Advancements in mathematical models, numerical algorithms and computing capabilities have brought many complex non-linear problems in science and engineering closer to understanding through simulations. The increased resolution and modeling fidelity comes with an increase in the complexity of developing and managing advanced simulations software. A key factor in effective utilization of HPC resources for simulation is the availability of sophisticated, scalable, and robust community codes. Because of their complexity, the development of such codes demands interdisciplinary, long-term collaborative efforts by teams of experts.
This award will fund a workshop to develop ideas on how to widen the user base on the high-end HPC platforms through the use of community codes and an integrated computing approach to science. Such a workshop on community codes is timely in view of the extraordinary challenges faced by computational scientists in planning and preparing for the next generation computing platforms. The workshop will address topics including; understanding what it takes to develop successful community codes; practical and effective software engineering for scientific codes; interdisciplinary interactions in integrated science computing; and data management and analysis for scientific simulations.
A. Dubey, E. Balaras and D.Q. Lamb The workshop brought attention to some of the most important issues in scientific software development that are critical for the codes and the research communities served by these codes. Having a robust and well supported community code provides a definite advantage to the community in terms of the start-up cost of research projects. For example, if a code provides infrastructure for IO, analytics and fundamental solvers, the end users do not have to spend any of their research resources developing them. Instead, they can concentrate on implementing custom capabilities applicable to their specific projects, and therefore obtain results much more quickly. Similarly, the projects that benefit from AMR don't have to develop their own interface to the AMR packages. Several teams would not be in a position to use high performance computing if it wasn't for the availability of the community codes that are designed for them. However, despite all these obvious benefits there continues to be a scarcity of successful community codes and a lack of willingness by the research communities to invest in them. One of the objectives of the workshop was to identify issues that keep the community codes from reaching their potential and to propose and evaluate solutions that would help mitigate this problem. The consensus among the participants in the workshop was that community codes have a unique value, and it is timely to have a forum for ongoing discussion about their requirements and concerns. It is equally important to evangelize the community codes' usefulness to research communities that use computing for obtaining scientific results. Community codes reduce the barrier to entry for new researchers in computing based research, provide more reliable means of achieving results and foster communications within communities and among communities. All of these factors benefit and enhance science and engineering research. Some of the barriers to community codes reaching their potential are: (1) tension between the science needs and long-term health of the scientific codes; (2) the sociological challenges posed by the interdisciplinary nature of community code development process; and (3) a lack of career path in academia for the broadly qualified software engineers and developers that are necessary for such coding projects. It was agreed that this workshop was a good start, but that similar workshops should be held regularly and an attempt should be made to broaden their reach. It is equally important to expose students to the benefits of community codes and good software practices.
