Compiler-assisted Operator Template Library for DNN Accelerators

doi:10.1007/s10766-021-00701-6

	Compiler-assisted Operator Template Library for DNN Accelerators
	Li, Jiansong 1,2; Cao, Wei 1; Dong, Xiao 1,2; Li, Guangli 1,2; Wang, Xueying 1,2; Zhao, Peng 1,2; Liu, Lei 1; Feng, Xiaobing 1,2
	2021-03-25
发表期刊	INTERNATIONAL JOURNAL OF PARALLEL PROGRAMMING
ISSN	0885-7458
页码	18
摘要	Despite many dedicated accelerators are gaining popularity for their performance and energy efficiency in the deep neural network (DNN) domain, high-level programming support for these accelerators remains thin. In contrast to existing researches targeting the whole DNNs, we choose to dive into details and review this problem from a finer-grained level, operators. Due to performance concerns, operator programmers may have to take hand-written assembly as their first choice, which is error-prone and involves many programming chores. To alleviate this problem, we propose TOpLib, a compiler-assisted template library. By providing a unified user-view abstraction, TOpLib allows programmers to express computational kernels with high-level tensor primitives, which will be automatically lowered into low-level intrinsic primitives via expression templates. Moreover, considering memory management is performance-critical and the optimization strategy of expression template is limited to enumeration based rewriting rules, we implement TOpLib with a compiler-assisted approach. We address the memory reuse challenges into the compiler, which allows TOpLib to make full use of on-chip buffers and result in better performance. Experiments over 55 typical DNN operators demonstrate that TOpLib can generate scalable code with performance faster than or on par with hand-written assembly versions.
关键词	DNN Accelerators Template Library Address Space Management
DOI	10.1007/s10766-021-00701-6
收录类别	SCI
语种	英语
资助项目	National Key R&D Program of China[2017YFB1003103] ; Science Fund for Creative Research Groups of the National Natural Science Foundation of China[61521092]
WOS研究方向	Computer Science
WOS类目	Computer Science, Theory & Methods
WOS记录号	WOS:000632749100001
出版者	SPRINGER/PLENUM PUBLISHERS
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/16758
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Li, Jiansong
作者单位	1.Chinese Acad Sci, Inst Comp Technol, Beijing, Peoples R China 2.Univ Chinese Acad Sci, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Li, Jiansong,Cao, Wei,Dong, Xiao,et al. Compiler-assisted Operator Template Library for DNN Accelerators[J]. INTERNATIONAL JOURNAL OF PARALLEL PROGRAMMING,2021:18.
APA	Li, Jiansong.,Cao, Wei.,Dong, Xiao.,Li, Guangli.,Wang, Xueying.,...&Feng, Xiaobing.(2021).Compiler-assisted Operator Template Library for DNN Accelerators.INTERNATIONAL JOURNAL OF PARALLEL PROGRAMMING,18.
MLA	Li, Jiansong,et al."Compiler-assisted Operator Template Library for DNN Accelerators".INTERNATIONAL JOURNAL OF PARALLEL PROGRAMMING (2021):18.