摘要
针对可重构片上系统软硬件划分问题,采用DAG建模,提出一种改进的图广度优先遍历法,将软硬件划分问题转化为带约束条件的0/1背包问题,提出基于小波变异的二进制粒子群算法。该算法改变BPSO的粒子参数计算模式,利用群体最优值和个体最优值决定粒子当前取值的概率,并引入小波变异以一定概率对粒子变异,得到最优计算结果。实验表明该算法提高了解的精度,得到令人满意的划分结果。
Hardware/Software partitioning is a significant problem in RSoC design. Based on a directed acyclic graph （DAG） model, an improved breadth first search （BFS） algorithm is proposed. The problem is then converted into the constrained 0/1 knapsack problem, and a wavelet mutation binary particle swarm optimization is presented. BPSO parameters of particle are changed in the algorithm, and the current values of the probability rested on the current value of global best position and the private best position. Then, with a certain probability to change the value of particle vector, wavelet mutation is introduced, and an optimal result is reached. Simulations show the algorithm improved？the accuracy of？the results, achieve an agreeable partitioning.
出处
《微计算机信息》
2011年第11期51-53,18,共4页
Control & Automation
基金
基金申请人:彭蔓蔓
项目名称:面向可重构片上系统的过程级动态软硬件划分研究
基金颁发部门:国家科技部(754209009)
关键词
DAG
广度优先遍历
粒子群优化算法
小波变异
directed acyclic graph
breadth first search
Particle Swarm Optimization
wavelet mutation
