访存与用户行为敏感的MPSoC应用映射  

Memory-Aware and User-Aware Mapping of Applications to MPSoCs

作　　者：王一拙[1] 左琦[2] 计卫星[1] 王小军[1] 石峰[1] 

机构地区：[1]北京理工大学计算机学院,北京100081 [2]北京市计算中心,北京100094

出　　处：《电子学报》2015年第4期631-638,共8页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61300011;No.61300010)

摘　　要：应用映射是MPSo C设计中的关键问题,针对多应用负载的MPSo C,提出一种访存与用户行为敏感的动态映射策略,该策略根据应用的数据访问特征区分热点与非热点应用,并对用户行为进行建模,根据用户行为模型,进一步在运行时区分关键与非关键应用.对每个进入系统的应用,按照应用的热点及关键性分类动态选择在线映射算法,让热点应用围绕存储器布局,非热点应用尽量避免占用存储器附近的资源;对关键应用,最小化应用内通信开销和链路竞争,对非关键应用,最小化应用间通信开销和链路竞争.实验表明,与单纯考虑访存或用户行为的映射策略相比,本文策略能够降低系统整体的通信能耗.Application mapping is one of the key issues in MPSoC design. For MPSoC systems nmning multiple appfications, a memory-aware and user-aware application mapping strategy is proposed. This strategy classifies the appfications into hot applications which access large amounts of data and non-hot applications with moderate memory access by offline analysis. Then, the applications are further identified as critical or non-critical application with respect to the user behavior at runtime. For each application, different mapping algorithm is applied according to the above classification. Hot applications are distributed as close as possible to the shared memory. The internal communication cost and contents are minimized for critical applications, and the external communication cost and contents are minimized for non-critical applications. Experimental results show that the proposed strategy saves the overall system communication energy, compat-ed with the single memory-aware or user-aware strategy.

关 键 词：多处理器片上系统 片上网络 应用映射 任务映射 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]