GPU accelerated ray tracing

Abstract

Creating a virtual simulation of the world around us that looks and behaves properly has always been the ultimate goal of old and modern graphic technologies that have long struggled with how light works, an effectively infinite number of photons zip around, reflecting off surfaces and passing through objects, all based on the molecular properties of each object. Simulating infinity with a finite resource like a computer's computational power is a recipe for disaster. Clever approximations were needed and new techniques were introduced. This work describes the implementation of a GPU accelerated path tracer, it was organized in the form of a case study, evaluating different techniques and their impact on the final product. This report discusses complications typically encountered when implementing a path tracer and how these can be bypassed. Considerations were taken concerning quality, rendering speed, extensibility, configurability, and compatibility when features were chosen for implementation. Light interaction with surfaces, diffuse reflections, physically accurate shadows, reflections and refractions, geometric properties and much other real world effects are introduced. To compare the effectiveness of the implementation held under the GPU platform, we have opted for implementing an equivalent algorithm and test it under a CPU platform. The final scenes are presented at the end of this thesis.