PhD Position: Optimization of Numerical Simulation Software for Graphical Processing Units, Belgium

A position is available to start a PhD on numerical simulations running on Graphical Processing Units (GPUs). In a collaboration between the DyNaMat and EELAB groups at UGent, we are currently developing this high-performance simulation package that allows to simulate, on the one hand, e.g. the energy efficiency of electrical steels as used in electromagnetic devices, and, on the other hand, the read and write processes on e.g. hard disks. Such calculations are numerically very intensive. Recently we started developing a multipurpose micromagnetic code — nicknamed “MuMax” — which can run on both GPU and CPU hardware. Due to the massively parallel architecture of GPUs, the program runs now about 100x faster on this hardware. Nevertheless, there is room for numerous computer-technical optimizations and many new features as e.g. the development of communication schemes for future multi-GPU implementations.

We are currently looking for a new PhD student with a masters degree in either informatics/computer science, engineering, physics or mathematics. The research topic is the optimization of numerical methods for GPU hardware. Although applied to micromagnetic simulations, no specific knowledge of physics or magnetism is needed. A general interest for numerical methods is required though.

You will join the work on topics like numerical techniques to solve large systems of coupled differential equations, fast summation techniques, fast Fourier transforms, etc. Both the underlying algorithms as well as their implementation on CPU and GPU hardware are to be optimized. A working knowledge of C/C++ is needed for the low-level GPU implementation. Other parts of the software are now written in C++/Java/Python/Go and scripting languages. They can be extended by modules written in any high-level programming language.

Scholarship Deadline: 2011-12-31

More info: View info

Writte a comment