Published December 15, 2020 | Version 1.1

BioExcel-2 D1.1 – Specification of software engineering and QA

Description

This document describes the current state and planning of the software engineering tasks within BioExcel2 WP11 . The quality assurance (QA) and testing practices for the application software GROMACS, HADDOCK and CP2K, as well as the GROMACS linked input preparation tool PMX are described. The plans on how to continue the implementation of software engineering best practices are outlined for the projects, to ensure that the planned improvements for usability, scalability and performance do not negatively affect the correctness of the software codes and their associated web portals.

In addition, this document reports on initial benchmarking and profiling of QM/MM biomolecular simulations in CP2K, which is critically important for BioExcel to understand how to produce high impact in this area. While such simulations have tremendous potential for impact at the exascale, biomolecular QM/MM is not as mature as e.g. classical MD or docking; all work in this area carries both high risk and reward. Very limited experience exists among biomolecular simulation users and developers, including within the BioExcel consortium.

The report includes details of the work undertaken, presented to illustrate the approaches used, including development of appropriate test cases, and the approaches used to ensure that the results obtained from (short) test cases are meaningful for production simulation runs. In the short period of time available, scaling graphs and profiles have been obtained. Scaling results have demonstrated limited scaling, consistent with several existing CP2K benchmarks which are based on similar system sizes. For these kinds of problems, CP2K does not appear to currently scale well beyond about 16 nodes, notably worse than the headline scalability figures often quoted for CP2K on other problem types. Several relevant routines have been identified for further investigation, but initial results indicate that improving scaling on these problems could be challenging. To mitigate the risk associated with investing time on the code with limited return, the consortium has agreed a future checkpoint at PM8 for a final decision on whether further effort should be invested in optimization of CP2K for biomolecular QM/MM.

Files

D1.1 – Updated specification of development process-R1.pdf

Files (2.5 MB)

Name Size
md5:4c7d7a936a30774ae270553d329141e1
2.5 MB Preview Download

Additional details

Related works

Obsoletes
Project deliverable: 10.5281/zenodo.263908 (DOI)

Funding

European Commission
BioExcel-2 - BioExcel Centre of Excellence for ComputationalBiomolecular Research 823830