Welcome!

I am a computational physicist with research interests in massively parallel and distributed simulation, GPU computing, and general relativistic computational fluid dynamics. I hold a PhD from the Technical University of Munich and the Max Planck Institute for Astrophysics, and I have spent some time at the Center for Computation and Technology at Louisiana State University and the Albert Einstein Institute in Potsdam before coming to Tübingen.

At present, I am mostly focused on employing graphics processing units (GPUs) to perform high-throughput simulations of magnetars, which are a special kind of neutron star with ultra-strong magnetic fields. These require three-dimensional simulations of the equations of magnetized fluid flow, in a variant which is consistent with Einstein's equations of general relativity.

Simulations of this kind are computationally expensive, and therefore require massively parallel and distributed computational techniques. Traditionally, these are performed on supercomputers, which are nowadays often large interconnected clusters of PC nodes.

Graphics processing units (GPUs) are a massively data-parallel architecture designed to produce real-time 3D graphics. Some of the operations performed in the traditional graphics pipeline like affine vertex transform, fragment processing and so on, have over the years evolved into generally programmable stages which are essentially data-parallel (SIMD) operations. Since this makes the GPU also a throughput-oriented processor, the same programmability can be used to perform more general computations, whether in graphics (for path tracing and photorealistic rendering) or in parallel simulation, like in computational fluid dynamics.

I have authored the Horizon code, an implementation of general relativistic magnetohydrodynamics on GPUs. With this code, we can investigate the oscillations and magnetic field dynamics of magnetars at unprecedented speeds, essentially establishing a neutron star model which can cover a second of physical evolution time on a single workstation, or allowing to use clusters of GPUs for very detailed investigations.