The topic of our project is the division of the Ray Tracing task upon 2 GPU's. Today, the Ray Tracing task is very time-consuming, and is therefore not applicable in computer games and other real-time applications. Our projects aims at making this process quicker, by performing the ray tracing task in parallel on 2 different processing units.
In this project we examined multiple methods of dividing the scene (between the two processors) so that rendering time is minimal. The difficulty was finding two sections of the image, whose rendering time is similar. This task was difficult because it's hard to estimate the rendering time of a section of the scene before the actual rending occurs.
We tested 3 different methods of division of the screen. Our conclusions were that no method can perform well on any input scene, however a combination of two of the methods (deciding which one to apply on each scene) works well for most scenes.