针对大多数并行Delaunay网格生成算法无法充分利用多核共享内存结构的问题,在原有面向共享内存的二维并行算法基础上,根据三维问题的特点提出基于OpenMP的三维并行Delaunay网格生成算法。算法采用划分求解区域为方格的方法实现候选点集的划分和并行插点。使用OpenMP对算法进行实现,并利用多种实现技术避免线程间的同步等待,提升算法效率。实验结果表明,该算法及所采用的实现技术可以在三维条件下快速生成大量网格单元,具有较高的并行效率,同时能够保证良好的网格质量。
For the fact that most parallel Delaunay mesh generation algorithms can' t suit the shared memory structure well, this paper proposed a parallel Delaunay mesh generation algorithm using OpenMP. It was based on the existing 2D parallel algorithm for shared memory structure and took into account the characteristics of the problem in 3D. The presented algorithm divided the domain into cubes in order to partition the candidate point set and inserted points in parallel. The algorithm was im- plemented with OpenMP. It used several methods to avoid the synchronization among threads and increased the efficiency of the algorithm. The experimental results show that the algorithm and implementation methods can rapidly generate large-scale mesh elements. It has a relatively good parallel efficiency together with good mesh quality.