支持软件预取的缓存WCET分析  

WCET Analysis of Unified Cache with Software Prefetching

作　　者：安立奎[1,2] 古志民[1] 付引霞[1] 赵鑫[1] 甘志华[1] 

机构地区：[1]北京理工大学计算机学院,北京100081 [2]渤海大学数理学院,辽宁锦州121013

出　　处：《北京理工大学学报》2015年第7期730-736,共7页Transactions of Beijing Institute of Technology

基　　金：国家自然科学基金资助项目(61370062)

摘　　要：许多高性能嵌入式处理器都引入了多级缓存、硬件预取及软件预取等机制,为使支持软件预取的硬实时任务具有执行时间的可预测性,提出一种支持软件预取的缓存WCET分析方法.该方法对多级缓存抽象解释模型进行了软件预取语义扩展,分析了软件预取对任务的最坏情况下性能和能耗的影响.实验结果表明,该方法能够对支持软件预取的多级缓存行为进行有效分析;同时软件预取优化技术可使某些访存缺失较大的硬实时任务WCET平均减少22.9%,能耗平均降低24.1%.Many high-performance embedded processors have been introduced multistage cache,hardware and software prefetching mechanisms. In order to support software prefetching of hardreal-time tasks with the predictability of the execution time, a kind of unified cache WCETanalysis method with software prefetching was presented. This method extended the softwareprefetching semantic for traditional cache abstract interpretation model to analyze the influence ofsoftware prefetching on the worse-case performance and energy consumption of the tasks.Experimental results show that the method presented in this paper can effectively support unifiedcache behavior analysis with software prefetching; At the same time, for some hard real-timetasks with larger cache misses, software prefetching may make the WCET decrease 22.9% onaverage and the energy consumption reduce 24. 1% on average.

关 键 词：软件预取 缓存行为分析 最坏执行时间 能耗分析 

分 类 号：TP314[自动化与计算机技术—计算机软件与理论]