摘要
流水线模式是网络处理器常用的一种编程模式,将任务映射到处理器处理引擎上去是NP-完全问题。针对以往基于遗传算法的解决方案过早收敛的局限性,提出m Ga Pipe算法。该算法采用优化交叉算子IMX和混合变异算子Hybrid M,避免遗传算法出现过早收敛,从而显著提高遗传算法解决此类问题的准确度。仿真结果显示m Ga Pipe算法在同等条件下将收敛到最优解的比率从传统遗传算法的解决方案的9.25%提升到52.25%。
Pipelined pattern is a generally usable programming pattern of network processors,and to map the tasks onto the processing engine of processor is an NP-Complete problem. But previous GA-based solution has some limitations,such as premature convergence. In light of this,we proposed an m Ga Pine algorithm. This algorithm adopts the optimised crossover operator（ IMX） and the hybrid mutation operator（ Hybrid M） and prevents the GA from being premature convergence,so that significantly improves the precision of GA in solving similar problems. Simulation results indicated that the mGaPipe algorithm raised the ratio of converging to best solution from 9. 25% of the solution of traditional genetic algorithm to 52. 25%.
出处
《计算机应用与软件》
CSCD
2015年第8期117-120,126,共5页
Computer Applications and Software
基金
国家科技支撑计划项目(2012BAH02B01)
中国科学院重点部署项目(KGZD-EW-103-2)
关键词
遗传算法
网络处理器
任务分配
流水线
Genetic algorithm（GA） Network processor Task allocation Pipeline
