Score-P

Score-P is an instrumentation and measurement framework, that, 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 behavior to pinpoint performance bottlenecks and their causes. It 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.

For a brief overview, please visit Score-P on the Helmholtz Research Software Directory.

For a detailed description, please refer to the article The Score-P performance tools ecosystem.

---

Everyone can become a contributor. Please see CONTRIBUTING.md for details.

---

You can build and install Score-P using ready-to-use tarballs or a Git checkout. For both options, please refer to INSTALL for a complete list of build options, and consult OPEN_ISSUES for known deficiencies.

Tarballs are availble at

• https://score-p.org/ (releases, dependencies as separate packages)
• https://perftools.pages.jsc.fz-juelich.de/cicd/scorep/ (releases, stable development)
• https://perftools.pages.jsc.fz-juelich.de/cicd/scorep/development.html (development snapshots)

The URL for the public Git repository is https://gitlab.com/score-p/scorep.

Build from a downloaded source package

Download a tarball and just do the following:

$ tar xf scorep-<version>.tar.gz
$ mkdir scorep-<version>/_build
$ cd scorep-<version>/_build
$ ../configure --prefix=<PREFIX> <OPTIONS…>
$ make
$ make install

If not already in $PATH, this will also install the Score-P dependencies CubeW, CubeLib, OTF2, and OPARI2 under PREFIX. Add PREFIX/bin to $PATH.

The configure output, also accessible via the file scorep.summary, will provide information about compilers used and components found. Rerun configure with different options until satisfied.

After make install you will find html and pdf documentation as well as usage examples under PREFIX/share/doc/scorep.

Build from Git checkout

If building and installing from the Git checkout, the following requirements are needed (assuming PREFIX/bin is in $PATH):

• AfS-dev (patched versions of Autotools):

  Install via

  $ wget https://perftools.pages.jsc.fz-juelich.de/utils/afs-dev/afs-dev-latest.tar.gz
  $ tar xf afs-dev-latest.tar.gz
  $ cd afs-dev-latest
  $ ./install-afs-dev.sh --continue-after-download --prefix=PREFIX

• Perftools-dev (code beautification and documentation generation):

  Install via

  $ wget https://perftools.pages.jsc.fz-juelich.de/utils/perftools-dev/perftools-dev-latest.tar.gz
  $ tar xf perftools-dev-latest.tar.gz
  $ cd perftools-dev-latest
  $ ./install-perftools-dev.sh --continue-after-download --prefix=PREFIX

• OPARI2, OTF2, CubeW, CubeLib

  Download from https://score-p.org/, extract, and install into PREFIX.

Once the requirements are set up, build and install like this:

$ ./bootstrap
$ mkdir _build
$ cd _build
$ ../configure --prefix=PREFIX <further options, see INSTALL>
$ make
$ make install

Build using containers

Docker images are provided to ease development for common frameworks.

Images are published in DockerHub and are available for the following base images:

• Ubuntu: scorep/devel-gcc and scorep/devel-clang
• NVIDIA HPC SDK: scorep/devel-nvhpc
• Intel oneAPI: scorep/devel-oneapi
• AMD ROCm: scorep/devel-rocm

Images tagged as latest are for use with the main branch. For releases, use the major.bugfix version of Score-P as tag. I.e., use scorep/devel-gcc:8.4 to build the Score-P 8.4 release in the GCC container.

Start the container with:

[host]$ docker run [--rm] -it \
            --cap-add SYS_PTRACE \
            --cap-add SYS_ADMIN \
            -u $(id -u):$(id -g) \
            --volume $PWD:/workspace \
            scorep/devel-gcc

[container]$ ./bootstrap
[container]$ mkdir _build
[container]$ cd _build
[container]$ ../configure
[container]$ make

Citing

Please refer to the Score-P measurement infrastructure by citing the overview article The Score-P performance tools ecosystem

Version-specific DOIs of the software can be found on Score-P’s Zenodo page.

See also CITATION.cff for machine-readable citation information.

Have fun!

---

Feel free to contact us via

• support@score-p.org, or
• Matrix chat.