A Scalable Implementation of a Finite-Volume Dynamical Core in the Community Atmosphere Model

W.B. Sawyer (Switzerland) and A.A. Mirin (USA)


numerical weather prediction, parallel pro gramming, MPI-2


A distributed memory message-passing parallel implemen tation of a finite-volume discretization of the primitive equations in the Community Atmosphere Model is pre sented. Due to the data dependencies resulting from the po lar singularity of the latitude-longitude coordinate system, it is necessary to employ two separate domain decomposi tions within the dynamical core. Data must be periodically redistributed between these two decompositions. In addi tion, the domains contain halo regions that cover the near est neighbor data dependencies. A combination of several techniques, such as one-sided communication and multi threading, are presented to optimize data movements. The resulting algorithm is shown to scale to very large machine configurations, even for relatively coarse resolutions.

Important Links:

Go Back