面向MPSoC系统多特征的模糊动态调度算法  

A Fuzzy Dynamic Scheduling Algorithm for Multiple Characteristics of MPSoC System

作　　者：李达[1] 侯义斌[2] 黄樟钦[2] 肖春华[2] 

机构地区：[1]西安交通大学计算机科学与技术系,西安710049 [2]北京工业大学嵌入式软件与系统研究所,北京100124

出　　处：《计算机辅助设计与图形学学报》2011年第8期1447-1454,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60773186);北京工业大学高层次人才培育项目

摘　　要：针对MPSoC系统中的不确定多特征,提出基于模糊集理论的多核系统动态启发式调度算法.利用模糊集描述系统的多种不确定特征,根据模糊隶属度计算确定任务和资源的调度优先级,以提高任务调度的可并发执行能力;综合资源的最早可用时间和利用率,适当选取最接近处理器评估函数均值的处理器,以提高处理器利用率和均衡处理器负载.仿真模拟结果表明,该算法的系统综合性能好于近视算法、节约算法和分组适度算法,更能适用于多样特征的复杂MPSoC系统任务调度.Aiming at the multiple uncertain characteristics of MPSoC system, a dynamic heuristic scheduling algorithm is proposed based on fuzzy set theory for multicore processor system. This algorithm uses fuzzy sets for describing the multiple uncertain characteristics of system, and fuzzy set membership calculation for determining scheduling priorities of tasks and resources to improve the capability of concurrent execution. It presents a comprehensive estimation of the earliest available time and utilization of resources. It selects the processor closest to the average estimation to increase utilization ratios and balance loads on processors. The simulation results show that the proposed algorithm outperforms other original algorithms in comprehensive performances, and is more suitable for task scheduling in complex MPSoC systems.

关 键 词：MPSoC系统 动态调度 多特征 优先级 模糊集 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]