Software Open Access

BioExcel exascale co-design benchmark cases

Berk Hess; Rodrigo Varga Honorato; Alexandre Bonvin; Arno Proeme; Mark Abraham

This document describes benchmark cases for co-design for the three main BioExcel application: GROMACS, HADDOCK and CP2K. We have analyzed the performance characteristics of the applications in bio-molecular type use cases and identified the one kernel or component for each code that has the most impact on performance. For GROMACS this is the non-bonded pair kernel, for HADDOCK I/O of many small files and for CP2K computation of the long-range part of the QM/MM potential. Benchmark tools have been built for these kernels/components. This document describes how these benchmarks are set up, how the tools work and discusses example output. The benchmark tools have been constructed such that a limited part of the code base is required. This is important for two reasons. One is that in co-design someone who is neither an expert of the code or the scientific domain should be able focus on the characteristics of the kernel. They should not have to understand, and certainly not benchmark, the whole code base or scientific problem. The second reason is that this makes it much easier to run code in an emulator to either inform hardware design decisions or optimize performance on hardware not yet available. Finally, possible extensions or additions of tools are discussed.

Files (118.5 MB)
Name Size
118.5 MB Download
All versions This version
Views 77
Downloads 55
Data volume 592.4 MB592.4 MB
Unique views 55
Unique downloads 33


Cite as