SynergyFlow: An Elastic Accelerator Architecture Supporting Batch Processing of Large-Scale Deep Neural Networks

doi:10.1145/3275243

	SynergyFlow: An Elastic Accelerator Architecture Supporting Batch Processing of Large-Scale Deep Neural Networks
	Li, Jiajun 1,2; Yan, Guihai 1,2; Lu, Wenyan 1,2; Gong, Shijun 1,2; Jiang, Shuhao 1,2; Wu, Jingya 1,2; Li, Xiaowei 1,2
	2019
发表期刊	ACM TRANSACTIONS ON DESIGN AUTOMATION OF ELECTRONIC SYSTEMS
ISSN	1084-4309
卷号	24 期号:1 页码:27
摘要	Neural networks (NNs) have achieved great success in a broad range of applications. As NN-based methods are often both computation and memory intensive, accelerator solutions have been proved to be highly promising in terms of both performance and energy efficiency. Although prior solutions can deliver high computational throughput for convolutional layers, they could incur severe performance degradation when accommodating the entire network model, because there exist very diverse computing and memory bandwidth requirements between convolutional layers and fully connected layers and, furthermore, among different NN models. To overcome this problem, we proposed an elastic accelerator architecture, called SynergyFlow, which intrinsically supports layer-level and model-level parallelism for large-scale deep neural networks. SynergyFlow boosts the resource utilization by exploiting the complementary effect of resource demanding in different layers and different NN models. SynergyFlow can dynamically reconfigure itself according to the workload characteristics, maintaining a high performance and high resource utilization among various models. As a case study, we implement SynergyFlow on a P395-AB FPGA board. Under 100MHz working frequency, our implementation improves the performance by 33.8% on average (up to 67.2% on AlexNet) compared to comparable provisioned previous architectures.
关键词	Deep neural networks convolutional neural networks accelerator architecture resource utilization complementary effect
DOI	10.1145/3275243
收录类别	SCI
语种	英语
资助项目	National Natural Science Foundation of China[61572470] ; National Natural Science Foundation of China[61872336] ; National Natural Science Foundation of China[61532017] ; National Natural Science Foundation of China[61432017] ; National Natural Science Foundation of China[61521092] ; National Natural Science Foundation of China[61376043] ; Youth Innovation Promotion Association, CAS[Y404441000]
WOS研究方向	Computer Science
WOS类目	Computer Science, Hardware & Architecture ; Computer Science, Software Engineering
WOS记录号	WOS:000455951700008
出版者	ASSOC COMPUTING MACHINERY
引用统计	被引频次：1[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/2XEOYT63/3471
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, Guihai; Li, Xiaowei
作者单位	1.Chinese Acad Sci, Inst Comp Technol, State Key Lab Comp Architecture, 6 Kexueyuan South Rd, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Li, Jiajun,Yan, Guihai,Lu, Wenyan,et al. SynergyFlow: An Elastic Accelerator Architecture Supporting Batch Processing of Large-Scale Deep Neural Networks[J]. ACM TRANSACTIONS ON DESIGN AUTOMATION OF ELECTRONIC SYSTEMS,2019,24(1):27.
APA	Li, Jiajun.,Yan, Guihai.,Lu, Wenyan.,Gong, Shijun.,Jiang, Shuhao.,...&Li, Xiaowei.(2019).SynergyFlow: An Elastic Accelerator Architecture Supporting Batch Processing of Large-Scale Deep Neural Networks.ACM TRANSACTIONS ON DESIGN AUTOMATION OF ELECTRONIC SYSTEMS,24(1),27.
MLA	Li, Jiajun,et al."SynergyFlow: An Elastic Accelerator Architecture Supporting Batch Processing of Large-Scale Deep Neural Networks".ACM TRANSACTIONS ON DESIGN AUTOMATION OF ELECTRONIC SYSTEMS 24.1(2019):27.