HPC Benchmark Game Source Code and Results (PECS 2025)

This archive contains the complete source code, measurement scripts,
measured data, postprocessing scripts, and figures for the paper:

Max Lübke and Dorian Stoll and Bettina Schnor and Stefan Petri:
HPC Benchmark Game: Comparing Programming Languages Regarding
Energy-Efficiency for Applications from the HPC Field.
Accepted for PECS2025 - Workshop on Performance and Energy Efficiency
in Concurrent and Distributed Systems, Co-located with Euro-Par 2025,
Dresden, Germany, August 26, 2025. https://pecs-workshop.github.io/2025

A preliminary PDF can be found here.

The HPC Benchmark Game is a collection of 5 benchmark programs, each
implemented in up to 5 programming languages. The intention was to
compare the impact of the programming language on runtime, energy and
memory usage. Unlike its namesake, the Computer Language Benchmark
Game, this focuses on parallelized, HPC-adjacent applications.

The source code for the benchmark applications is maintained on
https://gitup.uni-potsdam.de/bsvs/public/hpc-benchmark-game

For the energy measurement, it uses the EMA library as a
submodule. That is maintained on

https://github.com/PERFACCT/EMA

This archive includes a copy of the EMA code in its entirety.

One way to build the provided programs is by using the included
Dockerfile. See README.md for further instructions.

On the PIK HPC2024 cluster, which was used for the measurements shown
in the paper, using Docker and meson was not feasible, thus the build
and measurements were done on bare metal with shell scripts.  See
README-HPC2024.txt for further details.

The directory PECS2025-Figures contains the measured data,
postprocessing shell scripts, and figures that were used for the paper
(and some more that did not make it into the paper).