I found an interesting blog over at http://benryves.com. It belongs to a guy that seems to share my passion for low end systems. He's working on a 2.5D engine for the TI-83 and has a bunch of different hardware and software projects.
I've finally put all the pieces together, and it turned out pretty good imo. The video shows a rotating torus with a rotating directional light. I'd love to do point lights, but the inverse square root needed for normalization is extremely expensive in terms of CPU time needed.
Found an interesting article about fast fixed point division and multiplication on ARM processors over at Henry Thomas consulting. Originally I claimed that it didn't work for the GBA, but that was a lie. Some minor tweaks and now it runs like a charm. Fixed point division used to be a major bottleneck in my attempt at 3D rendering on the GBA, and now it's not a problem at all.