摘要
在科学计算领域,网格生成的规模及其速度和质量已经成为制约求解速度和精度的瓶颈,快速生成高质量的网格在许多领域需求迫切。本文针对现代并行计算机多级体系结构,研究共享内存模型下的并行Delaunay网格生成算法,采用任务并行策略,通过OpenMP并行模型实现Delaunay算法的并行。实验表明,该并行算法和串行算法最终生成的网格在质量上的区别很小,而同时其网格生成效率也可以满足大规模网格生成的需求。
The problem on the scale and quality of generated meshes has become a bottleneck to achieve precision and performance in scientific and industrial computation. In many applications, fast generation of high quality mesh is of essential importance. In this paper, we present a parallel algorithm for Delaunay triangulation exploiting the locality and concurrency of the mesh generating operations. The algorithm employes a task-parallel strategy and is implemented using OpenMP. Experimental results indicate that the algorithm is efficient and capable of generating high quality meshes.
出处
《科研信息化技术与应用》
2012年第5期20-28,共9页
E-science Technology & Application
基金
国家自然科学基金(91130019)
国家高技术研究发展计划(863计划)(2012AA01A304)
关键词
网格生成
DELAUNAY算法
任务并行
Unstructured mesh generation
Delaunay algorithm
Task parallelization plan
