My research group focuses on the challenge of developing modern, high-performance software to solve problems in fluid mechanics, the aeronautics industry and other topics in the science and engineering fields. Our interests are highly disciplinary, lying at the intersection of applied mathematics, computational engineering and high performance computing.
Currently some of our goals focus around understanding fluid turbulence: how steady laminar flows transition to this chaotic state and how this impacts on real-world problems in the aeronautics industry. To achieve this I am developing efficient, robust and massively parallel high-order spectral element software that, together with modern computing technology, will form the next generation of computational flow simulation software.
Appointment as Reader 1st August 2021
I’m very pleased to say that I have been appointed as a Reader in the Department of Engineering at King’s College London! See my profile page for further information.
Two new publications in Computer Methods in Applied
Mechanics and Engineering 1st July 2021
Two students in the group have published their first papers in CMAME! Ed Laughton has a paper out on different interpolation techniques for non-conformal discontinuous Galerkin methods. Mikkel Lykkegaard has his first paper on using deep neural networks for uncertantity quantification of groundwater flow. Well done both for all of your hard work!
New funded project: ELEMENT 1st April 2020
I’m very pleased to announce that we have received funding from EPSRC to form ELEMENT: the exascale mesh network. In this project we will work towards addressing the high-priority use case of mesh generation and adaption. This project forms one of the use cases under phase 1 of the strategic priorities fund ExCALBIUR programme.
New paper in SIAM Journal on Scientific Computing 25th March 2020
Very happy to say that my collaborators Roman Amici and Mike Kirby have a new paper to appear in SISC. We investigate the performance and potential for matrix-free solvers at high-order on the hybrid meshes needed to represent complex configurations and geometries, by exploiting vectorisation that is found on modern CPU hardware. With this approach we’re able to obtain 50-70\% of the peak performance on modern hardware.
New paper in Computer Physics Communications 22nd February 2020
Together with my collaborators Jan Eichstadt and Joaquim Peiro, we have a new paper that discusses performance portable implementations of implicit high-order solvers, and the influence of algorithmic choices under programming models such as OpenMP, OpenACC and Kokkos.