Parallelization on GPU of Mesh Generation Using Dynamic Bubble System
Résumé
The meshing method based on a dynamic bubble system is known for its ability to generate high quality meshes. However, the computation time remains a problem since the motion convergence is slow and it increases drastically with the number of nodes. Our goal is to significantly reduce the simulation time of bubbles motion by using a massively parallel computing based on the CUDA NVIDIA architecture. In this paper, we explain in detail the parallel computations and algorithms for the bubble packing meshing method and give the main results obtained.