high performance computing on graphics processing units: hgpu.org

Molecular dynamics simulations are today a widelyused tool in many research fields. Such simulations produce large time-dependent data sets, whichneed to be interactively visualised allowing efficient exploration. On the other hand, commonlyused point-based rendering of the individual particles usually fails to emphasise global contiguousstructures like particle clusters. To solve this issue,we want to visualise these data sets using metaballs.Free particles form individual spheres while clustered particles result in larger closed shapes. Usingimage-space hardware-accelerated techniques provides interactive frame rates and high visual quality. We present two approaches evaluating the metaballs shapes on the graphics hardware. The first approach uses a vicinity data structure stored in graphics memory to evaluate the metaballs shape in a single rendering pass. The second approach uses multiple rendering passes to approximate the metaballsshape on a per pixel basis.