Using Dataflow to Optimize Energy Efficiency of Deep Neural Network Accelerators

Author(s)
Chen, Yu-Hsin; Emer, Joel S; Sze, Vivienne
Thumbnail
Download_2017__Micro_Top_Picks__Eyeriss__Article_.pdf (280.9Kb)
Open Access Policy
Terms of use
Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/
Metadata
Show full item record
Abstract
The authors demonstrate the key role dataflows play in the optimization of energy efficiency for deep neural network (DNN) accelerators. By introducing a systematic approach to analyze the problem and a new dataflow, called Row-Stationary, which is up to 2.5 times more energy efficient than existing dataflows in processing a state-of-the-art DNN, this work provides guidelines for future DNN accelerator designs.
Date issued
2017-06
URI
https://hdl.handle.net/1721.1/130106
Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science; Massachusetts Institute of Technology. Microsystems Technology Laboratories
Journal
IEEE Micro
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Citation
Chen, Yu-Hsin et al. "Using Dataflow to Optimize Energy Efficiency of Deep Neural Network Accelerators." IEEE Micro 37, 3 (June 2017): 12 - 21. © 2017 IEEE
Version: Author's final manuscript
ISSN
0272-1732
Collections