As of now, I have a basic working raytracer implemented in opengl and c++. However, I am facing difficulties in porting it to CUDA because it is based on the cmake/make build system with which I have had no prior experience.
I still think it is possible to remain with this implementation except it will take longer than expected. For now, I plan to try and port it over to CUDA. At the same time, I will work on the CPU implementation except using BVH for ray-object intersection tests. If there are too many unexpected difficulties in porting the existing implementation to CUDA, I will look for another existing implementation and work on that instead.
For the parallelism competition, I hope to be able to show a live demonstration of the raytracer program. If that proves to be impossible, I will instead show some graphs comparing the speed of the different implementaions I mentioned above.
As a summary, these are the issues I am facing: