E-ERA: An energy-efficient reconfigurable architecture for RNNs using dynamically adaptive approximate computing

IEICE Electronics Express

Yu Gong¹  Longxing Shi¹  Wei Dong¹  Tingting Xu¹  Wei Ge¹  Bo Liu¹  Jinjiang Yang¹ 

[1] National ASIC System Engineering Technology Research Center, Southeast University

关键词: reconfigurable architecture;    recurrent neural network;    dynamically adaptive accuracy;   

DOI  :  10.1587/elex.14.20170637

学科分类：电子、光学、磁材料

来源: Denshi Jouhou Tsuushin Gakkai

【 摘 要 】

This paper proposes an Energy-Efficient Reconfigurable Architecture (E-ERA) for Recurrent Neural Networks (RNNs). In E-ERA, reconfigurable computing arrays with approximate multipliers and dynamically adaptive accuracy controlling mechanism are implemented to achieve high energy efficiency. The E-ERA prototype is implemented on TSMC 45 nm process. Experimental results show that, comparing with traditional designs, the power consumption of E-ERA is reduced by 28.6%∼52.3%, with only 5.3%∼9.2% loss in accuracy. Compared with state-of-the-art architectures, E-ERA outperforms up to 1.78X in power efficiency and can achieve 304 GOPS/W when processing RNNs for speech recognition.

【 授权许可】

CC BY