The “fun” part was getting the scanline rendering for the text which is very tight on time - I hit an optimiser bug on whatever version of gcc I was using at the time where it unrolled the loop too much and actually made it slower, so there’s a bunch of inline assembler to ensure it’s fast enough!