Compressed Grids for GPU Ray Tracing of Large Models

Ray tracing on GPUs is an area of ongoing research. GPUs are well suited for this parallel rendering algorithm. GPU computing devices typically have characteristics which make them quite different from CPUs: increased data parallelism, increased memory bandwidth, smaller caches, lower memory capacity. Presently it is difficult to visualize large scenes with tens of millions of triangles in these memory constrained platforms. In this paper we present a compressed grid data structure, capable of state of the art rendering performance, using up to 6× less memory than conventional grid storage schemes. The compressed grid is built and traversed on the GPU.