebner, PC Diniz & D Goehringer (eds), Applied Reconfigurable Computing: Architectures, Tools, and Applications - 14th International Symposium, ARC 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10824 LNCS, Springer Verlag, pp. 43-54, 14th International Symposium on Applied Reconfigurable Computing, ARC 2018, Santorini, Greece, 18/5/2. https://doi.org/10.1007/978-3-319-78890-6_4

Musha K, Kudoh T, Amano H. Deep learning on high performance FPGA switching boards: Flow-in-cloud. In Voros N, Keramidas G, Antonopoulos C, Huebner M, Diniz PC, Goehringer D, editors, Applied Reconfigurable Computing: Architectures, Tools, and Applications - 14th International Symposium, ARC 2018, Proceedings. Springer Verlag. 2018. p. 43-54. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-78890-6_4

Musha, Kazusa ; Kudoh, Tomohiro ; Amano, Hideharu. / Deep learning on high performance FPGA switching boards : Flow-in-cloud. Applied Reconfigurable Computing: Architectures, Tools, and Applications - 14th International Symposium, ARC 2018, Proceedings. editor / Nikolaos Voros ; Georgios Keramidas ; Christos Antonopoulos ; Michael Huebner ; Pedro C. Diniz ; Diana Goehringer. Springer Verlag, 2018. pp. 43-54 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{bd93afdf6adb4adcb850c1fb18e4a7a8,

title = "Deep learning on high performance FPGA switching boards: Flow-in-cloud",

abstract = "FiC (Flow-in-Cloud)-SW is an FPGA-based switching node for an efficient AI computing system. It is equipped with a number of serial links directly connected to other nodes. Unlike other multi-FPGA systems, the circuit switching fabric with the STDM (Static Time Division Multiplexing) is implemented on the FPGA for predictable communication and cost-efficient data broadcasting. Parallel convolution modules for AlexNet are implemented on FiC-SW1 prototype boards consisting of Kintex Ultrascale FPGA, and evaluation results show that the parallel execution with 20 boards achieved 4.6 times better performance than the state of art implementation on a single Virtex 7 FPGA board.",

author = "Kazusa Musha and Tomohiro Kudoh and Hideharu Amano",

note = "Funding Information: This paper is based on results obtained from a project commissioned by the New Energy and Industrial Technology Development Organization (NEDO). Publisher Copyright: {\textcopyright} Springer International Publishing AG, part of Springer Nature 2018.; 14th International Symposium on Applied Reconfigurable Computing, ARC 2018 ; Conference date: 02-05-2018 Through 04-05-2018",

year = "2018",

doi = "10.1007/978-3-319-78890-6_4",

language = "English",

isbn = "9783319788890",

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

publisher = "Springer Verlag",

pages = "43--54",

editor = "Nikolaos Voros and Georgios Keramidas and Christos Antonopoulos and Michael Huebner and Diniz, {Pedro C.} and Diana Goehringer",

booktitle = "Applied Reconfigurable Computing",

}

TY - GEN

T1 - Deep learning on high performance FPGA switching boards

T2 - 14th International Symposium on Applied Reconfigurable Computing, ARC 2018

AU - Musha, Kazusa

AU - Kudoh, Tomohiro

AU - Amano, Hideharu

N1 - Funding Information: This paper is based on results obtained from a project commissioned by the New Energy and Industrial Technology Development Organization (NEDO). Publisher Copyright: © Springer International Publishing AG, part of Springer Nature 2018.

PY - 2018

Y1 - 2018

N2 - FiC (Flow-in-Cloud)-SW is an FPGA-based switching node for an efficient AI computing system. It is equipped with a number of serial links directly connected to other nodes. Unlike other multi-FPGA systems, the circuit switching fabric with the STDM (Static Time Division Multiplexing) is implemented on the FPGA for predictable communication and cost-efficient data broadcasting. Parallel convolution modules for AlexNet are implemented on FiC-SW1 prototype boards consisting of Kintex Ultrascale FPGA, and evaluation results show that the parallel execution with 20 boards achieved 4.6 times better performance than the state of art implementation on a single Virtex 7 FPGA board.

AB - FiC (Flow-in-Cloud)-SW is an FPGA-based switching node for an efficient AI computing system. It is equipped with a number of serial links directly connected to other nodes. Unlike other multi-FPGA systems, the circuit switching fabric with the STDM (Static Time Division Multiplexing) is implemented on the FPGA for predictable communication and cost-efficient data broadcasting. Parallel convolution modules for AlexNet are implemented on FiC-SW1 prototype boards consisting of Kintex Ultrascale FPGA, and evaluation results show that the parallel execution with 20 boards achieved 4.6 times better performance than the state of art implementation on a single Virtex 7 FPGA board.

UR - http://www.scopus.com/inward/record.url?scp=85046291072&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85046291072&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-78890-6_4

DO - 10.1007/978-3-319-78890-6_4

M3 - Conference contribution

AN - SCOPUS:85046291072

SN - 9783319788890

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

SP - 43

EP - 54

BT - Applied Reconfigurable Computing

A2 - Voros, Nikolaos

A2 - Keramidas, Georgios

A2 - Antonopoulos, Christos

A2 - Huebner, Michael

A2 - Diniz, Pedro C.

A2 - Goehringer, Diana

PB - Springer Verlag

Y2 - 2 May 2018 through 4 May 2018

ER -

Deep learning on high performance FPGA switching boards: Flow-in-cloud

Abstract

Publication series

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this