A classification scheme for rendering algorithms on parallel computers

In recent years, the size of data sets in applications such as scientific visualization and virtual reality, have been continously growing. In High Performance Computing (HPC), parallel computers are used to reduce the total rendering time of these data sets. Shared memory architecture computers are nowadays widely used, but distributed memory computers promise better performance due to their scalability. At the Fraunhofer Institute for Computer Graphics, parallel computer graphics research is carried out using both these architectures. In this paper we discuss the control flow of rendering algorithms in general purpose multicomputers, which are connected to specialised rasterizer hardware. In the first section, we present the MANNA / VISA system. In the next section we have a general look at the rendering process on multicomputers, and finally we construct a classification scheme and evaluate it on the MANNA / VISA system.