摘要
基于MPI消息传递实现了小浪底水库三维数学模型的并行编程,采取网格分区和主从模式实现了并行计算,在全局网格和局部区域之间建立了映射关系,并对临界单元和进出口单元等特殊点进行了处理。结果表明:神威-SZNM集群的8CPU并行计算的加速比为5.2,远高于原串行程序在Windows操作系统下采用CVF编译器编译执行的计算速度,极大地缩短了计算时间,提高了计算准确度。
Based on the MPI message passing,it realized Xiaolangdi Reservoir three-dimensional mathematical model of parallel programming,using parallel computing grid partition and the main slave mode,the mapping relationship between the global grid and the local area,and the critical unit and into the the export unit and other special points had been dealt with.The results show that the accelerated the the divinity of SZNM cluster 8CPU parallel computing ratio is 5.2,which is much higher than the speed of original serial program of the CVF compiler under the Windows operating system speed,dramatically reducing the computation time and increasing the calculation accuracy degrees.
出处
《人民黄河》
CAS
北大核心
2012年第5期25-27,共3页
Yellow River
基金
中央级公益性科研院所基本科研业务费专项(HKY-JBYW-2009-18)
关键词
METIS
MPI
并行计算
三维数学模型
小浪底水库
METIS
MPI
parallel computing
three-dimensional mathematical model
Xiaolangdi Reservoir
