Runtime performance measurements of interactive visualisation algorithms

<p>Runtime performance measurements for GPU-based direct volume rendering and GPU-based raycasting of spherical particles on ten different discrete graphics processing units from AMD and NVIDIA. The data set at hand systematically evaluates typical factors influencing performance of interactive visualisations like view-port size, data set size, camera position, etc. The volume raycaster is implemented in OpenCL whereas the particle raycaster is written in Direct3D11.</p>

<p>The data set comprises the individual measurements as comma-separated values. The results for the volume raycasting and the particle rendering are stored in separate files compressed in zip archives (volume.zip and spheres.zip). Each line represents a measurement and all dependent and independent factors are given for that measurement. The machine-readable documentation in the additional JSON files (volume-description.json and spheres-description.json) describes these factors and specifies the data type in terms of .NET CLR types. The content of the XDF files volume.xdf and spheres.xdf is identical to the comma-separated values in volume.zip and spheres.zip. The files are provided for faster loading into R. The image files (grouped into the "renderings" folder in tree view) provide sample renderings of the data sets used for the performance measurements. These are intended for illustration purposes.</p>