Parallelization of a finite volume CFD code

This paper concerns the parallelization of an in-house three-dimensional unstructured finite volume computational fluid dynamics (CFD) code (FJ3SOLV). This is part of the code development project for store release trajectory prediction at the Institute for Aerospace Research of the National Research Council Canada. The parallelization was based on the Data Decomposition method. The additive Schwarz Domain Decomposition method was used to obtain solutions in the subdomains. The domain of interest (unstructured grid) was partitioned into a number of balanced subdomains using the ParMETIS library. The message passing interface (MPI) system was used to perform parallel computation and communication between the processors. A parallel high performance aerodynamics code (HiPAC) was developed. Parallel computations of inviscid flows around a Delta wing and the CF-18 aircraft were carried out using different numbers of processors. Excellent scalability and very good quality solutions were obtained.