摘要： 基于MPI消息传递实现了多点源高斯扩散模型的并行编程,以数据的分布存储作为区域划分的依据,实现了计算量的负载平衡;通过对算法的改进实现了粗粒度计算,大大减少通信量,提高了程序的执行效率,并对串行计算和并行计算的结果进行了比较。研究结果表明,将并行计算应用于多点源污染物浓度计算是可行的,并获得较高的加速比,并行效率最高达97.5%,与单机相比,不仅可以提高运行速度,大大缩短计算时间,而且可以扩大计算规模。

关键词: 并行计算, 大气污染扩散, 加速比

Abstract: This paper makes an experiment with parallel computing of multi-point sources GUSSI dispersion model,which is on the base of message passing interface(MPI).According to the data distribution storage,it divides the regions and achieves the load balance of computation capacity among computer nodes.By improving the method,it achieves coarse granularity computation.As a result,the traffic becomes less,and the implementation efficiency gets increased.Conclusion can be drawn that it is feasible to apply the parallel computing to multi-point sources pollution concentration computing.Comparing with the CPU time-consuming of single microcomputer,the computing speed is raised and the CPU time-consuming is reduced greatly,the computing scale is also enlarged.

中图分类号: 

• TP399