Parallel Spectral Element Methods for Viscous Flow Problems in General Domains

本课题旨在为高性能分布式存储并行处理器开发一种先进的计算流体动力学（CFD）软件包。这套代码将基于不可压缩Navier-Stokes方程的谱元方法，目标问题类别将是一般二维和三维区域内的流动。选择分布式内存并行架构的动机是它们的高性能和经济性，这源于它们内在的利用数据局部性的能力。这项工作将是首席研究员过去在平行光谱元素方法方面工作的延伸，重点是增加现有软件的几何灵活性和功能，通过合并不符合离散化和添加运行时可视化功能。

The proposed research is to develop an advanced computational fluid dynamics (CFD) software package for high-performance- distributed memory parallel processors. The suite of codes will be based on the spectral element method for the incompressible Navier-Stokes equations, and the target class of problems will be flow in general two- and three-dimensional domains at transitional Reynolds numbers. The choice of distributed memory parallel architectures is motivated by their high performance and economy which derive from their intrinsic ability to exploit data locality. The work will be an extension of the principal investigator's past work in parallel spectral element methods, with emphasis on increasing the geometric flexibility and functionality of existing software, through incorporation of nonconforming discretizations and addition of run-time visualization capabilities.