Abstract:

This paper constructs parallel model based on Open Multi-processing(OpenMP) and parallel model based on Message Passing Interface(MPI) in multi-core environment, and constructs parallel model based on MPI and parallel model based on MPI+OpenMP in the multi-core clusters. The preferred strategy of dynamic ant colony and the strategy of sub-cycle exchange are presented. Finding optimized path of road is used to compare the three models. Experimental results show that MPI-based model run faster and get higher speedup than OpenMP-based model. In multi-core environment, hybrid model of MPI+OpenMP gets higher speedup than MPI-based model with more processes.

Key words: ant colony algorithm, mutlicore, cluster, parallel model, information communication strategy

摘要：

在单机多核下分别构造基于OpenMP和MPI的并行蚁群算法模型，在多核集群机下构造基于MPI和MPI+OpenMP的并行蚁群算法模型，并提出动态蚁群择优策略及分段周期交流策略。基于实际路网的路径寻优问题对上述模型进行比较，实验结果表明，在单机多核下，基于MPI的模型与基于OpenMP的模型相比，运行时间短，加速比高，在多核集群机下，基于MPI+OpenMP的混合模型相比基于MPI的模型，在进程数较多时仍具有较高的加速比。

关键词: 蚁群算法, 多核, 集群机, 并行模型, 信息交流策略

CLC Number: 

• TP301.6