GPGPU Accelerated MAC-SOLA Immersed Boundary Method

A GPU accelerated discrete finite difference based efficient IBM is validated by studying the motion of passive flexible flaps in a two dimensional flow field. A FEM based beam model is used for the deforming flaps and the partitioned coupling method is used to study the FSI. Solution Algorithm (SOLA) and Marker and Cell (MAC) method are coupled to satisfy the mass conservation equations. This coupling avoids the cell regeneration and/or reconstruction algorithms in the moving boundary problems. The coupling algorithm improves the granularity in the estimation of pressure correction near the immersed boundary. The coupled pressure correction solver is accelerated on a NVIDIA P-100 GPGPU using OpenACC. The performance of accelerating the pressure correction solver on the GPU and through multi-core CPUs is also discussed.