A Fast Parallel Large-scale Grid Generator for Parallel Computing in Engineering Simulation

Due to the limitation of memory and time, the traditional serial grid generation algorithms are usually the bottleneck to make the distributed computing be widely used in engineering science. To solve this problem and make the true large-scale parallel simulation in engineering science practical, a fast parallel grid generation algorithm has been proposed and the corresponding grid generator has been developed. Firstly, the initial coarse volume mesh is generated by using advancing front method. Then, two-level domain decomposition is applied and each subdomain is distributed to its corresponding process. By doing so, the intra-node and inter-node communication are distinguished and the communication overheads are reduced. Lastly, the mesh refinement algorithm is performed to generate large-scale grids without communication. Applications of the proposed algorithm and the developed generator for large-scale grid generation on supercomputer (TIANHE-2) are carried out. The reasonably good parallel scalability is obtained and the effectiveness and correctness of the developed generator are verified.