Efficient GPU Rendering of Subdivision Surfaces using Adaptive Quadtrees

摘要

We present a novel method for real-time rendering of subdivisionrnsurfaces whose goal is to make subdivision faces as easy to renderrnas triangles, points, or lines. Our approach uses standard GPUrntessellation hardware and processes each face of a base mesh independently,rnthus allowing an entire model to be rendered in a singlernpass. The key idea of our method is to subdivide the u; v domain ofrneach face ahead of time, generating a quadtree structure, and thenrnsubmit one tessellated primitive per input face. By traversing thernquadtree for each post-tessellation vertex, we are able to accuratelyrnand efficiently evaluate the limit surface. Our method yields a morernuniform tessellation of the surface, and faster rendering, as fewerrnprimitives are submitted. We evaluate our method on a variety ofrnassets, and realize performance that can be three times faster thanrnstate-of-the-art approaches. In addition, our streaming formulationrnmakes it easier to integrate subdivision surfaces into applicationsrnand shader code written for polygonal models. We illustrate integrationrnof our technique into a full-featured video game engine.