Distributed Ray Tracer on GPU

Ray tracing is a method for producing photorealistic 3D computer generated imagery by modeling the interaction of light rays with a scene. Because each primary ray is independent of other primary rays being modeled, ray tracing offers massive degrees of parallelism that is suitable to parallel architectures like GPUs, multicore CPUs, and distributed computing environments. Our goal is to implement a ray tracer running on multicore CPUs and NVIDIA GeForce 8800 GPUs in multiple systems in a cluster that exchange scene data and rendered images to distribute the computational load. A dynamic scheduling algorithm assigns work units to each node to provide high sustained performance that scales with additional processors and GPUs.