For my Master’s thesis, I wanted to see if I could find and implement a scalable global illumination algorithm. One that could run well on low-end devices while not being so limited that it could not take advantage of the extra computation and storage available on high-end devices. I settled on implementing the real-time global illumination method from Real-time global illumination by precomputed local reconstruction from sparse radiance probes .
I found that it can scale down to mobile phones and has the potential to be useful in production; my implementation runs Sponza at 30FPS on my iPhone 7, but there are definitely many areas that would need optimisation before it would be production ready.
I plan to write some blog posts soon covering the implementation details in a more bite-sized way than my thesis does. In the meantime, or if you’re interested anyway, you can download my thesis below.