Compression-domain parallel rendering

The 3D triangle mesh is the dominant representation used in the parallel rendering of 3D geometric models. However, the explosive growth in the complexity of the mesh-based 3D models overwhelms the communication bandwidth of existing parallel rendering systems. An effective solution to this problem is to use a compressed mesh representation. In recent years, researchers have shown a great deal of interest in developing highly efficient mesh compression algorithms. However, using a compressed mesh in a parallel rendering architecture to achieve the highest possible end-to-end performance is a largely unexplored area. We have previously (1998, 2000) developed an efficient mesh compression/decompression algorithm, called "breadth-first traversal" (BFT). In this paper, we design and implement a parallel rendering architecture that can use a BFT mesh representation. The enabling technology is a novel algorithm that can perform a compression-domain subdivision of the BFT mesh for bandwidth-efficient distribution of submeshes to parallel processors. Parallel rendering using a BFT mesh reduces the communication requirement to about one third of that of uncompressed representation.