Bachelor Thesis

FZJ-2024-06525

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Extending a Performance Analysis Tool to Handle MPI Message Probing

Haus, K. (Corresponding author)FZJ*

2024

This record in other databases:

Please use a persistent id in citations: doi:10.34734/FZJ-2024-06525

Abstract: In the realm of high-performance computing (HPC), exascale systems have now reached a point of practical reality and offer unprecedented computing power. As these systems become more powerful, it is crucial to ensure that users can effectively harness this power. It is essential to consider scalability and performance optimization when developing applications running on such systems to be able to fully leverage their capabilities. Performance analysis is a fundamental aspect of optimizing parallel applications. This thesis addresses a specific gap in the Scalasca performance analysis tool: MPI message probing. Message probing is useful to determine the required buffer size for a pending message. This work presents an extended event model that captures probe calls and integrates them into Scalasca’s analysis framework. The enhanced tool is evaluated across various test cases, demonstrating its capability to identify inefficiencies related to MPI message probing, especially the Late Sender wait state. The results show that the extended analysis provides a more comprehensive insight into application behavior.

Note: Bachelorarbeit, FH Aachen - University of Applied Sciences, 2024

---

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. ATMLPP - ATML Parallel Performance (ATMLPP) (ATMLPP)

Appears in the scientific report 2024

---

Database coverage:

---