A Pipelining Loop Optimization Method for Dataflow Architecture  

作　　者：Xu Tan Xiao-Chun Ye Xiao-Wei Shen Yuan-Chao Xu Da Wang Lunkai Zhang Wen-Ming Li Dong-Rui Fan Zhi-Min Tang 

机构地区：[1]State Key Laboratory of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences Beijing 100190, China [2]School of Computer and Control Engineering, University of Chinese Academy of Sciences, Beijing 1000J9, China [3]State Key Laboratory of Mathematical Engineering and Advanced Computing, Wuxi 214125, China [4]College of Information Engineering, Capital Normal University, Beijing 100048, China [5]Department of Computer Science, The University of Chicago, Chicago, IL 60637, U.S.A.

出　　处：《Journal of Computer Science & Technology》2018年第1期116-130,共15页计算机科学技术学报（英文版）

基　　金：This work was supported by the National Key Research and Development Program of China under Grant No. 2016YFB0200501, tile National Natural Science Foundation of China under Grant Nos. 61332009 and 61521092, the Open Project Program of State Key Laboratory of Mathematical Engineering and Advanced Computing under Grant No. 2016A04 and tile Beijing Municipal Science and Technology Commission under Grant No. Z15010101009, the Open Project Program of State Key Laboratory of Computer Architecture under Grant No. CARCH201503, China Scholarship Council, and Beijing Advanced hmovation Center for hnaging Technology.

摘　　要：With the coming of exascale supercomputing era, power efficiency has become the most important obstacle to build an exascale system. Dataflow architecture has native advantage in achieving high power efficiency for scientific applications. However, the state-of-the-art dataflow architectures fail to exploit high parallelism for loop processing. To address this issue, we propose a pipelining loop optimization method （PLO）, which makes iterations in loops flow in the processing element （PE） array of dataflow accelerator. This method consists of two techniques, architecture-assisted hardware iteration and instruction-assisted software iteration. In hardware iteration execution model, an on-chip loop controller is designed to generate loop indexes, reducing the complexity of computing kernel and laying a good f（mndation for pipelining execution. In software iteration execution model, additional loop instructions are presented to solve the iteration dependency problem. Via these two techniques, the average number of instructions ready to execute per cycle is increased to keep floating-point unit busy. Simulation results show that our proposed method outperforms static and dynamic loop execution model in floating-point efficiency by 2.45x and 1.1x on average, respectively, while the hardware cost of these two techniques is acceptable.With the coming of exascale supercomputing era, power efficiency has become the most important obstacle to build an exascale system. Dataflow architecture has native advantage in achieving high power efficiency for scientific applications. However, the state-of-the-art dataflow architectures fail to exploit high parallelism for loop processing. To address this issue, we propose a pipelining loop optimization method （PLO）, which makes iterations in loops flow in the processing element （PE） array of dataflow accelerator. This method consists of two techniques, architecture-assisted hardware iteration and instruction-assisted software iteration. In hardware iteration execution model, an on-chip loop controller is designed to generate loop indexes, reducing the complexity of computing kernel and laying a good f（mndation for pipelining execution. In software iteration execution model, additional loop instructions are presented to solve the iteration dependency problem. Via these two techniques, the average number of instructions ready to execute per cycle is increased to keep floating-point unit busy. Simulation results show that our proposed method outperforms static and dynamic loop execution model in floating-point efficiency by 2.45x and 1.1x on average, respectively, while the hardware cost of these two techniques is acceptable.

关 键 词：dataflow model control-flow model loop optimization exascale computing scientific application 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论] TU-0[自动化与计算机技术—计算机科学与技术]