数值预报MAPS模式并行化移植初探

Primary Study on Parallel Transplantation of Numerical Prediction MAPS Model

通过武汉区域气象中心数值预报 MAPS(Mesoscale Analysis and ProcessSystem)模式并行化移植的实例 ,介绍了如何使用共享内存、消息传递的并行方式进行数值预报模式并行化工作 ,对不同条件下并行 MAPS模式的并行效率进行了比较。结果表明 ,对数值预报 MAPS模式完成程序优化和共享内存方式以及消息传递方式的并行化等工作后 ,MAPS模式运算时间大为缩短 ,满足预报业务时效的要求。

It is in this paper introduced how to use SMP and MPI to make numerical prediction model parallel by employing a case in which parallel transplantation of numerical prediction MAPS model was made by Wuhan Regional Meteorological Center.The parallel efficiency of MAPS model is also analysed and compared under different conditions.The result shows that the run time can be greatly cut short and satisfy the operation request after doing serial optimisation and using SMP?MPI to make the MAPS model parallel.