Abstract

Scientific software practices and development has shifted dramatically in recent years, driven by a need to improve transparency of how our scientific results are generated from software tools in order to ensure reproducibility. For newer codes, or those written from scratch, modern software development tools (which incorporate e.g. issue tracking, code review, CI/CD, container registries and other devops tools) can help to promote both a collaborative software practice and improve overall software quality. However, more established codes—particularly those which are frequently used in the high-performance computing environment—face a number of challenges in using these tools effectively. In this talk, I will introduce the Nektar+ spectral/hp element framework: a C+ codebase developed across the last decade, which aims to provide a high-performance scalable platform for the development of physical solvers based on the high-order finite element method. I will discuss the strategies and attempts we have made to underpin sustainable development of the code and improve general code quality, as well as to highlight our future plans in this area.