There is typically a disconnect between the tools used by domain scientists and research engineers exploring scientific problems through simulation, and the leading edge of high-performance solver technology. Bridging this gap is important to accelerate the exploitation of new computational and mathematical tools for enabling new scientific investigations and new engineering design. In the context of the FEniCS Project (http://fenicsproject.org/), we have been successful in making the creation of new, performant finite element-based solvers compact and expressive, and accessible and straightforward for domain scientists. A natural step is to place a wide range of high-performance solvers at the fingertips of researchers, allowing them explore the parameter space of sophisticated solvers for advanced applications with ease, with the objective of finding scalable, parallel solvers for new and changing models. I will present a number of examples of how this can be made easy, from elliptic problems with over 10 billion degrees of freedom, to multi-field equations, to some of the largest practical engineering simulations performed.

 

Garth Wells is the Hibbitt Reader in Solid Mechanics and University of Cambridge. His research interests span scientific computing, numerical analysis, mathematical software and high-performance computing. He is a leader of the FEniCS Project (http://fenicsproject.org), an international collaborative project on novel, open-source software for solving differential equations. Before joining University of Cambridge he held a faculty post at Delft University of Technology (Netherlands), and has held research/fellowship positions at Stanford University, University of Texas at Austin and University of Bologna.