The Simulation results can be see at ALE-FEM Project
This project focuses on using GPU to accelerate the mesh movement modules that comes in ALE based Finite Element method. In this paper, we propose strategies to calculate the mesh normals asynchronously in GPU by transferring the Finite element Data from CPU to GPU and to factorise the Mesh matrix in GPU asynchronously by overlapping that computation with the Navier Stokes computation in GPU. We have also discussed the timing difference between the direct solver and the sparse GPU solvers and the existing limitations in the lower level sparse Factorisation routines with less host interventions.
A schematic illustrating the proposed timeline for mesh movement computation in GPU.