基于遗传蚁群算法的CMP线程调度方法  

CMP thread scheduling method based on genetic ant colony algorithm

作　　者：刘清华[1] 吴悦[1] 杨洪斌[1] 

机构地区：[1]上海大学计算机工程与科学学院,上海200072

出　　处：《计算机工程与设计》2011年第6期2116-2118,2123,共4页Computer Engineering and Design

基　　金：上海市重点学科建设基金项目(J50103)

摘　　要：为提高多核处理器系统的调度效率,充分发挥多核处理器的性能,提出了一种新的线程调度算法。该方法利用遗传算法的快速随机全局搜索能力,生成蚁群算法所需的信息素分布,利用蚁群算法的正反馈性,并将其应用到CMP的线程调度中,以提高线程调度的效率。通过两种算法的结合,弥补了遗传算法随着求解范围增大而效率降低,蚁群算法需要信息素浓度的增加才能提高效率的不足,更好地发挥它们的优势,提升求解速度。实验结果表明,该算法能够很好地降低任务的执行时间,充分发挥多核处理器系统的优势。To improve schedule efficiency and performance of multi-core processor system, a new thread scheduling algorithm is presented. This method uses the fast and random global search capability of genetic algorithm to generate the pheromone required by ant colony algorithm, then uses the positive feedback of ant colony algorithm, and applies it to CMP thread scheduling problem to improve the efficiency of thread scheduling. Two algorithms are combined to compensate for the deficiency of the efficiency dropped of genetic algorithm along with the increase in solving range and ant colony algorithm needs the increasing pheromone concentration which could improve the efficiency, and allow full play to their advantages preferably to promoting the solving speed. Experiment shows that this hybrid algorithm can reduce the task execution time and give full play to the advantage of multi-core processors system.

关 键 词：CMP 线程调度 遗传算法 蚁群算法 执行时间 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]