Further development of the parallel program complex of SL-AV atmosphere model

Abstract

The SL-AV global semi-Lagrangian atmosphere model is applied to the operational medium-range weather forecast at Hydrometeorological center of Russia. The works on increasing the code scalability and using future computer architectures are described. The scalable parallel multigrid algorithm for solving the linear algebraic equations systems is implemented. It is expected that the multigrid algorithm will be used instead of direct algorithm based on fast Fourier transforms requiring global communications. The results for convergence and strong scalability of the multigrid method are given. The parallel scalability of the low-resolution versions of the SL-AV model for both seasonal and climate simulation has been evaluated at computer systems based on Intel Xeon Phi 2 (Knights Landing) processors. The results show a practical possibility to use these processors for the global atmosphere modelling with the efficiency comparable to the classical cluster systems.