NPB MG在分布式环境下的并行实现

Parallel Implementation of NPB MG in a Distributed System

ＮＡＳＰａｒａｌｅｌＢｅｎｃｈｍａｒｋｓ（ＮＰＢ）是为了测试和评价超级计算机性能而设计的并行基准测试程序集，ＭＧＢｅｎｃｈｍａｒｋ是其中的一个核心程序，它用多重网格方法求解三维Ｐｏｉｓｏｎ偏微分方程，本文介绍了其主要算法，并给出了在分布式环境下的并行算法，最后给出了在ＳＧＩＣｈａｌｅｎｇｅＰＶＭ平台上的测试结果。

The NAS Parallel Benchmarks(NPB) has been developed for the performance evaluation of highly parallel supercomputers.The MG Benchmark is a simplified multigrid kernel,which solves a 3 D Poisson PDE with constant coefficients.This paper presents the main algorithms of the MG Benchmark,then gives a parallel algorithm, and finally gives out the result obtained on the SGI Challenge PVM platform.