基于MPI+OpenMP混合模型的并行地震数据处理支撑库的研究
详细信息 本馆镜像全文    |  推荐本文 | | 获取馆网全文
摘要
计算机集群技术已经引起了石油地球物理界的广泛关注,如何将现有地震数据处理模块快速、高效地移植到集群上已成为地震数据处理需要解决的重大课题。本文将现有的基于消息传递(MPI)的并行地震处理模型与共享存储(OpenMP)模型相结合,实现了一个适合于SMP集群的并行地震数据支撑库,将涉及到消息传递的并行地震数据操作以直观的API的形式提供给开发人员。本文利用支撑库提供的API开发了一些测试模块。实验证明,支撑库可支持现有地震数据处理和显示模块的多种并行计算模型,并且能够获得较高的并行加速比和计算效率。
The computer cluster technology has greatly attracted the attention of the geologists in the petroleum enterprises worldwide,and how to transplant the serial processing modules to the cluster has become a very important issue of seismic data processing.This paper combines the existing parallel seismic data processing patterns based on MPI with the shared memory programming paradigm,and provides a parallel seismic data processing support library suitable for the SMP cluster.We have developed some test modules using the support library.The experiment demonstrates that the platform can thoroughly support the current seismic data processing patterns and get a high parallel speedup.
引文
[1]杨红霞,赵改善.21世纪的地震数据处理系统[J].石油物探,2001,13(4):126-141.
    [2]Hadjidoukas P E,Papatheodorou T S.OpenMP Extensionsfor Master-Slave Message Passing Computing[J].ParallelComputing,2005,31(10-12):1155-1167.
    [3]Chang Herman.3-D Prestack Kirchhoff Depth Migration:From Prototype to Production in a Massively Parallel Proces-sor Enviroment[J].Geophysics,1998,63(2):546-556.
    [4]Dai Hengchang.Parallel Processing of Prestack KirchhoffTime Migration on a PC Cluster[J].Computers and Geosci-ences,2005,31(7):891-899.
    [5]陈莉,竹岛由里子,藤代一成,等.大规模数据场的并行可视化[J].浙江大学学报(理学版),2001,28(2):222-226.
    [6]Perkins W.Parallel Geophysical Processing[J].The LeadingEdg,1996,15(12):1359-1361.
    [7]王宏琳,高绘生.地震并行处理模式与应用框架[J].计算机学报,2001,24(2):202-208.
    [8]陈勇,陈国良,李春生.SMP集群混合编程模型研究[J].小型微型计算机系统,2004,25(10):1763-1767.
    [9]赵永华,迟学斌.基于SMP集群的MPI+OpenMP混合编程模及有效实现[J].微电学与计算机,2005,22(10):7-11.

版权所有:© 2023 中国地质图书馆 中国地质调查局地学文献中心