文档介绍:粗粒度动态可重构处理器中的并发任务调度问题研究
方达
(电子科技大学通信与信息工程学院成都 611731)
【摘要】利用调度算法减少重构时间并提高执行并发任务的能力是提升可重构数字系统运行速度的一条重要途径。以粗粒度动态可重构处理器通用架构为基础,重点研究粗粒度动态可重构处理器执行并发任务调度问题的数学模型。通过数学推导,指出了影响动态可重构处理器执行并发任务时处理速度的两大关键因素——需配置的PE总个数和任务载入的总次数。针对影响处理速度的两大关键因素,比较了两种极端条件下相应的调度算法性能。仿真结果表明:执行时间较长时,调度算法对系统性能的改进程度低于配置时间较长时。
关键词:粗粒度;动态可重构;并发任务;调度算法
中图法分类号 TP302 文献标识码 A
Scheduling Algorithm of Concurrent Tasks in Dynamic Coarse-grained Reconfigurable Processors
Fang Da
(School munication and Information University of Electronic Science and Technology of China Chengdu 610054)
Abstract:One of the important ways to accelerate the operation of reconfigurable digital systems is to make use of scheduling alorithm so as to reduce reconfiguring time and to increase processing capacity. Based on the general architectures of dynamic coarse-grained reconfigurable processors, this paper formulate mathematics models when a number of tasks are concurrently implemented. By means of mathematical deduction, this paper points out two key factors influencing the processors’ processing speed when a number of tasks are concurrently implemented — the number of PEs reconfigured and the times of tasks loading. This paper pares the merits and drawbacks of the scheduling algorithm owning to the factors under two extreme conditions. Results of simulation show that the improvement rate produced by scheduling algorithm is lower when o