摘要
提出了一种新的混合并行蚁群算法,在单机多核机及多核集群机下分别实现了MPI并行蚁群算法及MPI+TBB并行蚁群算法,应用于真实路网车辆路径问题(Vehicle Routing Problem,VRP),对两者进行了实验对比,实验结果表明MPI并行蚁群算法具有较高的加速比,和问题规模关系不大,刚开始呈现线性加速比。较MPI并行蚁群算法,MPI+TBB混合并行蚁群算法具有更好的可扩展性,在进程数较多时仍具有较高的加速比。
A new hybrid parallel ant colony algorithm （ACA） is presented, respectively construct MPI parallel ACA and MPI ＋ TBB parallel ACA in the multi-core single machine and multi-core clusters to apply to the real road network for vehicle routing problem （VRP）. The experimental results show that MPI based parallel ACA make higher speedup , have nothing to do with problem size, and show linear speedup at the beginning. MPI ＋ TBB based parallel hybrid parallel ACA has better scalability and a more high speedup than the MPI based parallel ACA in the more processes.
出处
《信息技术》
2012年第7期184-187,共4页
Information Technology
关键词
MPI
TBB
蚁群算法
多核
集群
MPI
TBB
ant colony algorithm
multi-core
clusters
