一种基于数据划分和任务映射的并行调度算法  

A Parallel Scheduling Algorithm Based on Data Partitioning and Task Mapping

作　　者：马晓慧[1] 陈娟[1] 

机构地区：[1]山西大学商务学院信息学院,太原030031

出　　处：《现代计算机（中旬刊）》2012年第5期7-10,共4页Modern Computer

基　　金：山西省教育科学"十二五"规划课题(No.GH-11178)

摘　　要：提出一种基于数据划分和任务复制与映射的片上多核处理器并行调度算法,该算法基于数据划分策略,优先将同一进程的线程节点放在同一处理核上执行,有效地降低任务调度的时间复杂度。为了避免任务负载不均衡,采用处理器核上的极限值来控制过载。为了达到最优的任务执行时间,采用任务复制与映射的策略,寻找关键任务并予以复制,从而充分利用处理器各个核的资源来提高任务的调度效率。实验分析表明,随着任务调度数量的不断增加,算法的任务完成时间总是最少的。Presents a parallel scheduling algorithm of chip multi-core processor based on task clustering and duplication. This algorithm, using the strategy of task clustering, gives priority to the operation of thread nodes of the same process on the same processor and effectively reduces time complexity of task scheduling. In order to avoid the unbalanced task load on the processors, it will employ their ultimate values to control the load. For achieving the optimal time of task operations, this algorithm, with the adoption of task duplication strategy, looks for the key tasks and duplicates them so as to fully utilize the resources of each core on the processor and improve the efficiency of task scheduling. The analysis of the experiment shows that, with the increasing number of task scheduling, the time of task operation of this algorithm is always the least.

关 键 词：数据划分 任务映射 并行调度 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]