Pragma Directed Shared Memory Centric Optimizations on GPUs

doi:10.1007/s11390-016-1624-8

	Pragma Directed Shared Memory Centric Optimizations on GPUs
	Li, Jing 1,2; Liu, Lei 1; Wu, Yuan 3; Liu, Xiang-Hua 3; Gao, Yi 3; Feng, Xiao-Bing 1; Wu, Cheng-Yong 1
	2016-03-01
发表期刊	JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY
ISSN	1000-9000
卷号	31 期号:2 页码:235-252
摘要	GPUs become a ubiquitous choice as coprocessors since they have excellent ability in concurrent processing. In GPU architecture, shared memory plays a very important role in system performance as it can largely improve bandwidth utilization and accelerate memory operations. However, even for affine GPU applications that contain regular access patterns, optimizing for shared memory is not an easy work. It often requires programmer expertise and nontrivial parameter selection. Improper shared memory usage might even underutilize GPU resource. Even using state-of-the-art high level programming models (e.g., OpenACC and OpenHMPP), it is still hard to utilize shared memory since they lack inherent support in describing shared memory optimization and selecting suitable parameters, let alone maintaining high resource utilization. Targeting higher productivity for affine applications, we propose a data centric way to shared memory optimization on GPU. We design a pragma extension on OpenACC so as to convey data management hints of programmers to compiler. Meanwhile, we devise a compiler framework to automatically select optimal parameters for shared arrays, using the polyhedral model. We further propose optimization techniques to expose higher memory and instruction level parallelism. The experimental results show that our shared memory centric approaches effectively improve the performance of five typical GPU applications across four widely used platforms by 3.7x on average, and do not burden programmers with lots of pragmas.
关键词	GPU shared memory pragma directed data centric
DOI	10.1007/s11390-016-1624-8
收录类别	SCI
语种	英语
资助项目	National High Technology Research and Development 863 Program of China[2012AA010902] ; National Natural Science Foundation of China (NSFC)[61432018] ; Innovation Research Group of NSFC[61221062]
WOS研究方向	Computer Science
WOS类目	Computer Science, Hardware & Architecture ; Computer Science, Software Engineering
WOS记录号	WOS:000372154100002
出版者	SCIENCE PRESS
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/8638
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, Jing; Liu, Lei; Wu, Yuan; Liu, Xiang-Hua; Gao, Yi; Feng, Xiao-Bing; Wu, Cheng-Yong
作者单位	1.Chinese Acad Sci, Inst Comp Technol, State Key Lab Comp Architecture, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 3.Beijing Samsung Telecom Res & Dev Ctr, Beijing 100028, Peoples R China
推荐引用方式 GB/T 7714	Li, Jing,Liu, Lei,Wu, Yuan,et al. Pragma Directed Shared Memory Centric Optimizations on GPUs[J]. JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY,2016,31(2):235-252.
APA	Li, Jing.,Liu, Lei.,Wu, Yuan.,Liu, Xiang-Hua.,Gao, Yi.,...&Wu, Cheng-Yong.(2016).Pragma Directed Shared Memory Centric Optimizations on GPUs.JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY,31(2),235-252.
MLA	Li, Jing,et al."Pragma Directed Shared Memory Centric Optimizations on GPUs".JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY 31.2(2016):235-252.