基于改进人工鱼群算法的软硬件划分方法  

Hardware/Software Partitioning Method Based on Improved Artificial Fish Swarm Algorithm

作　　者：全浩军[1] 张涛[1] 郭继昌[1] 

机构地区：[1]天津大学电子信息工程学院,天津300072

出　　处：《天津大学学报（自然科学与工程技术版）》2013年第10期923-928,共6页Journal of Tianjin University：Science and Technology

基　　金：国家重大科技专项资助项目(2010ZX03004-003-03);国家自然科学基金资助项目(61179045

摘　　要：将人工鱼群算法应用于软硬件划分，从而提出一种软硬件划分方法．针对人工鱼群算法在应用于离散型问题时普遍存在的最优解出现概率低、收敛速度慢等问题，采用随机步长来改善鱼的游走行为，使用邻域搜索来获得邻域内的更优状态，并根据无效迭代次数来提前终止迭代、提高算法效率．在对不同结点数的随机 DAG 图划分实验中，改进后算法的平均耗时约为原算法的6.5%～34.5%，而最优解出现概率则为原算法的5～7倍．因此，改进后算法在寻优能力和收敛速度上均优于原始算法，可更高效地完成软硬件划分任务．Artificial fish swarm algorithm （AFSA） is adopted and a novel method for Hardware/Software （HW/SW） partitioning is proposed. When AFSA is applied to solve discrete problems, the optimum solution occurrence probabil-ity and the convergence speed are low. So, the fish behaviors are improved by random step, and then neighborhood searching is adopted to get a better state in the neighborhood. Finally, early termination is made and algorithm effi-ciency is improved based on the number of invalid-iteration. In the partitioning experiments of random directed acyclic graphs （DAGs） with different node numbers, the average time cost of improved AFSA is about 6.5%~34.5%of that of the original algorithm, and the optimum solution occurrence probability is 5-7 times that of the original algorithm. So the improved AFSA can achieve better results in search ability and convergence speed than the original algorithm. Thus the improved AFSA can perform HW/SW partitioning much more efficiently.

关 键 词：人工鱼群算法 软硬件划分 随机步长 邻域搜索 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]