University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Colloquium of the Department and the PhD Program

title

"Non-graphical" applications of programmmable graphics hardware

speaker

Prof. Dr. Andreas Kolb, University Siegen
Siegen, Germany

date & place

Tuesday, 14.06.2005, 14:15 h
Room C421

abstract

Traditionally, the graphics processing unit (GPU) is used in the process of real-time image synthesis. In the last years programmable GPUs were introduced and this programmability is increasing especially for commodity PC graphics cards. Thus, these chips are capable of performing more than the specific graphics computations for which they were designed. They are now capable coprocessors, and their high speed makes them useful for a variety of applications, like image processing and simulation tasks. At first, this talk will give a brief overview on programmable GPUs. After a short introduction to the graphics pipeline, various aspects of programmable GPUs are discussed. In order to implement any algorithm on a GPU, one has to evaluate the inherent structure of the algorithm. Since a programmable GPU can be seen as a specific type of parallel processor, problems that are not suitable for parallelization in general can rarely be implemented on GPUs successfully. Some examples of applications in the image processing context will be given. The second part of this talk will give some examples of GPU-based simulation algorithms. One example is the implementation of massive particle systems on the GPU. The simulation of particle systems can easily be parallelized if the particles are not coupled. These kind of particle systems can simulate up to 1 million particles in real-time, i.e. at about 20 frames per second. Coupled particles can be introduced, if a fixed, sparse coupling pattern is present, e.g. for cloth simulation. Additionally, an approach for dynamic coupling of particles is presented.