The Score-P Performance Tools Ecosystem

Christian Feld^1*Alexandru Calotoiu²Gregor Corbin¹Markus Geimer¹Marc-André Hermanns³Maximilian Knespel⁴Bernd Mohr¹Jan André Reuter¹Maximilian Sander⁴Pavel Saviankou¹Marc Schlütter¹Robert Schöne⁴Sameer S. Shende⁵Anke Visser¹Bert Wesarg^4,6William R. Williams⁴Felix Wolf⁷Brian J. N. Wylie¹Mikhail Zarubin⁴

• ¹Jülich Supercomputing Centre, Forschungszentrum Jülich GmbH, Jülich, Germany
• ²Scalable Parallel Computing Laboratory, Department of Computer Science, ETH Zürich, Zürich, Switzerland
• ³IT Center, RWTH Aachen University, Aachen, Germany
• ⁴Information Services and High Performance Computing, Center for Interdisciplinary Digital Sciences, TUD Dresden University of Technology, Dresden, Germany
• ⁵Performance Research Laboratory, OACISS, University of Oregon, Eugene, OR, United States
• ⁶GWT-TUD GmbH, Dresden, Germany
• ⁷Department of Computer Science, Technical University of Darmstadt, Darmstadt, Germany

With the first exascale computing systems in production use, tuning and scaling up HPC applications to fully exploit the available hardware resources is more important than ever. Thus, there is a strong need for software tools that assist application developers with this task. The Score-P instrumentation and measurement infrastructure plays a major role in filling this gap. Score-P is a community-driven, highly scalable tool suite, aimed to be easy to use, for profiling and event tracing of massively parallel HPC application codes. It provides measurement data via common data formats as well as runtime interfaces for a variety of complementary analysis tools developed by multiple institutions and companies, allowing users to gain insights into the communication, synchronization, input/output, and scaling behavior of their applications, pinpointing performance bottlenecks and their causes. In this article, we provide an overview of the current state of the Score-P infrastructure and its related tools ecosystem Cube, Extra-P, TAU, Scalasca, and Vampir. In particular, we detail Score-P's current design and architecture, which is both highly flexible and extensible. Moreover, we describe how Score-P interacts with the analysis tools mentioned above, and highlight the major extensions implemented over the past 10+ years to keep pace with the rapidly changing landscape of HPC hardware and parallel application programming interfaces. Furthermore, we discuss emerging challenges—in particular with respect to the ever growing heterogeneity in both hardware and software—for collecting and analyzing performance data from applications running on future top-tier computing systems.