Empirical Heterogeneous System Modeling for Efficient Real-Time Rendering

Real-Time Ray Tracing using computation hardware such as GPUs and CPUs requires new and innovative algorithms that take advantage of modern hardware architectures and their unique capability. In particular the combination of these devices creates heterogeneous architectures that require new algorithms and data structures for efficient ray tracing. This dissertation will investigate these new ray tracing algorithms and how they can effectively utilize heterogeneous. To improve the utilization, this thesis will also investigate the use of probabilistic models of heterogeneous architecture performance.