A RRAM-based coarse grain reconfigurable array for neural network accelerators

Zhengyu Chen, Hai Zhou, Jie Gu

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

Abstract

We propose a novel RRAM-based coarse grain reconfigurable array design for neural network computing. The proposed reconfigurable array design consists of RRAM-based reconfigurable AU array and the associated interconnects using novel RRAM-based multiplexer logic. A significant area saving is achieved compared with conventional design due to the simplification of logic expression as well as the saving of storage space and routing congestions from conventional switch controller. The experiments using 45nm CMOS technology show 47% area improvement and 27% performance enhancement can be achieved by using proposed RRAM-based reconfigure array technique.

Original languageEnglish (US)
Title of host publication2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781538676264
DOIs
StatePublished - Feb 11 2019
Event2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018 - Burlingame, United States
Duration: Oct 15 2018Oct 18 2018

Publication series

Name2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018

Conference

Conference2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018
CountryUnited States
CityBurlingame
Period10/15/1810/18/18

Fingerprint

Particle accelerators
accelerators
Neural networks
logic
space storage
congestion
simplification
CMOS
controllers
switches
Switches
Controllers
RRAM
augmentation
Experiments

ASJC Scopus subject areas

  • Hardware and Architecture
  • Electrical and Electronic Engineering
  • Safety, Risk, Reliability and Quality
  • Electronic, Optical and Magnetic Materials
  • Instrumentation

Cite this

Chen, Z., Zhou, H., & Gu, J. (2019). A RRAM-based coarse grain reconfigurable array for neural network accelerators. In 2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018 [8640182] (2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/S3S.2018.8640182
Chen, Zhengyu ; Zhou, Hai ; Gu, Jie. / A RRAM-based coarse grain reconfigurable array for neural network accelerators. 2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018. Institute of Electrical and Electronics Engineers Inc., 2019. (2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018).
@inproceedings{7abc3c3ac01647aa808c382c6a25450e,
title = "A RRAM-based coarse grain reconfigurable array for neural network accelerators",
abstract = "We propose a novel RRAM-based coarse grain reconfigurable array design for neural network computing. The proposed reconfigurable array design consists of RRAM-based reconfigurable AU array and the associated interconnects using novel RRAM-based multiplexer logic. A significant area saving is achieved compared with conventional design due to the simplification of logic expression as well as the saving of storage space and routing congestions from conventional switch controller. The experiments using 45nm CMOS technology show 47{\%} area improvement and 27{\%} performance enhancement can be achieved by using proposed RRAM-based reconfigure array technique.",
author = "Zhengyu Chen and Hai Zhou and Jie Gu",
year = "2019",
month = "2",
day = "11",
doi = "10.1109/S3S.2018.8640182",
language = "English (US)",
series = "2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
booktitle = "2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018",
address = "United States",

}

Chen, Z, Zhou, H & Gu, J 2019, A RRAM-based coarse grain reconfigurable array for neural network accelerators. in 2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018., 8640182, 2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018, Institute of Electrical and Electronics Engineers Inc., 2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018, Burlingame, United States, 10/15/18. https://doi.org/10.1109/S3S.2018.8640182

A RRAM-based coarse grain reconfigurable array for neural network accelerators. / Chen, Zhengyu; Zhou, Hai; Gu, Jie.

2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018. Institute of Electrical and Electronics Engineers Inc., 2019. 8640182 (2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018).

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

TY - GEN

T1 - A RRAM-based coarse grain reconfigurable array for neural network accelerators

AU - Chen, Zhengyu

AU - Zhou, Hai

AU - Gu, Jie

PY - 2019/2/11

Y1 - 2019/2/11

N2 - We propose a novel RRAM-based coarse grain reconfigurable array design for neural network computing. The proposed reconfigurable array design consists of RRAM-based reconfigurable AU array and the associated interconnects using novel RRAM-based multiplexer logic. A significant area saving is achieved compared with conventional design due to the simplification of logic expression as well as the saving of storage space and routing congestions from conventional switch controller. The experiments using 45nm CMOS technology show 47% area improvement and 27% performance enhancement can be achieved by using proposed RRAM-based reconfigure array technique.

AB - We propose a novel RRAM-based coarse grain reconfigurable array design for neural network computing. The proposed reconfigurable array design consists of RRAM-based reconfigurable AU array and the associated interconnects using novel RRAM-based multiplexer logic. A significant area saving is achieved compared with conventional design due to the simplification of logic expression as well as the saving of storage space and routing congestions from conventional switch controller. The experiments using 45nm CMOS technology show 47% area improvement and 27% performance enhancement can be achieved by using proposed RRAM-based reconfigure array technique.

UR - http://www.scopus.com/inward/record.url?scp=85063129940&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063129940&partnerID=8YFLogxK

U2 - 10.1109/S3S.2018.8640182

DO - 10.1109/S3S.2018.8640182

M3 - Conference contribution

T3 - 2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018

BT - 2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

Chen Z, Zhou H, Gu J. A RRAM-based coarse grain reconfigurable array for neural network accelerators. In 2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018. Institute of Electrical and Electronics Engineers Inc. 2019. 8640182. (2018 IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, S3S 2018). https://doi.org/10.1109/S3S.2018.8640182