Score-P

The instrumentation and measurement framework Score-P, together with analysis tools build on top of its output formats, provides insight into massively parallel HPC applications, their communication, synchronization, I/O, and scaling behaviour to pinpoint performance bottlenecks and their causes.

408
mentions
39
contributors

Cite this software

What Score-P can do for you

The Score-P instrumentation and measurement infrastructure is a highly scalable and easy-to-use tool suite for profiling (summarizing program execution) and event tracing (capturing events in chronological order) of HPC applications.

The scorep instrumentation command adds instrumentation hooks into a user's application by either prepending or replacing the compile and link commands. C, C++, Fortran, and Python codes as well as contemporary HPC programming models (MPI, threading, GPUs, I/O) are supported.

When running an instrumented application, measurement event data is provided by the instrumentation hooks to the measurement core. There, the events are augmented with high-accuracy timestamps and potentially hardware counters (a plugin-API allows querying additional metric sources). The augmented events are then passed to one or both of the built-in event consumers, profiling and tracing (a plugin-API allows creation of additional event consumers) which finally provide output in the formats CUBE4 and OTF2, respectively. These open and backwards-compatible output formats can be consumed by established analysis tools, e.g., like

  • CubeGUI (RSD entry), the performance report explorer for Scalasca and Score-P, a generic tool for displaying a multidimensional performance space,
  • Extra-P, an automatic performance-modelling tool that supports the user in the identification of scalability bugs,
  • TAU's ParaProf, a portable, scalable performance analysis tool, and PerfExplorer, a framework for parallel performance data mining and knowledge discovery,
  • Scalasca Trace Tools (RSD entry), a collection of trace-based performance analysis tools that have been specifically designed for use on large-scale systems featuring hundreds of thousands of CPU cores, automatically identifying potential communication and synchronization bottlenecks and offering guidance in exploring their causes, and
  • Vampir, a trace-based framework that enables users to quickly display and analyse arbitrary program behaviour.

The above can be summarized in this picture:
Score-P overview

Even without the analysis tools, insights into a profiling experiment can be obtained by scorep-score, giving the user the opportunity to reconfigure a measurement, e.g., by filtering events that are of no value for an intended analysis. An instrumented application can be run in several configurations, just by setting Score-P environment variables. A list of configuration variables, recognized by a given Score-P installation, is available via the scorep-info command.

Early work on concepts for the first BMBF and DOE funded Score-P projects (starting 2009) originated from the Helmholtz University Young Investigators Grant and the Helmholtz Virtual Institute – High Productivity Supercomputing (both Felix Wolf).

The first Score-P release v1.0 was announced in 2011. From there on, Score-P is continually maintained and improved (using a meritocratic governance model), staying up-to-date with HPC trends.

Early on, the distributed development team invested in automated builds, tests, and deployment (CICD) from a source code repository to improve the portability of Score-P, a must-have in the non-standard HPC world. Nowadays, GitLab's CICD automation creates public ready-to-install stable and work-in-progress development tarballs, releases and documentation. It detects portability and coding convention issues early during review-based feature development.

Score-P grew into a big player in the tool's community and is attracting HPC vendors to fund implementation work to support their latest hardware.

Score-P is widely adopted at many HPC sites across the world and particularly documented its strength in several performance assessments during the EU H2020 POP CoE projects. Since more than a decade, its usage is taught at numerous VI-HPS Tuning Workshops, conference tutorials, and international HPC summer schools.

Accessing Score-P

Please find the latest ready-to-use tarballs here: https://perftools.pages.jsc.fz-juelich.de/cicd/scorep/

The public GitLab repository, a mirror of the development repository, is located here: https://gitlab.com/score-p/scorep

Getting in contact

If you have any comments or questions regarding the use and installation of Score-P, want to report a bug you discovered, or have a feature request, please email support@score-p.org

Staying up-to-date

You can also sign up to the Score-P News mailing list to receive the latest news about releases, tutorials, workshops, and other Score-P-related events.

Citing Score-P

If you find Score-P helpful for your research, please mention the Score-P software DOI as well as the Score-P overview paper in your publications:

  • Knüpfer, A. et al. (2012). Score-P: A Joint Performance Measurement Run-Time Infrastructure for Periscope, Scalasca, TAU, and Vampir. In: Brunst, H., Müller, M., Nagel, W., Resch, M. (eds) Tools for High Performance Computing 2011. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-31476-6_7

Acknowledgements

This work is supported by:

Score-P funding

Logo of Score-P
Keywords
Programming languages
  • C 74%
  • C++ 10%
  • M4Sugar 6%
  • Makefile 4%
  • Shell 2%
License
  • BSD-3-Clause
</>Source code

Participating organisations

Forschungszentrum Jülich
TU Dresden
RWTH Aachen University
German Aerospace Center (DLR)
Technische Universität München
Technical University of Darmstadt
University of Oregon

Reference papers

Mentions

Contributors

CF
Christian Feld
DL
Daniel Lorenz
DS
Dirk Schmidl
RT
Ronny Tschüter
German Aerospace Center (DLR)
YO
Yury Oleynik
DE
Dominic Eschweiler
JS
Johannes Spazier
SS

Related software

CubeGUI

CU

Cube, which is used as performance report explorer for Scalasca and Score-P, is a generic tool for displaying a multi-dimensional performance space consisting of the dimensions (i) performance metric, (ii) call path, and (iii) system resource.

Updated 11 months ago
83 7

CubeLib

CU

Cube, which is used as performance report explorer for Scalasca and Score-P, is a generic tool for displaying a multi-dimensional performance space consisting of the dimensions (i) performance metric, (ii) call path, and (iii) system resource.

Updated 11 months ago
11 4

CubeW

CU

Cube, which is used as performance report explorer for Scalasca and Score-P, is a generic tool for displaying a multi-dimensional performance space consisting of the dimensions (i) performance metric, (ii) call path, and (iii) system resource.

Updated 11 months ago
11 4

Scalasca

SC

The Scalasca Trace Tools support performance optimization of parallel programs with a collection of highly scalable trace-based tools for in-depth analyses of concurrent behavior, in particular with respect to communication and synchronization, and offers guidance in exploring their causes.

Updated 11 months ago
107 16