首页> 外文会议>AIAA Computational Fluid Dynamics Conference >Development of Stencil-Based Mesh Partitioning for Parallel Unstructured CFD Solvers
【24h】

Development of Stencil-Based Mesh Partitioning for Parallel Unstructured CFD Solvers

机译:平行非结构化CFD求解器的模板基网眼分区的发展

获取原文

摘要

A study evaluating unstructured mesh partitioning for computational fluid dynamics (CFD) simulations on parallel computers is presented. Considerations for effective partitioning of computational unstructured meshes for a family of implicit time-integration methods using the line-relaxation algorithm are outlined. Mesh partitioning using the Metis library is evaluated for two- and three-dimensional meshes. Different combinations of computational stencil information provided to the partitioning library are outlined and comparison of the resulted load balance and communication volume measures are presented. An augmentation to the existing mesh partitioning method via a heuristic algorithm is proposed. The heuristic algorithm relies on an initial partitioning of the mesh, which is then refined iteratively. Results using the new method are compared with the standard partitioning methods. This is done both from a theoretical perspective and by running the CFD code. Elapsed time results of the CFD code are presented from simulations on a parallel computer. Results show that the improved partitioning using the stencil information and the heuristic algorithm result in faster turn-around times. The stencil-based partitioning and the proposed heuristic algorithm are simple modifications to existing codes and are applicable to general simulations employing domain decomposition.
机译:呈现了对并行计算机上的计算流体动力学(CFD)模拟的评估评估非结构化网格分区。概述了用于使用线放松算法的隐式时间集成方法系列的计算非结构化网格的有效分区的考虑。使用Metis库进行网格分区,用于两维网格。概述了提供给分区库的计算模板信息的不同组合,并呈现了所产生的负载平衡和通信体积测量的比较。提出了通过启发式算法的现有网格分区方法的增强。启发式算法依赖于网格的初始分区,然后迭代地改进。使用新方法的结果与标准分区方法进行比较。这是从理论透视和运行CFD代码来完成的。在并行计算机上的模拟中呈现了CFD代码的经过时间结果。结果表明,使用模板信息和启发式算法的改进分区导致旋转时间更快。基于模板的分区和所提出的启发式算法对现有代码简单地修改,并且适用于采用域分解的一般模拟。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号