负载自适应的异构MPSoC任务调度算法研究

发布时间：2018-06-04 11:27

本文选题：异构MPSoC + 负载自适应　；参考：《工程科学与技术》2017年02期

【摘要】：在异构MPSoC中,并行任务通过调度算法被分配到各个处理器核上运行,因而任务调度算法的优劣将直接影响异构MPSoC的应用性能。根据处理器核类型和任务间依赖关系,以减小任务间通信开销为目标,提出一种具备负载自适应能力的异构MPSoC任务调度算法。首先,将待调度任务集划分为多个并行任务子集;其次,在考虑处理器核负载的基础上,根据并行任务子集集合、处理器核集合及任务子集在各个核上的执行效率生成赋权二部图;最后,利用赋权二部图最大权匹配方法,将并行任务子集合理地调度到负载适应的处理器核上运行,以降低任务集的平均调度长度,并提高处理器核利用率,从而实现异构MPSoC应用性能的提升。仿真实验在不同的任务总数、任务最大前驱数、核类型、核数量的应用场景下,通过任务集平均调度长度、处理器核利用率两项指标对提出算法进行了定量分析。结果表明,提出算法能有效降低任务集平均调度长度,在实现负载自适应的同时提高异构MPSoC处理器核的利用率。
[Abstract]:In heterogeneous MPSoC, parallel tasks are assigned to each processor core by scheduling algorithm, so the advantages and disadvantages of task scheduling algorithm will directly affect the application performance of heterogeneous MPSoC. In order to reduce the communication overhead between tasks, a heterogeneous MPSoC task scheduling algorithm with adaptive load capability is proposed according to the dependency relationship between processor cores and tasks. First, the task set to be scheduled is divided into multiple parallel task subsets; secondly, based on the processor kernel load, the set of parallel task subsets is proposed. The execution efficiency of the processor core set and task subset on each core generates the weighted bipartite graph. Finally, using the weighted bipartite graph maximum weight matching method, the parallel task subset is reasonably scheduled to run on the load-adaptive processor core. In order to reduce the average scheduling length of the task set and improve the processor core utilization, the performance of heterogeneous MPSoC applications can be improved. The simulation results show that the proposed algorithm is quantitatively analyzed by the average scheduling length of the task set and the processor kernel utilization ratio under different application scenarios of the total number of tasks, the maximum number of tasks, the type of cores and the number of cores. The results show that the proposed algorithm can effectively reduce the average scheduling length of task sets and improve the utilization of heterogeneous MPSoC processor cores while implementing load adaptation.
【作者单位】：中国科学院成都计算机应用研究所;中国科学院大学;西南民族大学计算机科学与技术学院;广西民族大学广西混杂计算与集成电路设计分析重点实验室;
【基金】：国家自然科学基金资助项目(11371003;11461006) 广西自然科学基金资助项目(2012GXNSFGA060003) 广西教育厅科研资助项目(201012MS274) 西南民族大学中央高校基本科研业务费专项资金资助项目(2015NZYQN28)
【分类号】：TP301.6

【相似文献】