An efficient scheduling algorithm for dataflow architecture using loop-pipelining

doi:10.1016/j.ins.2020.09.029

	An efficient scheduling algorithm for dataflow architecture using loop-pipelining
	Li, Yi 1,2; Wu, Meng 1; Ye, Xiaochun 1,3; Li, Wenming 1; Xue, Rui 1,2; Wang, Da 1; Zhang, Hao 1; Fan, Dongrui 1,2
	2021-02-08
发表期刊	INFORMATION SCIENCES
ISSN	0020-0255
卷号	547 页码:1136-1153
摘要	Dataflow architecture has native advantages in achieving high instruction parallelism and power efficiency for today's emerging applications such as high performance computing and deep neural network. In dataflow computing, the execution of instructions is driven by data, so the data transfer efficiency of the network on chip (NoC) is a key factor affecting performance. However, the NoC performance degrades due to the increasing use of multi-cast communications in many applications. The existing dataflow architecture instruction scheduling algorithms do not optimize multicast communication between the instruction and its successor instructions, so the routing paths of many multicast packets have forks which cause bandwidth waste and potential network congestion. We propose a sharing path awareness (SPA) algorithm to optimize multicast communication in the dataflow architecture. The algorithm shares the routing paths from the instruction to its child node to reduce the NoC bandwidth waste through the instruction scheduler. For applications using software iteration, we further extend the loop optimization to the SPA algorithm to sufficiently exploit instruction-level parallelism. Compared with the state-of-the-art algorithm, we show that the SPA algorithm achieves 20.21% average performance improvement and 15.11% energy consumption reduction for our experimental workloads. (C) 2020 Elsevier Inc. All rights reserved.
关键词	Dataflow architecture Instruction scheduling Multicast Sharing path Loop optimization
DOI	10.1016/j.ins.2020.09.029
收录类别	SCI
语种	英语
资助项目	National Key Research and Development Program[2018YFB1003501] ; National Natural Science Foundation of China[61732018] ; National Natural Science Foundation of China[61872335] ; National Natural Science Foundation of China[61802367] ; National Natural Science Foundation of China[61672499] ; Austrian-Chinese Cooperative RD Project (FFG)[171111KYSB20170032] ; Austrian-Chinese Cooperative RD Project (CAS)[171111KYSB20170032] ; Strategic Priority Research Program of Chinese Academy of Sciences[XDC05000000] ; Open Project Program of the State Key Laboratory of Mathematical Engineering and Advanced Computing[2019A07]
WOS研究方向	Computer Science
WOS类目	Computer Science, Information Systems
WOS记录号	WOS:000590678700008
出版者	ELSEVIER SCIENCE INC
引用统计	被引频次：5[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/2XEOYT63/16526
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Ye, Xiaochun
作者单位	1.Chinese Acad Sci, State Key Lab Comp Architecture, ICT, Beijing, Peoples R China 2.UCAS, Sch Comp & Control Engn, Beijing, Peoples R China 3.State Key Lab Math Engn & Adv Comp, Wuxi, Jiangsu, Peoples R China
推荐引用方式 GB/T 7714	Li, Yi,Wu, Meng,Ye, Xiaochun,et al. An efficient scheduling algorithm for dataflow architecture using loop-pipelining[J]. INFORMATION SCIENCES,2021,547:1136-1153.
APA	Li, Yi.,Wu, Meng.,Ye, Xiaochun.,Li, Wenming.,Xue, Rui.,...&Fan, Dongrui.(2021).An efficient scheduling algorithm for dataflow architecture using loop-pipelining.INFORMATION SCIENCES,547,1136-1153.
MLA	Li, Yi,et al."An efficient scheduling algorithm for dataflow architecture using loop-pipelining".INFORMATION SCIENCES 547(2021):1136-1153.