Accelerating Manycore Fluid Flow Predictions Including Wave/Ship Motions Using Parareal

使用 Parareal 加速众核流体流动预测，包括波浪/船舶运动

• 批准号: RGPIN-2017-04247
• 负责人: Gerber, Andrew
• 金额: $ 1.82万
• 依托单位: University of New Brunswick
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=760177
• 关键词: Accelerating; Manycore; Fluid; Flow; Predictions

Exascale high-performance computing trends point to substantial computing resources at a low cost moving into the future, but new software designs and parallelization strategies are required to take full advantage of these emerging manycore hardware systems. Over the past six years, the applicant has led the development of a manycore Computational Fluid Dynamics (CFD) code called EXN/Aero that is now demonstrating significant solution speed-up (using GPUs) versus traditional CPU solutions obtained with spatial domain decompositions. The EXN/Aero development has also included a combined space-time parallel computing approach called parareal. Here, the domain decomposition includes both space and time, and is an emerging research area in exascale computing since it provides a way to reduce computing time when a highly parallel application is memory-bound. There is, however, a need to refine the methodology for high-Reynolds number and multi-physics CFD applications. The applications of choice for the proposed research is in the area of environmental and ship/submarine ocean flows. Building an efficient set 3D unsteady turbulent flow CFD models in these areas is important for comprehensive acoustic source modeling, including subsequent input into a wide range of signatures studies. The applicant has worked for a number of years with Defense Research and Development Canada on their submarine program (developing a range of CFD simulation capabilities including EXN/Aero) and the extension to ships is therefore a natural one. In particular, the proposal will focus on objectives that develop the parareal method for ship simulations that include the important couplings of waves, ship motion and propulsion (including cavitation). All of these aspects are important, when coupled, to predicting the highly turbulent flow environment from which noise sources arise. The research will allow Canadian industry and government to tap into future CFD supercomputing technologies which will be essential to developing complex systems such as ship and submarine platforms, and environmental flows such as that for tidal energy development.

Exascale高性能计算的趋势指向未来以低成本获得大量计算资源，但需要新的软件设计和并行化策略来充分利用这些新兴的众核硬件系统。在过去的六年中，申请人领导了名为EXN/Aero的多核计算流体动力学（CFD）代码的开发，该代码现在证明了与通过空间域分解获得的传统CPU解决方案相比，解决方案速度显著提高（使用GPU）。EXN/Aero的开发还包括一种称为parareal的时空并行计算方法。在这里，区域分解包括空间和时间，是一个新兴的研究领域，在exascale计算，因为它提供了一种方法，以减少计算时间时，一个高度并行的应用程序是内存限制。然而，有必要改进高雷诺数和多物理场CFD应用的方法。拟议研究的应用选择是在环境和船舶/潜艇海洋流动领域。 在这些领域建立一套有效的三维非定常湍流CFD模型对于全面的声源建模非常重要，包括随后输入到广泛的特征研究中。 申请人已与加拿大国防研究与发展部合作多年，致力于其潜艇项目（开发一系列CFD模拟功能，包括EXN/Aero），因此扩展到船舶是很自然的。特别是，该提案将侧重于开发船舶模拟的平行区域方法的目标，其中包括波浪、船舶运动和推进（包括空化）的重要耦合。 所有这些方面都很重要，当耦合时，预测噪声源产生的高度湍流环境。这项研究将使加拿大工业和政府能够利用未来的CFD超级计算技术，这对于开发船舶和潜艇平台等复杂系统以及潮汐能开发等环境流至关重要。