多核处理平台上任务图模型的并行调度策略研究  

Research on Parallel Scheduling Algorithm of Task Graph Model on Multi-core Processing Platform

作　　者：周本海[1] 乔建忠[1,2] 林树宽[1] 

机构地区：[1]东北大学信息科学与工程学院,沈阳110819 [2]辽宁省软件开发与应用重点实验室,沈阳110819

出　　处：《小型微型计算机系统》2012年第11期2485-2492,共8页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60873009)资助;辽宁省重点科技攻关项目(R200801036)资助

摘　　要：凭借着高性能,低功耗的特性,多核处理器已经占据了目前的主要市场.提出一种多核处理平台上基于任务图模型的调度策略.建立了多核平台上任务图的空间与时间并行调度模型;针对任务图的空间并行与时间并行调度模型提出了并行节点合并、分配的优化算法与流水线并行的优化算法.最后,提出将优化的空间与时间并行调度技术相结合的并行调度策略.通过实验验证,本文提出的算法比其他多核并行调度算法降低了处理器核心间的通信与同步开销,提高了系统的计算效率与吞吐量.Nowadays,with high-performance and low power consumption,multi-core processors have occupied the main market.This paper proposes a task graph scheduling policy on multi-core processing platform.Firstly,the paper sets up the space and temporary parallel scheduling model;Secondly,aiming at the space and temporary parallel scheduling model,the optimized mergence and partition algorithms of parallel nodes and pipeline are proposed respectively;Thirdly,the paper presents a parallel scheduling method combining the space and temporary parallel technology.The experiment result shows that the new algorithm is able to reduce the communication and synchronous overhead and increase the computational efficiency and system throughout compared with other multi-core parallel algorithms.

关 键 词：多核处理器 调度 空间并行 时间并行 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]