Hardware Acceleration for GCNs via Bidirectional Fusion

doi:10.1109/LCA.2021.3077956

	Hardware Acceleration for GCNs via Bidirectional Fusion
	Li, Han 1,2; Yan, Mingyu 1,2; Yang, Xiaocheng 1; Deng, Lei 3; Li, Wenming 1; Ye, Xiaochun 1; Fan, Dongrui 1,2; Xie, Yuan 4
	2021
发表期刊	IEEE COMPUTER ARCHITECTURE LETTERS
ISSN	1556-6056
卷号	20 期号:1 页码:4
摘要	Derived from the fusion of graph traversal and neural networks, graph convolutional neural networks (GCNs) have achieved state-of-the-art performance in graph learning. However, the hybrid execution pattern, caused by the opposite characteristics of graph traversal based phase and neural network based transformation phase, poses huge challenges to the efficient execution of traditional architectures. Although GCN accelerators have emerged to address these challenges, they fail to harvest both bidirectional execution and inter-phase fusion opportunities exposed by the alternate execution phases in GCNs. Previous works either concentrate on a single execution direction or exchange the execution order of phases without inter-phase fusion, hence failing to further improve performance and efficiency. Therefore, we propose a novel hardware unit named BiFusion, which can be easily applied to existing GCN accelerators with hybrid architecture in order to harvest both of the above opportunities. BiFusion enables dynamic direction selection and inter-phase fusion, and helps significantly reduce the amounts of data access and computation. Experiments show that integrating the BiFusion unit helps the state-of-the-art GCN accelerator achieve 2x speedup on average.
关键词	Random access memory Computational modeling Analytical models Hardware Engines Computer architecture Transforms Graph convolutional neural networks hardware accelerator bidirectional execution inter-phase fusion
DOI	10.1109/LCA.2021.3077956
收录类别	SCI
语种	英语
资助项目	National Natural Science Foundation of China[61732018] ; National Natural Science Foundation of China[61872335] ; National Natural Science Foundation of China[61802367]
WOS研究方向	Computer Science
WOS类目	Computer Science, Hardware & Architecture
WOS记录号	WOS:000658323400003
出版者	IEEE COMPUTER SOC
引用统计	被引频次：4[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/17655
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Yan, Mingyu
作者单位	1.Chinese Acad Sci, Inst Comp Technol, Beijing 100864, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 3.Tsinghua Univ, Beijing 100084, Peoples R China 4.Univ Calif Santa Barbara, Santa Barbara, CA 93106 USA
推荐引用方式 GB/T 7714	Li, Han,Yan, Mingyu,Yang, Xiaocheng,et al. Hardware Acceleration for GCNs via Bidirectional Fusion[J]. IEEE COMPUTER ARCHITECTURE LETTERS,2021,20(1):4.
APA	Li, Han.,Yan, Mingyu.,Yang, Xiaocheng.,Deng, Lei.,Li, Wenming.,...&Xie, Yuan.(2021).Hardware Acceleration for GCNs via Bidirectional Fusion.IEEE COMPUTER ARCHITECTURE LETTERS,20(1),4.
MLA	Li, Han,et al."Hardware Acceleration for GCNs via Bidirectional Fusion".IEEE COMPUTER ARCHITECTURE LETTERS 20.1(2021):4.