资源限制性并行任务固定优先级可调度性分析

Schedulability Analysis for Resource-restricted Multi-core Parallel Tasks Under Global Fixed-priorities

异构多核平台的发展,导致并行任务需要执行在具有多样性资源的多核平台上.虽然,并行任务的某个程序片段只能在规定的资源上执行,但是这样操作可以充分利用各类不同资源的特性,达到更加快速节能处理任务的目的.同时,具有资源限制任务的可调度性研究在实时嵌入式系统领域已有一定的研究成果,但是采用的任务模型相对简单,分析方法不够精确.鉴于此,本文对具有资源限制性的并行任务在全局固定优先级调度策略下的可调度性问题进行了研究,基于单并行任务的分析方法提出了基于全局固定优先级调度策略的分析方法.首先,基于分解策略提出了高优先级任务干涉的分析方法.然后,将高优先级任务干涉分析方法和单并行任务提出的路径抽象技术相结合,推导出并行任务的最差响应时间算法.最后,通过仿真实验进行验证所提出的算法在可调度性、精确度层面的性能.实验结果表明,提出的算法在各个参数下的接受率实验符合实验预期,分析时间相对降低,但平均分析时间仍然在离线分析的可接受范围内,提出的算法能够对实时系统并行软件设计提供一定的指导价值.

With the development of heterogeneous multi-core platforms,parallel tasks need to be executed on multi-core platforms with diverse resources.A certain program segment of a parallel task can only be executed on specified resources.This operation can make full use of the characteristics of different resources and achieve the purpose of faster and more energy-efficient task processing.At the same time,the schedulability of resource-restricted tasks has been studied in the field of real-time embedded systems,but the task model is relatively simple and the analysis method is not accurate enough.In view of this,this paper studies the schedulability of resourcerestricted parallel tasks under the global fixed priority scheduling strategy,and presents an analysis method based on the global fixed priority scheduling strategy based on the single parallel task analysis method.First,a method is presented based on the decomposition strategy for analyzing high-priority task interference.A worst-case response time algorithm for parallel tasks is derived by combining the high-priority task interference analysis method with the path abstraction technique used for single parallel tasks.Finally,simulation experiments are performed to verify the performance of the developed algorithm.The experimental results show that the acceptance rate of the proposed algorithm under each parameter is in line with the experiments.The analysis time was reduced,but the average analysis time is still within the acceptable range of offline analysis.The proposed algorithm can provide some guiding values for the parallel software design of real-time systems.