东南大学学报（自然科学版）2017，Vol.47Issue(2)：225-230,6.DOI:10.3969/j.issn.1001-0505.2017.02.005
求解大规模软硬件划分问题的爬山淘汰粒子群算法
Elimination particle swarm optimization with hill climbing algorithm for solving large-scale hardware/software partitioning problem
摘要
Abstract
To solve the large-scale hardware/software (HW/SW) partitioning problem,an elimination particle swarm optimization with hill climbing (EPSO-HC) algorithm is proposed.First,the Darwin's theory of evolution is stimulated,and the particles in the immediate vicinity of the current global worst position are eliminated to keep population diversity and avoid premature convergence.Secondly,the search mechanism of the hill climbing (HC) algorithm is improved by regarding the particles' own best positions as the search directions.Focus search is carded out near the current global position and the solution quality is improved.Then,the graphic processing unit (GPU) is adopted to compute the HW/SW communication cost in parallel to reduce the runtime of the EPSOHC algorithm.Finally,the experiments on benchmark and large-scale HW/SW partitioning tasks are conducted to evaluate the algorithm performance.The experimental results show that,compared with the other HW/SW partitioning algorithms for 23 HW/SW partitioning tasks,the proposed algorithm achieves higher quality solution and shorter runtime.关键词
软硬件划分/粒子群优化算法/爬山法/通信代价/并行计算Key words
hardware/software partitioning/particle swarm optimization algorithm/hill climbing algorithm/communication cost/parallel computing分类
信息技术与安全科学引用本文复制引用
鄢小虎,何发智,陈壹林..求解大规模软硬件划分问题的爬山淘汰粒子群算法[J].东南大学学报（自然科学版）,2017,47(2):225-230,6.基金项目
国家自然科学基金资助项目(61472289)、国家重点研发计划资助项目(2016YFC0106305). （61472289）
