基于线程队列动态规划法的GPU性能优化  

GPU Performance Optimization Based on Thread Queue Dynamic Programming Method

作　　者：魏雄 胡倩 王秋娴 闫坤 许萍萍 WEI Xiong;HU Qian;WANG Qiuxian;YAN Kun;XU Pingping(School of Mathematics and Computer,Wuhan Textile University,Wuhan 430000,China;School of Computer,National University of Defense Technology,Changsha 410073,China;Hubei Urban Construction Vocational and Technological College,Wuhan 430205,China)

机构地区：[1]武汉纺织大学数学与计算机学院,湖北武汉430000 [2]国防科技大学计算机学院,湖南长沙410073 [3]湖北城市建设职业技术学院,湖北武汉430205

出　　处：《计算机与网络》2021年第10期67-71,共5页Computer & Network

摘　　要：针对目前线程队列管理方法中执行时间长的线程无法得到响应的问题,对线程队列进行了研究,提出了线程队列动态规划法(Thread Queue Dynamic Programming,TQDP)。通过实验验证,对比了评估性能的参数,如Weighted Speedup,STP,ANTT等,TQDP与++DynCTAT,Mod+Bypass,PBS,BF相比,Weighted Speedup分别提高了14%,13.7%,13.3%,3.4%。与SMK,SMK-(P+W)相比,STP分别提高了8.5%,1.4%,表明提出的方法实验效果好,优化了系统性能。Aiming at the problem that threads with long execution time in the current thread queue management methods cannot be responded to,the thread queue is studied and a thread queue dynamic programming method(TQDP)is proposed,and compared the parameters for performance evaluation such as Weighted Speedup,system throughput(STP),average normalized turnaround time(ANTT),etc.,the results of evaluation shows that the proposed schemes significantly improve the weighted speedup of ++DynCTAT,Mod+Bypass,PBS,BF,by 14%,13.7%,13.3%,3.4%on average,respectively,and the system throughput of SMK and SMK-(P+W),by 8.5%and 1.4%,respectively.The experimental results show that the method proposed in this paper is effective,and optimizes the systemperformance.

关 键 词：线程队列 线程动态规划 加速比 线程调度 并行计算 

分 类 号：TP338.6[自动化与计算机技术—计算机系统结构]