Content area

Abstract

The cryosphere plays a crucial role in the 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 compared to central processing units (CPUs). However, harnessing the full theoretical performance of GPUs often requires significant effort in redesigning algorithms and careful implementation. Recently, several frameworks have emerged that aim 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 6-fold speedup on the GPU compared to our OpenMP-based CPU code, while maintaining competitiveness when run on the CPU. Additionally, we explore the use of lower-precision floating-point types on the GPU, showing that switching to single precision can further accelerate sea-ice codes.

Details

1009240
Business indexing term
Title
A GPU parallelization of the neXtSIM-DG dynamical core (v0.3.1)
Author
Jendersie, Robert 1   VIAFID ORCID Logo  ; Lessig, Christian 2   VIAFID ORCID Logo  ; Richter, Thomas 3   VIAFID ORCID Logo 

 Institute of Simulation and Graphics, Otto von Guericke University, Magdeburg, Germany; Institute of Analysis and Numerics, Otto von Guericke University, Magdeburg, Germany 
 Institute of Simulation and Graphics, Otto von Guericke University, Magdeburg, Germany; European Centre for Medium-Range Weather Forecasts, Bonn, Germany 
 Institute of Analysis and Numerics, Otto von Guericke University, Magdeburg, Germany 
Publication title
Volume
18
Issue
10
Pages
3017-3040
Publication year
2025
Publication date
2025
Publisher
Copernicus GmbH
Place of publication
Katlenburg-Lindau
Country of publication
Germany
Publication subject
ISSN
1991962X
e-ISSN
19919603
Source type
Scholarly Journal
Language of publication
English
Document type
Journal Article
Publication history
 
 
Milestone dates
2024-08-12 (Received); 2024-09-25 (Revision request); 2025-01-20 (Revision received); 2025-03-03 (Accepted)
ProQuest document ID
3207630135
Document URL
https://www.proquest.com/scholarly-journals/gpu-parallelization-nextsim-dg-dynamical-core-v0/docview/3207630135/se-2?accountid=208611
Copyright
© 2025. This work is published under https://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Last updated
2025-07-22
Database
ProQuest One Academic