High-Performance Scientific Computing

  1. University
  2. Faculties
  3. Faculty of Mathematics, Natural Sciences, and Materials Engineering
  4. Institute of Mathematics
  5. Chairs
  6. High-Performance Scientific Computing
  7. News

Talk: Sustainable research software for accessible high-performance computing

On 27th May 2025, Michael gave a talk titled "Sustainable research software for accessible high-performance computing" during the 39th Workshop on Sustained Simulation Performance (WSSP) at the High-Performance Computing Center Stuttgart (HLRS). As usual, the WSSP was a great event with many experienced scientists exchanging ideas, problems, and solutions on topics revolving around HPC.

Abstract

Modern supercomputers are becoming increasingly heterogeneous, incorporating hardware components from multiple vendors. At the same time, high-performance computing software development has grown more collaborative, often uniting research groups and institutions across different regions. Coupled with the constant addition of new features and performance optimizations, this raises critical questions about sustainability: How can we handle hardware complexity, coordinate diverse development teams, and maintain evolving research software within an academic environment while still writing energy-efficient, high-performance code?

In this talk, we present some of our strategies for tackling these challenges through the Trixi Framework. We introduce Trixi.jl, a high-order numerical simulation environment for conservation laws built in Julia, along with its spin-off packages (TrixiShallowWater.jl, TrixiAtmo.jl) and its sister project, TrixiParticles.jl. We then discuss how we handle software architectures, automation, code reuse, and organizational practices to balance extensibility with accessible high performance on heterogeneous systems. Finally, we point out remaining open questions and outline plans for future development.

Together with Arpit Babbar and Hendrik Ranocha, we have submitted our paper "Automatic differentiation for Lax-Wendroff-type discretizations".

 

arXiv:2506.11719 reproduce me!

 

 

Abstract

Lax-Wendroff methods combined with discontinuous Galerkin/flux reconstruction spatial discretization provide a high-order, single-stage, quadrature-free method for solving hyperbolic conservation laws. In this work, we introduce automatic differentiation (AD) in the element-local time average flux computation step (the predictor step) of Lax-Wendroff methods. The application of AD is similar for methods of any order and does not need positivity corrections during the predictor step. This contrasts with the approximate Lax-Wendroff procedure, which requires different finite difference formulas for different orders of the method and positivity corrections in the predictor step for fluxes that can only be computed on admissible states. The method is Jacobian-free and problem-independent, allowing direct application to any physical flux function. Numerical experiments demonstrate the order and positivity preservation of the method. Additionally, performance comparisons indicate that the wall-clock time of automatic differentiation is always on par with the approximate Lax-Wendroff method.

Search