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

Kazusa Musha, Tomohiro Kudoh, Hideharu Amano

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (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 languageEnglish
Title of host publicationApplied Reconfigurable Computing
Subtitle of host publicationArchitectures, Tools, and Applications - 14th International Symposium, ARC 2018, Proceedings
PublisherSpringer Verlag
Pages43-54
Number of pages12
ISBN (Print)9783319788890
DOIs
Publication statusPublished - 2018 Jan 1
Event14th International Symposium on Applied Reconfigurable Computing, ARC 2018 - Santorini, Greece
Duration: 2018 May 22018 May 4

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10824 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other14th International Symposium on Applied Reconfigurable Computing, ARC 2018
CountryGreece
CitySantorini
Period18/5/218/5/4

Fingerprint

Field Programmable Gate Array
Field programmable gate arrays (FPGA)
High Performance
Data Broadcasting
Circuit Switching
Switching circuits
Time division multiplexing
Multiplexing
Vertex of a graph
Broadcasting
Convolution
Division
Learning
Deep learning
Prototype
Module
Computing
Communication
Evaluation
Costs

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Musha, K., Kudoh, T., & Amano, H. (2018). Deep learning on high performance FPGA switching boards: Flow-in-cloud. In 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. 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 proceedingConference contribution

Musha, K, Kudoh, T & Amano, H 2018, Deep learning on high performance FPGA switching boards: Flow-in-cloud. in 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 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)). https://doi.org/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. 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",
year = "2018",
month = "1",
day = "1",
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",
booktitle = "Applied Reconfigurable Computing",
address = "Germany",

}

TY - GEN

T1 - Deep learning on high performance FPGA switching boards

T2 - Flow-in-cloud

AU - Musha, Kazusa

AU - Kudoh, Tomohiro

AU - Amano, Hideharu

PY - 2018/1/1

Y1 - 2018/1/1

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

PB - Springer Verlag

ER -