Code generation for unstructured mesh computations
With Paul Kelly and David Ham, I work on code generation for numerical simulations on unstructured meshes. Our current work has a particular focus on the finite element method. Building on work in the FEniCS project, we are interested in how to generate efficient code for different architectures (such as multicore CPUs, or GPUs) from a single high-level, architecture-agnostic source.
With Wyn Williams, and our PhD students Les Nagy and Patrick Conway, I work on the development of models of micromagnetism using the finite element method. Model development is carried out using the FEniCS framework. We use our models to study the paleomagnetic record, the historical record of the Earth's magnetic field. Having an accurate picture of the historical field is necessary to understand the behaviour of fluids deep within the Earth. Rock records are unable to provide all of the necessary information, and so we use simulation models to augment experimental data.
Experimental work on this, and studies of magnetotactic bacteria, is carried out in collaboration with Adrian Muxworthy.
Parallel statistical analysis
I am one of the lead developers of the SPRINT framework, an R package developed to parallelise core statistical analyses necessary in workflows for the analysis of genomic and post-genomic data. This work is a collaboration between EPCC and the Division of Pathway Medicine at the University of Edinburgh.
Eike Mueller, University of Bath, Multigrid solvers for numerical weather prediction., 2013
Professor Wyn Williams, The University of Edinburgh, Micromagnetic models, 2012
Research Student Supervision
Nagy,L, Parallelisation of a micromagnetic model