Dataflow Computing for Engineering Fluid Simulations (DFC4EFS)

点击翻译按钮获取中文摘要

HPC has given many companies the ability to perform more complex fluid simulation on large clusters of CPU or GPU based nodes. These simulations consume large amounts of power, but are required for engineering more efficient products. This project will explore the use of the innovative Dataflow Computing paradigm, pioneered by Maxeler Technologies with their Dataflow Engines, to significantly increase the performance of CFD simulation, while improving on the energy efficiency of the whole simulation process. The main objective is therefore to map an engineering CFD code onto Maxeler's Dataflow Engines and demonstrate through the running of typical turbomachinery testcases the improvements in energy consumption of the simulation while affording a step-change in computational performance. The innovative aspects of the project are (1) the use of Dataflow computing programming paradigm for CFD; (2) the deployment of Dataflow Engines in an industrial strength engineering simulation application.