Analysis and Implementation of Local Subdivision Algorithms in the GPU

Being able to refine a mesh without comprimising the memory bandwidth is a powerful tool that might be explored with the new Tessellator part of the graphics pipeline. In this paper we provide a detailed implementation of two local subdivision algorithms - PN-Triangles and Phong Tessellation - using the new Tessellator pipeline. Moreover, a explicit quality and performance comparison between both algorithms is made. Our results showed that Phong Tessellation has a considerable performance gain in comparison with PN-Triangles when implemented in the current Tessellation hardware. However, the visual quality of PN-Triangles algorithm is visually smoother.