Paddy

Paddy Ó Conbhuí Ph.D

Computational Scientist
MSc High Performance Computing, University of Edinburgh
Ph.D GeoSciences, University of Edinburgh
+353 1 524 1608 (ext 34)
HPC Hub Dublin

Pádraig "Paddy" Ó Conbhuí joined ICHEC’s Oil & Gas team in November 2017 to work on the LERO project: ExSeisDat. The ExSeisDat (Extreme-Scale Seismic Data) project is a free and open-source, high performance, parallel I/O library for seismic data files (e.g. SEG-Y), with a geophysicist-friendly interface, along with a dynamically optimized seismic workflow library. Pádraig and the ExSeisDat team work in close collaboration with Tullow Oil to ensure the project is relevant and useful to the Oil & Gas industry, and with DataDirect Networks (DDN) to ensure it's optimized for next-generation I/O hardware and technology.

Pádraig has extensive experience in computational physics, and in developing and optimising physical models that translate into high performance numerical algorithms. He is adept in a wide variety languages, particularly C, C++, Fortran, and Python, and in building software bindings between these languages. He has consulted and collaborated on a number of projects with colleagues from Imperial College London, the University of Cambridge, ETH Zurich, and the Geological Survey of Norway, on topics from Paleomagnetism and Rock Magnetism, to Biomagnetism for medical applications. His work included developing and extending software to enable research in Rock and Mineral Magnetism, by building high performance applications with user friendly interfaces.

During his B.A. Mod in Theoretical Physics in Trinity College, Dublin, Pádraig completed an undergraduate dissertation on developing high performance multiply-trace kernels for complex-valued matrices in CUDA C using NVidia Tesla GPUs for Lattice QCD applications. By fusing a number of multiplications together, he beat the performance of cuBLAS by around 20%, and by fusing a number of traces together, rearranging them into a matrix multiplication, he beat a naïve pure-trace implementation by around 20x.

He then went on to complete MSc is in High Performance Computing with EPCC in the University of Edinburgh, completing a dissertation on identifying and alleviating bottlenecks in parallel Molecular Dynamics algorithms using MPI and OpenMP. By combining a Replicated Data and Systolic Loop approach, he developed a new list comparison algorithm (intended for computing a Lennard-Jones potential with no cut-off) that had near perfect speedup, scaling as O(N^2/P), with N the number of particles, and P the number of processors, up to P = N, and even continued scaling with reduced speedup beyond that. The result was a list comparison that could be completed in O(N) (or less!), given enough processors.

In 2017, Pádraig received his Ph.D. from the School of Geosciences in the University of Edinburgh. The topic of his thesis was on including magnetostriction into micromagnetic models of nanoscale ferromagnets: modelling how stress and strain effect how magnetic minerals might store magnetic signals, particularly over geological timescales. However, he also worked on a wide number of projects in the fields of Numerical Rock Magnetism and Paleomagnetism, and on software to lower the barrier to entry for researchers using micromagnetic modelling software. The project involved developing expertise and making significant contributions in a wide range of topics including Magnetism, Ferromagnetism, Elasticity, Plasticity / Dislocations, Electron Holography, Finite Element Methods, Non-linear Optimization, Computer Geometry, and Tetrahedral Volume and Triangular Surface Mesh Generation.

He received a number of awards for presentations delivered at conferences throughout his studentship.

Publications

Ó Conbhuí, P., Williams, W., Fabian, K., Ridley, P., Nagy, L., & Muxworthy, A. (2017). MERRILL: Micromagnetic Earth Related Robust Interpreted Language Laboratory. Geochemistry, Geophysics, Geosystems (Under Review)

Nagy, L., Williams, W., Muxworthy, A. R., Fabian, K., Almeida, T. P., Ó Conbhuí, P., & Shcherbakov, V. P. (2017). Stability of equidimensional pseudo–single-domain magnetite over billion-year timescales. PNAS 114(39), 10356-10360.

Almeida, T. P., Muxworthy, A. R., Kovács, A., Williams, W., Nagy, L., Ó Conbhuí, P., Frandsen, C., Supakulopas, R., & Dunin-Borkowski, R. E. (2016). Direct observation of the thermal demagnetization of magnetic vortex structures in nonideal magnetite recorders. Geophysical Research Letters, 43(16), 8426-8434.

Einsle, J. F., Harrison, R. J., Kasama, T., Ó Conbhuí, P., Fabian, K., Williams, W., Woodland, L., Fu, R.R., Weiss, B.P., & Midgley, P. A. (2016). Multi-scale three-dimensional characterization of iron particles in dusty olivine: Implications for paleomagnetism of chondritic meteorites. American Mineralogist, 101(9), 2070-2084.

Einsle, J. F., Fu, R. R., Weiss, B. P., Kasama, T., Fabian, K., Ó Conbhuí, P., Williams, W., Midgley, P., & Harrison, R. J. (2015). Focused ion beam nanotomography of chondritic meteorites: Closing the mesoscale length gap in paleomagnetic studies. Microscopy and Microanalysis, 21(S3), 2261-2262.