Micropolygon Rendering on the GPU (DAEV)

Recent advances in the programmability of graphics processors(GPUs) have made it feasible to replace the hardware-implemented graphics pipeline with one implemented completely in software. This allows us to use methods previously only used in the field of software rendering to improve the visual fidelity of real-time graphics. One such method is the Reyes rendering algorithm which allows the rendering of curved and displaced surfaces by splitting them into sub-pixel sized polygons(micropolygons) before drawing.

This thesis presents an OpenCL implementation of Reyes running on the GPU. Our major contribution is the implementation of parallel surface subdivision with a bounded memory usage.