摘要
通过网络计算机群,采用信息传递接口(MPI)为通讯库,建立了分蓄洪区洪水演进计算的并行数值模拟方法.以荆江分洪区为对象,模拟计算了实施分洪后洪水的演进过程.计算表明,网络并行计算系统机群的计算速度与计算规模、进程数等因素有关;并行计算相对于单机计算能够提高计算速度,并且加速比及并行效率会随着计算规模的加大而加大;同时,加速比会随着进程数的增加而先增加后减小,存在着一个最优的进程数.
A numerical simulation method for parallel algorithm is established to simulate the flood rouzing in flood diversion-detention area. The message passing interface is applied. As an example, the flood routing in Jingjiang reaches flood diversion-detention area is simulated. The result shows that the speed of workstation cluster is relevant to the computation scale and the process numbers etc.. Compared with computing in single computer, parallel algorithm can improve computing speed, and the parallel speed-up rate and efficiency also can increase as the computation scale enlarges. Furthermore, the parallel speed-up rate will be increased firstly and decreased later along with the increase of the process numbers; and an optimum process number can be found.
出处
《武汉大学学报(工学版)》
CAS
CSCD
北大核心
2005年第5期24-29,共6页
Engineering Journal of Wuhan University
基金
中国欧盟国际合作ANFAS资助项目
关键词
并行计算
并行程序
MPI
洪水演进
parallel computing
parallel program
message passing interface
flood routing