Immersed Boundary Method for Shared-memory Architectures

In this report, we propose a novel, massively parallelizable algorithm for the immersed boundary method based on the fluid solver of Guermond and Minev [11]. This solver employs a directional-splitting technique that allows the incompressible Navier–Stokes equation to be efficiently parallelized on both shared and distributed memory architectures. An implementation of the numerical scheme was constructed in Python and C/C++, where the bottlenecks were parallelized using OpenMP. The parallelization techniques used in our implementation are discussed. Furthermore, results for a fluid-structure interaction problem are shown along with a corresponding performance study.