(Hmmm. I'm sort of on a retrocomputing kick today...)
A few years ago, I wandered into Tom's office one day to find him typing away. I asked "whatcha doing?" and he replied "I'm writing an emulator for the PDP-1 so I can play the original Spacewar!" It tells you a lot about Tom to hear stories like this. It tells you something about me that I didn't respond with "why?" or "that's stupid" and instead just went back to my desk and wrote my own. I had never seen a PDP-1 (although the Computer History Museum has since completed restoration of theirs, where you can see Spacewar! running on real hardware). It took me a couple of days to tinker the emulator sufficient to run Spacewar! into shape, and I learned a lot about how to write emulators.
A couple of days ago, Eric sent out a link to his RetroChallenge project, an FPGA-ELF:
The ELF was a microcomputer based around the RCA-1802 processor, and was published back in 1976 in the pages of Popular Electronics. I remember reading this as a precocious 12 year old, but it would be four years until I got my first computer, an Atari 400. Still, these older machines seem like interesting things, so I thought I'd investigate what it would take to make a high quality emulator (or even a crappy one) for the 1802.
Here's the first little nugget that I found. The 1802 has sixteen 16-bit registers, labelled R0-R15. One of the more interesting features that it has is that the PC isn't a dedicated register: there is a 4 bit P register, which tells you which of the sixteen R registers serves as the PC. In each cycle, it fetches the byte pointed at by R[P], increments R[P], and then executes the instruction. I found that there is an interesting case that occurs with opcode 0x20. The high nibble of 2 indicates that it is a DEC instruction, and the low nibble specifies the register. This says "DEC R0". But in my case, R0 was serving as the PC (P was 0 as well), so this instruction essentially undoes the normal increment, restoring its previous contents, and we have a single byte infinite loop.
At least, that's what I think should happen. I don't have an actual computer, so debugging this kind of stuff is what writing emulators like this is all about.
Getting the basic instruction set working shouldn't be too hard, but there are other things like handling DMA that might be a bit trickier. Still, I'm destined to learn a lot, about a subject that is of no importance whatsoever. But then, you should expect that from me and this blog.
Addendum: Eric wrote up some nodes about his FPGA-ELF on his blog. Factoid: it runs about 100x faster than the original ELF.