Contribution to a conference proceedings

FZJ-2026-00840

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png What Will the Grace Hopper-Powered Jupiter Supercomputer Bring for Sparse Linear Algebra?

Tsai, Y.-H. ; Bode, M.FZJ* ; Anzt, H.

2026
ACM New York, NY, USA New York, NY, USA

This record in other databases:  

Please use a persistent id in citations: doi:10.1145/3773656.3773691  doi:10.34734/FZJ-2026-00840

Abstract: The first exascale supercomputer in Europe, JUPITER, is currently being built using the NVIDIA Grace Hopper superchips as main building blocks. JUPITER is designed to provide computing power for both data-driven (AI) workloads and numerics-based simulation workloads. For both workload types, and particularly for PDE-based simulations, high-performance sparse linear algebra operations are crucial. In this paper, we analyze the performance levels that sparse linear algebra operations can achieve on the JUPITER supercomputer and identify algorithmic modifications that can improve performance by acknowledging the Grace Hopper architecture.

---

Contributing Institute(s):

1. Jülich Supercomputing Center (JSC)

Research Program(s):

1. 5112 - Cross-Domain Algorithms, Tools, Methods Labs (ATMLs) and Research Groups (POF4-511) (POF4-511)
2. Inno4Scale - Innovative Algorithms for Applications on European Exascale Supercomputers (101118139) (101118139)

Appears in the scientific report 2026

---

Database coverage:
;

---