Adapting and optimising Fluidity for high-fidelity coastal modelling. (30th May 2018)

• Record Type: Journal Article
• Authors: Creech, Angus C.W.; Jackson, Adrian; Maddison, James R.
• Is Part Of: Computers & fluids. Volume 168(2018)
• Page Start: 46
• Page End: 53
• Publication Date: 2018-05-30
• Subjects: Fluidity -- Load balancing -- Mesh decomposition -- Code optimisation -- Parallel performance; Fluid dynamics -- Data processing -- Periodicals; 532.050285
• DOI: 10.1016/j.compfluid.2018.03.066
• Languages: English
• ISSNs: 0045-7930

Abstract

Highlights: Optimised a CFD application to provide 3x the performance for tidal modelling. Mesh re-ordering significantly improves performance without requiring code changes. Load balancing domain decomposition across parallel processes improves performance. Abstract: Work undertaken to improve the performance of Fluidity, an open-source finite-element computational fluid dynamics solver from Imperial College London, for both general computational fluid dynamics and tidal modelling problems is outlined. Optimising the general computational structure of Fluidity, along with work to improve the data decomposition and parallel load balancing enabled simulations to be run overthree times faster than with the original code, even when using thousands of computational cores. This changes the level of detail at which fluids problems can be studied with Fluidity, and impacts upon research that examines high Reynolds number turbulent flows. This is of particular relevance in areas such as engineering aerodynamics, wind energy, marine energy, and environmental or pollution modelling.