Massively Parallel Computational Fluid Dynamics with Julia and Trixi.jl
DescriptionHigh performance is often cited as one of the core design goals for the Julia programming language. While there is ample evidence that this is true for serial computations, there are far fewer examples of parallel computations in traditional HPC areas such as computational fluid dynamics. In this talk, we will therefore present massively parallel scaling results for Trixi.jl, a numerical simulation framework for conservation laws written in Julia. We will demonstrate the ability to scale a flow simulation to more than 50,000 CPU cores and illustrate some of the Julia-specific challenges we needed to overcome to achieve this. Furthermore, we will discuss some lessons learned while developing a research code in Julia for computational science at scale.
TimeWednesday, June 2814:30 - 15:00 CEST
Computer Science, Machine Learning, and Applied Mathematics