公平运行同时多线程处理器中的线程  

Co-Scheduling Threads in SMT Processors Fairly

作　　者：孙彩霞[1] 张民选[1] 

机构地区：[1]国防科学技术大学计算机学院,湖南长沙410073

出　　处：《电子学报》2008年第2期224-229,共6页Acta Electronica Sinica

基　　金：国家863高技术研究发展计划重大项目(No.2005AA110020)

摘　　要：同时多线程(SMT,Simultaneous Multithreading)处理器中,取指策略隐式的决定了共享资源在线程之间的分配,进而决定了吞吐量和公平性.然而,前面对取指策略的研究大都集中在吞吐量优化上,对公平性的研究极少.本文把公平性作为优化目标,提出了一种新颖的取指策略FAIR.实验结果表明:对于所有类型的负载,FAIR都能获得很好的公平性,RPRrange都不超过5%.而且,FAIR并没有以牺牲吞吐量来获取公平性.与ICOUNT这种典型的以优化吞吐量为目标的取指策略相比,FAIR的吞吐量平均只降低了3.8%.In Simultaneous Multithreading（SMT）processors,the instnction fetch policy determines the way of allccating shared resources among co-scheduled threads implicitly, and affects throughput and fairness consequently. However, prior work on fetch policies almost focuses on throughput optimization. The issue of fairness between threads in progress rates is studied rarely. In this paper, we purchase fairness as the main optimization goal and propose a novel fetch policy called FAIR. Results show that for all types of workloads,FAIR can achieve good fairness and RPRrange（a fairness metric defined in this paper）is less than 5% .Furthennore,FAIR does not sacrifice throughput severely. Compared to fetch policies orienting towards throughput maximization such as ICOUNT, degradation of FAIR in throughput is 3.8 % on average.

关 键 词：同时多线程 取指策略 吞吐量 公平性 

分 类 号：TP303[自动化与计算机技术—计算机系统结构]