异构多核系统多媒体流计算实时任务调度策略研究

发布时间：2018-12-26 17:20

【摘要】：随着嵌入式移动多媒体应用的日益丰富与普及，保证智能终端的实时性显得非常突出。能保证智能终端的处理性能的处理器的快速发展也使多核处理器上的实时任务调度问题成为计算机技术领域研究的热点。目前的研究热点在于，大量的并行任务运行时，需要平衡任务的实时性和系统的效率，减少总任务的执行时间，并且保证整个系统负载均衡。设计高效的多媒体流计算任务调度策略是提高异构多核处理器系统性能的最直接有效的方法，，具有重要的研究意义。异构多核系统上进行任务的分配及调度问题已经被证明是NP完全问题。普遍认为，采用启发式算法解决该问题能得到比较好的解。基于此，针对视频解码这一典型的多媒体流计算，提出了一种基于蚁群算法的新型的调度算法来解决异构多核处理器系统的实时任务调度问题。首先，为实现该算法建立系统模型，以便模拟实际异构多核环境。然后设计了算法的公式，包括任务选择公式、处理器选择公式以及任务信息素和处理器信息素更新公式。最后详细阐述了算法的实现，包括任务的选择及处理核的选择。为了满足异构多核系统中任务的实时性要求，设计任务模型时考虑了任务的到达时间，估计运算量以及截止时间等因素。为了模拟异构多核环境，设计处理器模型时考虑了各个内核不同的执行速度以及内核之间的通信带宽等。该算法能较好的满足任务间的执行优先约束关系以及截止时间的要求，取得较优的总任务调度长度。使用Matlab7.0实现了该算法，以评估上述异构多核系统中基于蚁群的实时任务调度算法的性能，并对算法参数的取值进行讨论。实验表明，该算法能用于解决异构多核系统中实时任务的调度问题，并且具有较好的收敛性。
[Abstract]:With the increasing richness and popularity of embedded mobile multimedia applications, it is very important to ensure the real-time performance of intelligent terminals. With the rapid development of processors which can guarantee the processing performance of intelligent terminals, the real-time task scheduling problem on multi-core processors has become a hot topic in the field of computer technology. The current research focus is that when a large number of parallel tasks are running, it is necessary to balance the real time of the task and the efficiency of the system, reduce the execution time of the total task, and ensure the load balance of the whole system. The design of efficient task scheduling strategy for multimedia stream computing is the most direct and effective method to improve the performance of heterogeneous multi-core processors. Task assignment and scheduling on heterogeneous multicore systems have been proved to be NP complete problems. It is generally believed that the heuristic algorithm can get a better solution to the problem. Based on this, a new scheduling algorithm based on ant colony algorithm is proposed to solve the real-time task scheduling problem in heterogeneous multi-core processor systems. Firstly, the system model is built to simulate the real heterogeneous multi-core environment. Then the formula of the algorithm is designed, including the formula of task selection, the formula of processor selection and the formula of task pheromone and processor pheromone update. Finally, the implementation of the algorithm, including task selection and processing kernel selection, is described in detail. In order to meet the real-time requirements of tasks in heterogeneous multi-core systems, the task model is designed taking into account such factors as the time of arrival, the estimation of computation and the deadline, etc. In order to simulate heterogeneous multi-core environment, the processor model is designed with different execution speed and communication bandwidth between cores. The algorithm can satisfy the requirements of execution priority constraint and deadline between tasks, and obtain a better total task scheduling length. The algorithm is implemented with Matlab7.0 to evaluate the performance of the real-time task scheduling algorithm based on ant colony in the heterogeneous multi-core system, and the parameters of the algorithm are discussed. Experiments show that the algorithm can be used to solve the real-time task scheduling problem in heterogeneous multi-core systems and has good convergence.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP332

【参考文献】