A CUDA-based implementation of OpenGL-compatible rasterization library prototype

Traditional 3D graphics pipelines have been mostly implemented as fixed-function special-purpose hardware chips. Recently, introducing programmable graphics pipelines, there was dramatic changes in their implementations. Additionally, introducing general purpose GPUs (graphics processing units), we got parallel computing pipelines such as CUDA and OpenCL, based on the parallel processing facilities of those GPUs. In this paper, we aimed to implement a full-software 3D graphics pipeline on those parallel computing pipelines, to finally show that the new full-software implementation can substitute the traditional graphics pipelines. In the previous work [1], they implemented the 3D rasterization engine on the CUDA architecture, for some special applications. Based on that CUDA-based 3D rasterization engine, we added the OpenGL 1.2-specific API (application program interface) functions, to build up a source-level compatible 3D graphics library for existing OpenGL graphics application programs. In near future, we will implement a wholly new OpenGL graphics library features on the CUDA pipeline. Those efforts are focused on the demonstration of feasibility for the unification of 3D graphics pipelines and parallel computing pipelines. We expect that this unification will give us remarkable advances in the future graphics hardware design.