Parallel implementation for a mhd solver with equilibrium chemistry

Abstract

A parallel implementation for a hypersonic MHD solver was developed. The original serial code utilizes a 4th-order modified Runge-Kutta scheme with TVD to solve the MHD equations for hypersonic flow regimes. To extend the capabilities of the serial solver, both a purely MPI parallel model and a hybrid MPI/OpenMP model were implemented and tested. The strong and weak scaling of the MPI-only implementation were subsequently investigated. Significant performance increases were observed, as is expected in a parallel implementation. However, the solver exhibited poor parallel efficiency for larger numbers of processes, which could be attributed to a communication intensive numerical scheme. The maximum speedup obtained was 128 on 256 cores. The addition of OpenMP did little to improve the performance of the MPI implementation except for cases with low numbers of cores. To extend the investigation of scalability to larger problems, the parallel program was then be used to simulate a double compression corner problem at Mach 9. For this larger problem, near ideal weak scaling was observed for up to thousands of cores.