Extending the graphic pipeline with new GPU-accelerated primitives
Rodrigo Toledo, Bruno Lévy · 2005
Graphics hardware is optimized for rasterizing and solving the vis-ibility of points, lines and polygons. In this paper, we propose a GPU implementation of new graphics primitives (e.g. spheres, cylinders, ellipsoids), which are compatible with the standard pipeline. Our technique consists of two steps, both executed on the GPU: (1) candidate pixels are pre-selected by rasterizing a standard prim-itive whose projection encloses the projection of our primitive; (2) for each candidate pixel, a fragment program computes the in-tersection and lighting, in a ray-tracing way, based on an analytic representation of our primitives. In addition, it replaces the Z-buffer value of the fragment according to the equation of our primitives. Our primitives are highly efficient. They are perfectly smooth under any arbitrary zoom. They consume much lower bandwidth than tesselated primitives (e.g. drawing a sphere only requires to send a GL POINT), and they have correct visibility/intersections with other primitives (including the standard ones).