PC-Cluster下的FFT并行算法分析

在地震资料处理中,大数据量的FFT运算非常耗费机时,在PC机群上实现FFT的并行运算,是一种低成本高效率的解决手段,但并行算法设计不是简单地串行运算的推广,仅仅考虑计算功能的实现是不够的,还需要考虑到通信的设计等与并行环境相关的因素才能充分挖掘并行系统的计算能力。本文讨论了FFT在PC机群环境下的MPI并行算法的设计思路,并对实际运行效果比较分析,得出了一些提高FFT并行效率的解决方法。

SEISMIC DATA PROCESSING'S PARALLEL COMPUTE BY NETWORK

In the seismic data processing field,although the speed of multiple data FFT operation is very slow,the parallel FFT works in a way of lowcost and high efficiency in a PCCluster.But the design of the parallel computing is not simply computing,so just considering the function's realization is not enough.Some other factors like communication design must be considered.This paper discusses the method of the MPI parallel programming in PCCluster environment.Then it analyses the effect of its running.Finally,it puts forward the solutions to improve the efficiency of parallel FFT program.