Two small steps in software land: With 16-bit subtract we can bounce a ball properly. With ROM tables we can finally render the 5x8 font I designed some time ago. (In the video below I forgot to add letter spacing, so the characters are all smashed together.)

16-bit subtraction turned out to be a bit harder than I hoped for. The difficulty is that without a status register you must somehow reconstruct the carry in software. This boils down to ugly bit fiddling with nasty edge cases, and is frustrating, knowing that the minimalistic hardware has just thrown the same carry away. But it can be done and needs to be solved just once. The interpreter can now do it at the same speed as 16-bit addition: 28 cycles including interpreter overhead.

For the more time-critical sound oscillators, they run in every horizontal blank, I use a much faster 15-bit scheme (7+8), reserving 1 bit in the lower byte for carry. That works, but for the application interpreter I want proper math.

We now also have ROM tables accessible from within the interpreter. This means that it becomes easier to stuff lots of data and programs in ROM, and load any of it into RAM whenever needed and without disturbing the video loop.

Two nice steps towards a useful application environment.