SGER: Scaling a CFD Code for Ocean Processes to 1000s of Processors

SGER：将海洋过程的 CFD 代码扩展到数千个处理器

• 批准号: 0738004
• 负责人: Kraig Winters
• 金额: --
• 依托单位: University of California-San Diego Scripps Inst of Oceanography
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2007
• 资助国家: 美国
• 起止时间: 2007-07-01 至 2008-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0738004&HistoricalAwards=false
• 关键词: SGER; Scaling; CFD; Code; Ocean

Intellectual MeritThis exploratory project is to modify an existing fluid dynamics code to make it suitable for parallel execution on thousands of processors and to quantify the resulting scalability using performance analysis tools. The code is designed for high-resolution, three-dimensional simulations of stratified fluid flow and is typically used for process oriented studies in small-scale physical oceanography, atmospheric science and fluid dynamics. The code uses MPI and has been run on laptops, single and dual processor workstations, workstation clusters with various interconnect technologies, several high performance comptuers. The maximum number of processors on which it has been run is 768 on the National Center for Atmospheric Research (NCAR) Blue Ice machine. The primary activities for which support is requested are to incorporate a two-dimensional data decomposition and to utilize performance analysis tools to assess performance and document scalability.Broader ImpactsThe primary goal of this effort is to produce a well-tested, flexible fluid dynamics code based on high order numerical techniques that scales well to new high performance architectures. The source code and related tools will be made freely available and will be useful to researchers in oceanography, mathematics, fluid dynamics, physics and engineering. This effort will also enhance the capabilities of the open-source parallel, 3D FFT library distributed by SDSC by incorporating additional globally-based differentiation schemes.

智力MeritThis探索性项目是修改现有的流体动力学代码，使其适合在数千个处理器上并行执行，并使用性能分析工具量化所产生的可扩展性。该代码是为分层流体流的高分辨率三维模拟而设计的，通常用于小规模物理海洋学、大气科学和流体动力学中面向过程的研究。该代码使用MPI，并已运行在笔记本电脑，单处理器和双处理器工作站，工作站集群与各种互连技术，几个高性能计算器。在美国国家大气研究中心（NCAR）的蓝冰机器上，运行它的最大处理器数量是768。支持的主要活动是要求将一个二维的数据分解，并利用性能分析工具来评估性能和文件scalability.Broader ImpactsThe这项工作的主要目标是产生一个良好的测试，灵活的流体动力学代码的基础上高阶数值技术，以及新的高性能架构的规模。源代码和相关工具将免费提供，对海洋学、数学、流体动力学、物理学和工程学研究人员有用。这一努力还将通过纳入额外的基于全局的差分方案来增强SDSC分发的开源并行3D FFT库的功能。