DMazeRunner: Optimizing Convolutions on Dataflow Accelerators

Shail Dave; Aviral Shrivastava; Youngbin Kim; Sasikanth Avancha; Kyoungwoo Lee

doi:10.1109/ICASSP40776.2020.9054275

DMazeRunner: Optimizing Convolutions on Dataflow Accelerators

Shail Dave, Aviral Shrivastava, Youngbin Kim, Sasikanth Avancha, Kyoungwoo Lee

College of Computing

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

8 Citations (Scopus)

Abstract

Convolution neural networks (CNNs) can be efficiently executed on dataflow accelerators. However, the vast space of executing convolutions on computational and memory resources of accelerators makes difficult for programmers to automatically and efficiently accelerate the convolutions and for architects to achieve efficient accelerator designs. We propose dMazeRunner framework, which allows users to optimize execution methods for accelerating convolution and matrix multiplication on a given architecture and to explore dataflow accelerator designs for efficiently executing CNN models. dMazeRunner determines efficient dataflows tailored for CNN layers and achieves efficient execution methods for CNN models within several seconds.

Original language	English
Title of host publication	2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	1544-1548
Number of pages	5
ISBN (Electronic)	9781509066315
DOIs	https://doi.org/10.1109/ICASSP40776.2020.9054275
Publication status	Published - 2020 May
Event	2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Barcelona, Spain Duration: 2020 May 4 → 2020 May 8

Publication series

Name	ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume	2020-May
ISSN (Print)	1520-6149

Conference

Conference	2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020
Country/Territory	Spain
City	Barcelona
Period	20/5/4 → 20/5/8

Bibliographical note

Funding Information:
This research was partially supported by funding from NSF grant CCF 1723476 - NSF/Intel joint research center for Computer Assisted Programming for Heterogeneous Architectures (CAPA), and from grants NRF-2015M3C4A7065522 and 2014-3-00035, funded by MSIT. Any opinions, findings, and conclusions presented in this material are those of the authors and do not necessarily reflect the views of their employers or the sponsoring agencies.

Publisher Copyright:
© 2020 IEEE.

All Science Journal Classification (ASJC) codes

Software
Signal Processing
Electrical and Electronic Engineering

Access to Document

10.1109/ICASSP40776.2020.9054275

Cite this

Dave, S., Shrivastava, A., Kim, Y., Avancha, S., & Lee, K. (2020). DMazeRunner: Optimizing Convolutions on Dataflow Accelerators. In 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings (pp. 1544-1548). [9054275] (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2020-May). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICASSP40776.2020.9054275

Dave, Shail ; Shrivastava, Aviral ; Kim, Youngbin et al. / DMazeRunner : Optimizing Convolutions on Dataflow Accelerators. 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2020. pp. 1544-1548 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings).

@inproceedings{7e424e9f041940c28a088bef1050d1df,

title = "DMazeRunner: Optimizing Convolutions on Dataflow Accelerators",

abstract = "Convolution neural networks (CNNs) can be efficiently executed on dataflow accelerators. However, the vast space of executing convolutions on computational and memory resources of accelerators makes difficult for programmers to automatically and efficiently accelerate the convolutions and for architects to achieve efficient accelerator designs. We propose dMazeRunner framework, which allows users to optimize execution methods for accelerating convolution and matrix multiplication on a given architecture and to explore dataflow accelerator designs for efficiently executing CNN models. dMazeRunner determines efficient dataflows tailored for CNN layers and achieves efficient execution methods for CNN models within several seconds.",

author = "Shail Dave and Aviral Shrivastava and Youngbin Kim and Sasikanth Avancha and Kyoungwoo Lee",

note = "Funding Information: This research was partially supported by funding from NSF grant CCF 1723476 - NSF/Intel joint research center for Computer Assisted Programming for Heterogeneous Architectures (CAPA), and from grants NRF-2015M3C4A7065522 and 2014-3-00035, funded by MSIT. Any opinions, findings, and conclusions presented in this material are those of the authors and do not necessarily reflect the views of their employers or the sponsoring agencies. Publisher Copyright: {\textcopyright} 2020 IEEE.; 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 ; Conference date: 04-05-2020 Through 08-05-2020",

year = "2020",

month = may,

doi = "10.1109/ICASSP40776.2020.9054275",

language = "English",

series = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "1544--1548",

booktitle = "2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings",

address = "United States",

}

Dave, S, Shrivastava, A, Kim, Y, Avancha, S & Lee, K 2020, DMazeRunner: Optimizing Convolutions on Dataflow Accelerators. in 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings., 9054275, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, vol. 2020-May, Institute of Electrical and Electronics Engineers Inc., pp. 1544-1548, 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020, Barcelona, Spain, 20/5/4. https://doi.org/10.1109/ICASSP40776.2020.9054275

DMazeRunner : Optimizing Convolutions on Dataflow Accelerators. / Dave, Shail; Shrivastava, Aviral; Kim, Youngbin et al.

2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2020. p. 1544-1548 9054275 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2020-May).

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

TY - GEN

T1 - DMazeRunner

T2 - 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020

AU - Dave, Shail

AU - Shrivastava, Aviral

AU - Kim, Youngbin

AU - Avancha, Sasikanth

AU - Lee, Kyoungwoo

N1 - Funding Information: This research was partially supported by funding from NSF grant CCF 1723476 - NSF/Intel joint research center for Computer Assisted Programming for Heterogeneous Architectures (CAPA), and from grants NRF-2015M3C4A7065522 and 2014-3-00035, funded by MSIT. Any opinions, findings, and conclusions presented in this material are those of the authors and do not necessarily reflect the views of their employers or the sponsoring agencies. Publisher Copyright: © 2020 IEEE.

PY - 2020/5

Y1 - 2020/5

N2 - Convolution neural networks (CNNs) can be efficiently executed on dataflow accelerators. However, the vast space of executing convolutions on computational and memory resources of accelerators makes difficult for programmers to automatically and efficiently accelerate the convolutions and for architects to achieve efficient accelerator designs. We propose dMazeRunner framework, which allows users to optimize execution methods for accelerating convolution and matrix multiplication on a given architecture and to explore dataflow accelerator designs for efficiently executing CNN models. dMazeRunner determines efficient dataflows tailored for CNN layers and achieves efficient execution methods for CNN models within several seconds.

AB - Convolution neural networks (CNNs) can be efficiently executed on dataflow accelerators. However, the vast space of executing convolutions on computational and memory resources of accelerators makes difficult for programmers to automatically and efficiently accelerate the convolutions and for architects to achieve efficient accelerator designs. We propose dMazeRunner framework, which allows users to optimize execution methods for accelerating convolution and matrix multiplication on a given architecture and to explore dataflow accelerator designs for efficiently executing CNN models. dMazeRunner determines efficient dataflows tailored for CNN layers and achieves efficient execution methods for CNN models within several seconds.

UR - http://www.scopus.com/inward/record.url?scp=85089228240&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85089228240&partnerID=8YFLogxK

U2 - 10.1109/ICASSP40776.2020.9054275

DO - 10.1109/ICASSP40776.2020.9054275

M3 - Conference contribution

AN - SCOPUS:85089228240

T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

SP - 1544

EP - 1548

BT - 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 4 May 2020 through 8 May 2020

ER -

Dave S, Shrivastava A, Kim Y, Avancha S, Lee K. DMazeRunner: Optimizing Convolutions on Dataflow Accelerators. In 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2020. p. 1544-1548. 9054275. (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings). doi: 10.1109/ICASSP40776.2020.9054275

DMazeRunner: Optimizing Convolutions on Dataflow Accelerators

Abstract

Publication series

Conference

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this