Contribution to a conference proceedings/Contribution to a book

FZJ-2026-02261

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Score-P with Arm(s) around the world...

Feld, C. (Corresponding author)FZJ* ; Corbin, G.FZJ* ; Wylie, B. J. N.FZJ*

2026
ACM New York NY, USA

This record in other databases:  

Please use a persistent id in citations: doi:10.1145/3784828.3785348  doi:10.34734/FZJ-2026-02261

Abstract: Now that the first Arm-based exascale computing systems have been deployed, it is more important than ever to tune and scale up HPC applications to fully exploit the available hardware resources. Therefore, there is a strong need for software tools that can assist application developers with this task. The Score-P performance tools ecosystem plays a major role in filling this gap. It consists of the highly scalable Score-P instrumentation and measurement infrastructure for profiling and event tracing of massively parallel HPC application codes, as well as complementary analysis tools that work on the common data formats CUBE4 and OTF2, allowing users to gain insights into their applications’ communication, synchronization, input/output, and scaling behavior, pinpointing performance bottlenecks and their causes. Score-P aims to be easy to use, originating from x86-64, SPARC and POWER systems and now also supporting Arm-based systems. In this article, we detail our experiences configuring and installing Score-P with various compiler and MPI combinations on the Fujitsu A64FX systems Fugaku and Deucalion, as well as the Nvidia Grace-Hopper exascale system, JEDI/JUPITER. Moreover, we explore differences in run time and measurement overheads between the systems and toolchains. Finally, we present the results of a performance assessment of the Neko CFD code on JEDI, leveraging the Scalasca, CubeGUI and Vampir analysis tools. We thus demonstrate that Score-P is ready to be used in the current Arm environments, identify remaining limitations and further opportunities to improve the software.

---

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. JLESC - Joint Laboratory for Extreme Scale Computing (JLESC-20150708) (JLESC-20150708)
3. POP3 - Performance Optimisation and Productivity 3 (101143931) (101143931)
4. ATMLPP - ATML Parallel Performance (ATMLPP) (ATMLPP)

Appears in the scientific report 2026

---

Database coverage:
;

---