李东洋, 王云岚. 基于MPI的进程拓扑感知映射研究[J]. 微电子学与计算机, 2013, 30(5): 67-71.
引用本文: 李东洋, 王云岚. 基于MPI的进程拓扑感知映射研究[J]. 微电子学与计算机, 2013, 30(5): 67-71.
LI Dong-yang, WANG Yun-lan. Research on MPI Topology-Aware Process Mapping[J]. Microelectronics & Computer, 2013, 30(5): 67-71.
Citation: LI Dong-yang, WANG Yun-lan. Research on MPI Topology-Aware Process Mapping[J]. Microelectronics & Computer, 2013, 30(5): 67-71.

基于MPI的进程拓扑感知映射研究

Research on MPI Topology-Aware Process Mapping

  • 摘要: 本文首先介绍了获取应用程序通信拓扑和物理拓扑的方法,结合进程间通信量和通信频次定义了进程间亲和度,并针对具有分层网络、NUMA效应的硬件系统,设计了相应的进程映射算法.该算法根据进程间的亲和度对进程集合进行划分,在划分过程中构造出和物理拓扑树同构的树状结构,随后在两棵同构树的对应节点上做映射来解决应用程序进程到物理节点、处理核心的映射问题.在一个多核集群系统上的实验表明,使用本文提出的进程映射方法可以有效提高并行应用程序的通信效率,显著减少程序的总体执行时间.

     

    Abstract: This paper firstly introduce the method used to obtain application commumication topology and hardware topology.After that process proximity is defined considering both communication volume and communication frequency among processes.And then based on this work a process mapping method is proposed for computing systems with hierarchical networks and NUMA effect.This algorithm partitions the sets of processes base on process proximity.During the processes partition,a tree—like structure is built which is isomorphic to the physical topology tree.Then a mapping between any two correspondent nodes on the two isomorphic trees is performed. Experimental evaluations on a multicore cluster show that the process mapping method presented in this paper can effectively improve the communication efficiency of MPI application and cut down the whole execution time significantly in most cases.

     

/

返回文章
返回