Shadow Volumes on Programmable Graphics Hardware

One of the best choices for fast, high quality shadows is the shadow volume algorithm. However, for real time applications the extraction of silhouette edges can significantly burden the CPU, especially with highly tessellated input geometry or when complex geometry shaders are applied. In this paper we show how this last, expensive part of the shadow volume method can be implemented on programmable graphics hardware. This way, the originally hybrid shadow volumes algorithm can now be reformulated as a purely hardware-accelerated approach. The benefits of this implementation is not only the increase in speed. Firstly, all computations now run on the same hardware resulting in consistent precision within all steps of the algorithm. Secondly, programmable vertex transformations are no longer problematic when applied to shadow casting objects.