3D NoC映射问题的动态蚁群算法

A Dynamic Ant Colony Optimization Algorithm for 3D NoC Mapping

摘要: 3D NoC映射通常涉及大量IP核及节点,使传统映射算法效率较低.为减少映射算法的执行时间,提高其优化能力,在传统蚁群算法(ACA)的基础上,提出一种动态蚁群算法(DACA).该算法采用逻辑斯蒂S形函数的变化形式,在每轮迭代开始前,依据当前迭代次数动态调整参数α,β及蚂蚁总数M.实验结果表明,与ACA相比,DACA可以缩短执行时间,提高算法性能;在面向随机任务时,其单位时间优化能力可以提升38.2%~65.9%;而当面向多媒体系统的真实应用时,其单位时间优化能力可以提升25.3%~32.7%.

Abstract: Considering that large number of nodes and tasks are involved during the mapping process of 3D NoC,traditional mapping algorithms are inefficient.To save the execution time and improve the optimization capacity,the dynamic ant colony algorithm(DACA) is proposed based on the ant colony algorithm(ACA) in this paper.In DACA,parameters α,β and the ant number M are all adjusted dynamically during the iteration by the logistic sigmoid function.Experimental results show that DACA can save the execution time and improve the performance compared with ACA.For random generated tasks,the improvement of the optimization capacity per second can reach 38.2%~65.9%.And,for a multimedia system,it can achieve an improvement of 25.3%~32.7%.