Numerical computations on graphics hardware

Daniel Pralle, Leibniz Universität Hannover, bachelor thesis
04/2007

In the past decades the computer processing power, measured in floating
operations per second (flops), increased significantly. This circumstance
allowed researchers to either get solutions more quickly for their numerical
models or create more complex models. Besides the CPU to achieve real-time
performance and high realism in virtual reality another component of the
computer has become a powerful processing unit, the graphic card or more
precisely: the GPU. Although the GPU of today’s graphic cards outperforms
the newest main processors easily in flops, it is not straight forward to
utilize them for general computations. The talk will introduce the graphics
processor and will show by an example what speed gains can be achieved by
the GPU.

Kontakt: Guido Böttcher