Abstract:
Objective The computation task in high resolution Earth's gravity field modeling is very huge, as the dataset and the parameters are dramatically. We focus on high performance and scientific computation method in high‑resolution Earth's gravity field modeling to make model determination feasible.
Methods The basic methods we used for high performance and computing are the semi-analysis method, open multi-processing (OpenMP) and message passing interface (MPI) parallel computing method. A new high performance and scientific computing method is designed. The semi-analysis method is also known as the lumped-coefficients method. The OpenMP and MPI technique are to make use of the huge-computing machine. And afterward the method is validated by numerical experiments on computing error and computing efficiency.
Results The semi-analysis can effectively reduce the computation and storage consumption in high‑resolutoin gravity field modeling. The time consumption of OpenMP+MPI hybrid parallel computing is only 1/18.4 that of sequential computing with 20 cores. The proposed computing method with semi-analysis and OpenMP+MPI computing can be used on super computers for Earth's gravity field modeling, and the modeling of coefficients up to degree and order 5 399 can be completed within one day using 80 central processing unit on 4 nodes.
Conclusions High performance and scientific computing is useful for high-resolution gravity field modeling, the method need to be improved tested for finer resolution field gravity field modeling.
下载:
