Preprints
https://doi.org/10.5194/egusphere-2024-2539
https://doi.org/10.5194/egusphere-2024-2539
25 Sep 2024
 | 25 Sep 2024
Status: this preprint is open for discussion.

A GPU-parallelization of the neXtSIM-DG dynamical core (v0.3.1)

Robert Jendersie, Christian Lessig, and Thomas Richter

Abstract. The cryosphere plays a crucial role in Earth’s climate system, making accurate sea ice simulation essential for improving climate projections. To achieve higher resolution simulations, graphics processing units (GPUs) have become increasingly appealing due to their higher floating point peak performance and superior energy efficiency compared to CPUs. However, harnessing the full theoretical performance of GPUs often requires significant effort in redesigning algorithms and careful implementation. Recently, several frameworks have emerged, aiming to simplify general-purpose GPU programming. In this study, we evaluate multiple such frameworks, including CUDA, SYCL, Kokkos, and PyTorch, for the parallelization of neXtSIM-DG, a finite-element-based dynamical core for sea ice. Based on our assessment of usability and performance, CUDA demonstrates the best performance, while Kokkos is a suitable option for its robust heterogeneous computing capabilities. Our complete implementation of the momentum equation using Kokkos achieves a sixfold speedup on the GPU compared to our OpenMP-based CPU code, while maintaining competitiveness when run on the CPU. Additionally, we explore the impact of different discretization orders and the use of lower precision floating-point types on the GPU, showing that switching to single precision can further accelerate sea ice codes.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.
Robert Jendersie, Christian Lessig, and Thomas Richter

Status: open (until 20 Nov 2024)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
Robert Jendersie, Christian Lessig, and Thomas Richter
Robert Jendersie, Christian Lessig, and Thomas Richter

Viewed

Total article views: 55 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
42 9 4 55 0 4
  • HTML: 42
  • PDF: 9
  • XML: 4
  • Total: 55
  • BibTeX: 0
  • EndNote: 4
Views and downloads (calculated since 25 Sep 2024)
Cumulative views and downloads (calculated since 25 Sep 2024)

Viewed (geographical distribution)

Total article views: 45 (including HTML, PDF, and XML) Thereof 45 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 05 Oct 2024
Download
Short summary
Climate change is an important subject for current and future generations. Computer simulations are critical to understand how climate change will effect local communities. An important part of such simulations is sea ice, which affects even distant areas in the long term. In our work, we explore how GPUs, computer chips originally designed for gaming, allow for the faster and more energy efficient simulation of a new computer program for sea ice simulations, the neXtSIM-DG dynamical core.