Rendering Subdivision Surfaces Efficiently on the GPU

Subdivision surfaces have become popular recently in computer graphics, because subdivision algorithms can easily provide smoother versions of an initially coarse base mesh of arbitrary topology. In this paper we present a GPU-based implementation of the Catmull-Clark subdivision surface scheme. This solution computes subdivisions in the geometry shader of Direct3D 10 compatible GPUs. This is not only much faster than doing it on the CPU, but it makes also possible for the CPU to work on other important tasks.