That's not the whole GPU, just the timing part.

The rest appears to be in framebuffer.S, still not a lot of code though.

Makes me wonder how easy it would be to implement something like this with WASM and either WebGL or WebGPU.

Depending on how 'correct' the video emulation is (e.g. per pixel, per scanline or per frame), it won't make much sense to offload the video emulation to the GPU because it must run in lockstep with the rest of the system emulation. Moving the entire emulation to the GPU also makes little sense because such old-school system emulations are hard to parallelize (because of the above mentioned lock-step problem - for a cycle-correct emulator you basically need to emulate one clock cycle for all 'subsystems' after another). You could probably emulate a ton of fairly slow gameboy emulators in parallel though :)

This is all of course a very rudimentary GB emulator, not accurate at all.

Hey hey hey. Whatsamatta with my implementation? It should be fairly cycle-accurate...