FATE: Fast and accurate timing error prediction framework for low power DNN accelerator design

Jeff Jun Zhang, Siddharth Garg

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

Abstract

Deep neural networks (DNN) are increasingly being accelerated on application-specific hardware such as the Google TPU designed especially for deep learning. Timing speculation is a promising approach to further increase the energy efficiency of DNN accelerators. Architectural exploration for timing speculation requires detailed gate-level timing simulations that can be time-consuming for large DNNs which execute millions of multiply-and-accumulate (MAC) operations. In this paper we propose FATE, a new methodology for fast and accurate timing simulations of DNN accelerators like the Google TPU. FATE proposes two novel ideas: (i) DelayNet, a DNN based timing model for MAC units; and (ii) a statistical sampling methodology that reduces the number of MAC operations for which timing simulations are performed. We show that FATE results in between 8X - 58X speed-up in timing simulations, while introducing less than 2% error in classification accuracy estimates. We demonstrate the use of FATE by comparing a conventional DNN accelerator that uses 2's complement (2C) arithmetic with one that uses signed magnitude representation (SMR). We show that that the SMR implementation provides 18% more energy savings for the same classification accuracy than 2C, a result that might be of independent interest.

Original languageEnglish (US)
Title of host publication2018 IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - Digest of Technical Papers
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781450359504
DOIs
StatePublished - Nov 5 2018
Event37th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - San Diego, United States
Duration: Nov 5 2018Nov 8 2018

Other

Other37th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018
CountryUnited States
CitySan Diego
Period11/5/1811/8/18

Fingerprint

Particle accelerators
Energy efficiency
Energy conservation
Deep neural networks
Sampling
Hardware

ASJC Scopus subject areas

  • Software
  • Computer Science Applications
  • Computer Graphics and Computer-Aided Design

Cite this

Zhang, J. J., & Garg, S. (2018). FATE: Fast and accurate timing error prediction framework for low power DNN accelerator design. In 2018 IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - Digest of Technical Papers [a24] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1145/3240765.3240809

FATE : Fast and accurate timing error prediction framework for low power DNN accelerator design. / Zhang, Jeff Jun; Garg, Siddharth.

2018 IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - Digest of Technical Papers. Institute of Electrical and Electronics Engineers Inc., 2018. a24.

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

Zhang, JJ & Garg, S 2018, FATE: Fast and accurate timing error prediction framework for low power DNN accelerator design. in 2018 IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - Digest of Technical Papers., a24, Institute of Electrical and Electronics Engineers Inc., 37th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018, San Diego, United States, 11/5/18. https://doi.org/10.1145/3240765.3240809
Zhang JJ, Garg S. FATE: Fast and accurate timing error prediction framework for low power DNN accelerator design. In 2018 IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - Digest of Technical Papers. Institute of Electrical and Electronics Engineers Inc. 2018. a24 https://doi.org/10.1145/3240765.3240809
Zhang, Jeff Jun ; Garg, Siddharth. / FATE : Fast and accurate timing error prediction framework for low power DNN accelerator design. 2018 IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - Digest of Technical Papers. Institute of Electrical and Electronics Engineers Inc., 2018.
@inproceedings{aef3c2c946c04a52a4d9457c3f7b9866,
title = "FATE: Fast and accurate timing error prediction framework for low power DNN accelerator design",
abstract = "Deep neural networks (DNN) are increasingly being accelerated on application-specific hardware such as the Google TPU designed especially for deep learning. Timing speculation is a promising approach to further increase the energy efficiency of DNN accelerators. Architectural exploration for timing speculation requires detailed gate-level timing simulations that can be time-consuming for large DNNs which execute millions of multiply-and-accumulate (MAC) operations. In this paper we propose FATE, a new methodology for fast and accurate timing simulations of DNN accelerators like the Google TPU. FATE proposes two novel ideas: (i) DelayNet, a DNN based timing model for MAC units; and (ii) a statistical sampling methodology that reduces the number of MAC operations for which timing simulations are performed. We show that FATE results in between 8X - 58X speed-up in timing simulations, while introducing less than 2{\%} error in classification accuracy estimates. We demonstrate the use of FATE by comparing a conventional DNN accelerator that uses 2's complement (2C) arithmetic with one that uses signed magnitude representation (SMR). We show that that the SMR implementation provides 18{\%} more energy savings for the same classification accuracy than 2C, a result that might be of independent interest.",
author = "Zhang, {Jeff Jun} and Siddharth Garg",
year = "2018",
month = "11",
day = "5",
doi = "10.1145/3240765.3240809",
language = "English (US)",
booktitle = "2018 IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - Digest of Technical Papers",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - FATE

T2 - Fast and accurate timing error prediction framework for low power DNN accelerator design

AU - Zhang, Jeff Jun

AU - Garg, Siddharth

PY - 2018/11/5

Y1 - 2018/11/5

N2 - Deep neural networks (DNN) are increasingly being accelerated on application-specific hardware such as the Google TPU designed especially for deep learning. Timing speculation is a promising approach to further increase the energy efficiency of DNN accelerators. Architectural exploration for timing speculation requires detailed gate-level timing simulations that can be time-consuming for large DNNs which execute millions of multiply-and-accumulate (MAC) operations. In this paper we propose FATE, a new methodology for fast and accurate timing simulations of DNN accelerators like the Google TPU. FATE proposes two novel ideas: (i) DelayNet, a DNN based timing model for MAC units; and (ii) a statistical sampling methodology that reduces the number of MAC operations for which timing simulations are performed. We show that FATE results in between 8X - 58X speed-up in timing simulations, while introducing less than 2% error in classification accuracy estimates. We demonstrate the use of FATE by comparing a conventional DNN accelerator that uses 2's complement (2C) arithmetic with one that uses signed magnitude representation (SMR). We show that that the SMR implementation provides 18% more energy savings for the same classification accuracy than 2C, a result that might be of independent interest.

AB - Deep neural networks (DNN) are increasingly being accelerated on application-specific hardware such as the Google TPU designed especially for deep learning. Timing speculation is a promising approach to further increase the energy efficiency of DNN accelerators. Architectural exploration for timing speculation requires detailed gate-level timing simulations that can be time-consuming for large DNNs which execute millions of multiply-and-accumulate (MAC) operations. In this paper we propose FATE, a new methodology for fast and accurate timing simulations of DNN accelerators like the Google TPU. FATE proposes two novel ideas: (i) DelayNet, a DNN based timing model for MAC units; and (ii) a statistical sampling methodology that reduces the number of MAC operations for which timing simulations are performed. We show that FATE results in between 8X - 58X speed-up in timing simulations, while introducing less than 2% error in classification accuracy estimates. We demonstrate the use of FATE by comparing a conventional DNN accelerator that uses 2's complement (2C) arithmetic with one that uses signed magnitude representation (SMR). We show that that the SMR implementation provides 18% more energy savings for the same classification accuracy than 2C, a result that might be of independent interest.

UR - http://www.scopus.com/inward/record.url?scp=85058171498&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058171498&partnerID=8YFLogxK

U2 - 10.1145/3240765.3240809

DO - 10.1145/3240765.3240809

M3 - Conference contribution

AN - SCOPUS:85058171498

BT - 2018 IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2018 - Digest of Technical Papers

PB - Institute of Electrical and Electronics Engineers Inc.

ER -