In addition to the technical programming, we will be presenting our last year of work at the EuroHPC Joint Undertaking booth J30 on Wednesday afternoon. Come hear about our latest work and then head over to Niclas' talk on Neko later in the afternoon.
At ISC High-Performance 2024, we present Neko - a portable framework for high-fidelity spectral element flow simulations. Unlike prior work, Neko adopts a modern object-oriented Fortran 2008 approach, allowing multi-tier abstractions of the solver stack and facilitating various hard- ware backends ranging from general-purpose processors, accelerators down to exotic vector processors and Field Programmable Gate Arrays (FPGAs) via Neko’s device abstraction layer. Focusing on Neko’s performance and exascale readiness, we outline the optimisation and algorithmic work necessary to ensure scalability and performance portability across a wide range of platforms. Finally, we present performance measurements on a wide range of accelerated computing platforms, including the EuroHPC pre-exascale systems LUMI and Leonardo, where Neko achieves excellent parallel efficiency for an extreme-scale direct numerical simulation (DNS) of turbulent thermal convection using up to 80% of the entire LUMI supercomputer.
If you have a workshop pass, please come hear about application development in the European HPC Centres of Excellence including CEEC!
"this ISC workshop has been organised to provide a forum to consider common challenges, ideas, solutions, and opportunities from the point of view of HPC applications developers preparing for exa-scale."
If you'll be at ECCOMAS 2024, check out 'Neko: A modern, Portable, and Scalable Framework for High-Fidelity Computational Fluid Dynamics' presented by Niclas Jansson on the terrace.CEEC will have a robust presence at ECCOMAS24 this summer in Lisbon, Portugal! Over the course of the conference, we will be represented in three mini-simposium talks ranging from progress with our wind turbine lighthouse case to our latest work on improving FLEXI.
Machine learning (ML) in scientific applications including computational fluid dynamics (CFD) is a growing field of research. However, ML can be less stable and more prone to errors in CFD because of its complexity relative to e.g. game theory. Thus, recent research has concentrated on reinforcement learning (RL) or physics-informed methods applied to CFD. Another continuously growing field of research which alleviates the common problems of ML in CFD is physics-informed neural networks (PINNs). Recently, modified versions of classical PINNs have been proposed to push their limitations and make them more tailored to CFD. With these considerations in mind, this minisymposium will discuss the applicability, predictive performance and limitations of state-of-the-art ML methods in CFD.
If you're not in the mini-symposium with Anna, make sure to come see Samuel Kemmler's talk 'A Fluid-Solid Coupled Micromechanical Simulation of Piping Erosion During the Installation of a Suction Bucket for the Foundation of an Offshore Wind Turbine'. It is part of the mini-symposium 020 - Advancements in Offshore Wind Structures , which will be held in Room 2.02.
If you're not in the mini-symposium with Samual, make sure to see the talk 'Entropy stable subcell shock capturing scheme for high-order discontinuous Galerkin methods on moving meshes' presented by Anna Schwarz in room 1.14.