Enabling mixed-precision with the help of tools: A Nekbone case study

TL;DR

This paper presents a methodology for enabling mixed-precision computing in Nekbone, demonstrating significant reductions in computation time and energy consumption using tools and the roofline model.

Contribution

It introduces a novel approach combining computer arithmetic tools and roofline analysis to effectively implement mixed-precision in a CFD mini-application.

Findings

Reduced time-to-solution by 40.7%.

Lowered energy-to-solution by 47%.

Validated the methodology on 128 MPI ranks.

Abstract

Mixed-precision computing has the potential to significantly reduce the cost of exascale computations, but determining when and how to implement it in programs can be challenging. In this article, we consider Nekbone, a mini-application for the CFD solver Nek5000, as a case study, and propose a methodology for enabling mixed-precision with the help of computer arithmetic tools and roofline model. We evaluate the derived mixed-precision program by combining metrics in three dimensions: accuracy, time-to-solution, and energy-to-solution. Notably, the introduction of mixed-precision in Nekbone, reducing time-to-solution by 40.7% and energy-to-solution by 47% on 128 MPI ranks.