PreDem -- Democratization of the Coupling Library preCICE

点击翻译按钮获取中文摘要

An increasingly important class of applications for high-performance computing are multi-physics and multi-code scenarios. They are characterised by the underlying model that typically consists of different types of systems of equations that are physically coupled. Typical examples are interactions between fluid flow and elastic deforming structures in aerospace engineering or medical applications or the thermal coupling between flow and rigid structures in cooling or heating systems. These complex models are established in order to ensure a very high accuracy which implies that we have to ensure a comparably high accuracy in the respective numerical simulation. This requires running the simulation on high performance computing platforms. To target the resulting challenges, we have developed a library called preCICE, which enables the efficient, robust, and parallel coupling of separate (even black-box) single-physics solvers. This approach leads to very short software development times and very fast simulations. preCICE is in the stage of a prototype. It has been coupled to various community codes such as OpenFOAM, Calculix, or ANSYS Fluent, but also with in-house codes of several academic institutions. preCICE is open-source under the LGPL3 license. During the last years, the requests for using preCICE in scientific projects increased significantly -- a dozen institutes during the last two years. Minimally invasive coupling is an ever-growing need for many applications. At the current state, however, the use of preCICE without support from the developer team is hardly feasible. The objective of the project is to derive best-practices on how to make a research software usable for non-experts. This is an unsolved issue not only for preCICE but for academic software in general that typically provides highly sophisticated state-of-the-art methods, which, however, can not be fully exploited by the community due to a lack of usability. With the main reason being limited time for usability improvements in the developer teams, our objective is to provide infrastructure, methods, guidelines and experiences that allow other developer teams to achieve a good usability of their code in a faster and easier way. We realize this objective by an exemplary study of the coupling library preCICE. The derived best practices can then be applied to other research software. Democratization of a research software, i.e., making it usable for non-experts, encompasses several aspects, which we group in four categories tackled in the projects work packages: Infrastructure (the research software should be easy to get and easy to set up), sustainability (the long-term sustainability of the research software has to be guaranteed), usability (the research software should be easy to use), outreach (the research software should be known and trusted).