English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Thanks to the computational power of modern cluster machines, numericalsimulations can provide, with an unprecedented level of details, new insightsinto fluid mechanics. However, taking full advantage of this hardware remainschallenging since data communication remains a significant bottleneck toreaching peak performances. Reducing floating point precision is a simple andeffective way to reduce data movement and improve the computational speed ofmost applications. Nevertheless, special care needs to be taken to ensure thequality and convergence of computed solutions, especially when dealing withcomplex fluid simulations. In this work, we analyse the impact of reduced(single and mixed compared to double) precision on computational performanceand accuracy for computational fluid dynamics. Using the open source libraryOpenFOAM, we consider incompressible, compressible, and multiphase fluidsolvers for testing on relevant benchmarks for flows in the laminar andturbulent regime and in the presence of shock waves. Computational gain andchanges in the scalability of applications in reduced precision are alsodiscussed. In particular, an ad hoc theoretical model for the strong scalingallows us to interpret and understand the observed behaviors, as a function offloating point precision and hardware specifics. Finally, we show how reducedprecision can significantly speed up a hybrid CPU-GPU implementation, madeavailable to OpenFOAM end-users recently, that simply relies on a GPU linearalgebra solver developed by hardware vendors.

On floating point precision in computational fluid dynamics using OpenFOAM