摘要
针对现有异构多核处理器(CMP)任务调度算法中存在的优先级排序机制不健全、任务调度效率不高的问题,提出一种基于启发式的综合任务调度算法。该算法以表调度为基础,面向平台异构性和依赖任务特性,给出一种按照加权优先级构造任务调度列表的方式。在此基础上,采用多任务复制技术减少核间依赖任务的通信时延,提早任务的最早开始时间,并在任务分配阶段采取区间插入的方式将任务分配到合适的处理器内核上执行,以提高处理器资源利用率。设计模拟实验对该算法性能和已有的2个较为高效的启发式任务调度算法进行对比实验,结果证明该算法能有效提升任务调度性能。
Based on the shortcoming that the bad priority mechanism and the low task scheduling efficiency of existing task scheduling algorithm for heterogeneous Multi-core Processor (CMP),this paper proposes a comprehensive task scheduling algorithm which is based on heuristic algorithm.The algorithm belongs to list scheduling algorithm with a weighted scheduling priority list approach based on the characteristics of heterogeneous platform and the dependent task features.On this basis,the algorithm takes multi-duplication technology to reduce the community between the dependent tasks located on the different core,which advances the earliest start time of the task.And in the allocating strategy,it takes interval insertion as the task allocating scheme,which improves the resource utilization.Finally,a simulation experiment is designed to test the performance of the new algorithm and two compared heuristic algorithm.The results show that the new algorithm is effective to enhance the performance of task scheduling.
出处
《计算机工程》
CAS
CSCD
2014年第8期242-245,252,共5页
Computer Engineering
基金
国家自然科学基金资助项目(61003036)
中央高校基本科研业务费专项基金资助项目(HEUCF100607)
黑龙江省教育厅科学研究支撑基金资助项目(12513048)
关键词
异构多核
启发式算法
表调度
任务复制
heterogeneous multi-core
heuristic algorithm
list scheduling
task duplication