The ever increasing complexity of 3D models generated by 3D scanners and through modeling requires ever faster graphics cards to render them at interactive rates. Unfortunately, the rate at which the performance of graphics cards increases is outpaced by the advances in scanning technology, making todays 3D models too large to render interactively. Because of the sheer amount of geometry to render, datasets such as the USGS Earth model are also too large to fit into memory, making the rendering even more challenging. The most common solution to these problems is the use of Level of Detail systems that simplify the original model, removing some of its complexity while maintaining as much of the original detail as possible.
We present a new multiresolution hierarchy and associated algorithms that provide adaptive granularity. This multi-grained hierarchy allows independent control of the number of hierarchy nodes processed on the CPU and the number of triangles to be rendered on the GPU. We employ a seamless texture atlas style of geometry image as a GPU-friendly data organization, enabling efficient rendering and GPU-based stitching of patch borders. We demonstrate our approach on both large triangle meshes and terrains with up to billions of vertices.
The proposed method also takes advantage of the parallelism available in the latest generation of computers. The availability of multi-core, multi-GPU machines allows for significant performance improvements when taken advantage of. In the proposed method adapt tiles, render tiles, and machine tiles are associated with CPUs, GPUs, and PCs, respectively, to efficiently parallelize the workload with good resource utilization. Adaptive tile sizes provide load balancing while our level of detail system allows total and independent management of the load on CPUs and GPUs. We demonstrate our approach on parallel configurations consisting of both single PCs and a cluster of PCs.
Speaker Biography
Krzysztof Niski received his M.S. from Johns Hopkins University and his B.S. from Shepherd University. His research interests lie in the computer graphics field, focusing on large-scale rendering systems, programmable graphics hardware and scientific data visualization.