A High-Efficient Scalable Solver for the Global Ocean/Sea-Ice Model MPIOM

This paper presents the work of the "Scalable-Earth-System-Models for high productivity climate simulations" project in improving the linear solver of the barotropic sub-system in the global ocean/sea-ice model MPIOM developed by the Max-Planck-Institute in Hamburg, Germany. We demonstrate the implementation of the conjugate gradient method and an incomplete Cholesky preconditioner with fill-in p in order to achieve high scalability and performance on an IBM POWER6 based supercomputer. Benchmarks of the new solver compared to the traditionally used Successive-Over-Relaxation-Method are given and analyzed with respect to the number of iterations and runtime.