介于消息传递界面和并行应用之间的通信库系统功能分析

介于消息传递界面和并行应用之间的通信库系统对开发高性能的可靠、易维护的大型并行应用提供了支持。文章以BLACS和OilCL为例,从进程抽象命名,基于源的消息选择,非阻塞通信和集体通信等方面分析了介于消息传递系统和并行应用之间的通信库系统功能。     

不确定性MPI／PVM程序的完全调试

讨论如何完全地调试不确定性MPI／PVM并行程序．在循环调试过程中,不确定性导致前次遇到的错误在以后的执行中很可能无法再现．基于MPI／PVM的FIFO通信模型,给出一种记录－重放技术的实现．通过可控制的重放,用户可以覆盖所有可能的程序执行路径,从而达到完全调试的目的．和其它方法相比,所提供的方法所需时空开销要小得多．此技术已在两种消息传递体系结构上得到实现：一种是曙光－2000超级服务器（由国家智能计算机研究中心开发）,它由单处理器（PowerPC）结点经MESH网互联而成;另一种是国家高性能计算中心（     

基于PVM的C++对象分布并行机制的初步研究

对象具有内在的并行性.将面向对象程序设计与分布并行处理相结合,可产生既具有面向对象特征,又充分利用资源,还可缩短作业运行时间的对象分布并行系统.提出了一种基于PVM(parallelvirtualmachine)的C++对象的分布并行机制.该机制以协议和pvmlib作了向后兼容扩充的PVM系统为对象分布并行支撑;用预处理器将用户作业的并行类分离,并派遣到(PVM)系统中的目标机进行异地编译、加载执行;通过把并行类映射为PVM任务,请求对象消息映射为请求PVM任务消息来实现并行类对象的分布并行.实验结果表明,此对象分布并行机制(当问题规模达到一定程度时)可提高系统资源利用率和程序运行效率,并能简化PVM应用编程.     

A fast convergent iterative boundary element method on PVM cluster

This paper reports a fast convergent boundary element method on a Parallel Virtual Machine (PVM) (Geist et al., PVM: Parallel Virtual Machine, A Users' Guide and Tutorial for Networked Parallel Computing. MIT Press, Cambridge, 1994) cluster using the SIMD computing model (Single Instructions Multiple Data). The method uses the strategy of subdividing the domain into a number of smaller subdomains in order to reduce the size of the system matrix and to achieve overall speedup. Unlike traditional subregioning methods, where equations from all subregions are assembled into a single linear algebraic system, the present scheme is iterative and each subdomain is handled by a separate PVM node in parallel. The iterative nature of the overall solution procedure arises due to the introduction of the artificial boundaries. However, the system equations for each subdomain is now smaller and solved by direct Gaussian elimination within each iteration. Furthermore, the boundary conditions at the artificial interfaces are estimated from the result of the previous iteration by a reapplication of the boundary integral equation for internal points. This method provides a consistent mechanism for the specification of boundary conditions on artificial interfaces, both initially and during the iterative process. The method is fast convergent in comparison with other methods in the literature. The achievements of this method are therefore: (a) simplicity and consistency of methodology and implementation; (b) more flexible choice of type of boundary conditions at the artificial interfaces; (c) fast convergence; and (d) the potential to solve large problems on very affordable PVM clusters. The present parallel method is suitable where (a) one has a distributed computing environment; (b) the problem is big enough to benefit from the speedup achieved by coarse-grained parallelisation; and (c) the subregioning is such that communication overhead is only a small percentage of total computation time.     

基于PVM的并行调试环境

并行软件开发环境的研究已日益成为并行计算和并行处理的重点。本文简介了可移植的消息传递环境ＰＶＭ，讨论了以此为目标的图形监视环境ＸＰＶＭ，阐明ＸＰＶＭ环境与实际需要的ＰＶＭ并行调试环境之间的差距，并在此基础上探讨了并行调试环境开发中的技术难点及其设计要求。     

基于PVM的并行Lisp机制与实现

本文提出了一种基于分布存储结构和消息传递平台的并行Ｌｉｓｐ语言，给出了Ｌｉｓｐ语言的并行机制，举例说明了并行Ｌｉｓｐ程序的设计方法，并用一个实际系统ＰＡＲＬｉｓｐ阐明了在ＰＶＭ上的主要实现方法，给出了性能测试结果。     

A study on optimal task decomposition of networked parallel computing using PVM

A numerical study is performed to investigate the effect of task decomposition on networked parallel processes using Parallel Virtual Machine (PVM), In our study, a PVM program distributed over a network of workstations is used in solving a finite difference version of a one dimensional heat equation, where natural choice of PVM programming structure would be the master-slave paradigm with the aim of finding an optimal configuration resulting in least computing time including communication overhead among machines. Given a set of PVM tasks comprised of one master and five slave programs, it is found that there exists a pseudo optimal number of machines, which does not necessarily coincide with the number of tasks, that yields the best performance when the network is under a light usage. Increasing the number of machines beyond this optimal one does not improve computing performance since increase in communication overhead among the excess number of machines offsets the decrease in CPU time obtained by distributing the PVM tasks among these machines. However, when the network traffic is heavy, the results exhibit a more random characteristic that is explained by the random nature of data transfer time.     

PVM与网络并行计算

松散耦合的异构型并行处理系统近年来又有较大的发展，这种只能支持粗粒度并行的计算环境，由于通讯硬件的更新和软件环境的开发，其性能已经能够与某些ＭＰＰ机相媲美，因而形成并行处理领域一个强有力的分支－网络计算，本文围绕ＰＶＭ，讨论网络计算的软件环境，概述其特色和发展，并与其它基于消息传递的软件环境相比较，最后预测ＰＶＭ未来可能面临的问题和发展方向。     

基于PVM的动态可伸缩并行调试器设计

ＰＶＭ（ｐａｒａｌｌｅｌ ｖｉｒｔｕａｌ ｍａｃｈｉｎｅ）正在被越来越广泛地使用，但ＰＶＭ应用程序的并行调试还是一个有待解决的难题。文中介绍了自行设计的基于ＰＶＭ的动态可伸缩并行调试器ｄｄｂｇ，主要内容包括并行调试界面、总体设计和一些关键的实现技术。ｄｄｂｇ与其它已有的消息传递并行调试器相比，最大特点是其动态可伸缩结构，它能根据应用的需要自动分布调试服务。     

粒子跟踪算法的并行实现

针对粒子跟踪算法提出了并行方案,设计了相应的并行绘制模型，并对该方案进行了算法设计和程序验证，以理想的加速比和较高的效率实现了粒子跟踪算法。