Ξ

im somewhat biased towards cisc in this case since id think theyre a little easier to program. you can worry a little less about addressing modes and such and still get a good idea of what the computer is actually doing with the data you give it, so i think it lends itself well to education

this has the same energy as my "put a Cortex-M0+ with exposed bus on a DIP" idea, with the only difference being that Cortex-M0+ already has a Von Neumann bus so no hacking is required

Eeeehhhhhhhh, probably more of a lukewarm take. Back in The Day, there was a holy war between the "sixers" and the "eighters"; that is, the Motorola/MOS way versus the Intel/Zilog way. If you look up old magazines, newsletters, and the like, you'll find people on both sides claiming the other party's ISAs are crap. And yeah, the 68HC11 was a better take on the 6800, though HARVARD!!!!!!!!!

Unfortunately, MIPS refuses to let anyone else have the materials. I didn't mention it, but they actually sued Loongson over the continued production of MIPS-compatible CPUs, but thankfully lost the case. However, Loongson's take is a hybrid of MIPS and RISC-V and is also very much overkill for the goals I have outlined.

Also, very nice to see that you actually found mirrors of the files! Shame that the only known mirrors are literally paywalled. And they actually prevented the Wayback Machine from archiving the files, the dicks :(

Now the problem is the availability of that specific FPGA. The manufacturer is more likely to kill it off without a replacement, bringing us back to the same problem. Seriously, I was looking at replicating a project someone did that used an FPGA to emulate the mapper chips on Game Boy carts to create a flash/repro cart, but the FPGA that was used was discontinued by the manufacturer and they had no drop-in replacement, which meant that if I were to revive this project I would need to find a suitable replacement FPGA, rewrite all the verilog to work with the new FPGA, and then have to make a cart PCB from scratch because the whole gimmick was that the FPGA was able to fit in the same footprint as the actual mapper chip on the real PCB, but everything else on the market wouldn't be able to properly fit.

But a bigger, more real problem is actually cost. FPGAs aren't cheap! I've seen some FPGAs on modules that fit in DIP sockets, and they cost far more than the most expensive chips mentioned in my post. However, such a module would be a good solution to the "make it human solderable" issue, in that even a BGA part could be put onto a daughterboard that the end user puts into a DIP socket.

oh so it's even more war crimes than I gave it credit for

From what I was able to find, all RISC-V parts that aren't Harvard-style microcontrollers are BGA parts, which fails the "human solderable" requirement. Furthermore, they're all very complicated SoC designs, and the vibe I've been getting is that you're not supposed to write any bare metal code, so for the intended use case it's not really worth it. I really need to write a followup to this chost.