4 ms·
The 68k has an excellent instruction set but imho it suffers from two things: a) It's big-endian. This is a bit of a religious thing, but in the end little-endi
by cmrdporcupine 5y ago
The 68k has an excellent instruction set but imho it suffers from two things: a) It's big-endian. This is a bit of a religious thing, but in the end little-endian won out, and probably for good reasons. It's arguably easier to read a big-endian hexdump but harder for the processor itself to manage and harder for some bit twiddling. b) Separate "data" and "address" register sets. Awkward to program, and makes the instruction set more complicated.
Looking back, the similar but not nearly as successful NS320xx architecture was better. Atari apparently even prototyped it for use in what became the (68k based) Atari ST, but the first chips produced in that series were too buggy.
Both the 68k series and the NS32k series were heavily 'inspired' by the VAX.
One thing that makes the 68k cool is that it's about as "retro" of an architecture as you can get that still supports a modern C/C++ toolchain. GCC still (mostly) supports it.
- kabdib 5y agoI worked on the ST and helped evaluate the NS32032/32016. We started counting clock cycles for various operations and got some bad vibes, and then we started finding out about chip bugs (definitely not pretty, and National was coy about them and just said stuff like "we're working on it"), we so decided to use the 68K. The NS parts were really a due-diligence affair, and we looked at them just to make sure we weren't missing out on something really great (we weren't). Definitely a prettier architecture than the 68K, with better toolchains and development support at the time. Didn't make up for the bad, though.
- cmrdporcupine 5y agoThanks, I always appreciate it when you chime in. Looking back, did Atari Corp actually build any hardware (prototype, wirewrap, etc.) with the 320xx? Or was it just evaluation from the POV of reading datasheets and the like? I've heard conflicting stories around this. That and the story about Atari talking with Microsoft about an early version of Windows for their hardware.
- kabdib 5y agoWhether or not anyone wire-wrapped a board (pretty sure no one did), we didn't write a single line of code for the National parts. There were a number of visits by National sales reps and they dropped a lot of documentation on us, though. We were generally doubtful that the chip would be fast enough, and needed to make a processor decision very quickly; in late July 1984 we were talking about it, and I don't recall anything past August or so. By October the software folks were mostly in Monterey with a bunch of 68K-based systems. Another factor in the 68K's favor was that it was pretty easy to buy off-the-shelf 68K-based workstations (e.g., Motorola VME-10, the Apple Lisa), which let us do 4-5 months of software development while waiting for the actual ST hardware.
- cmrdporcupine 5y agoThis might bring back some flashback memories for you :-) https://www.youtube.com/watch?v=TDruBwysYoc https://www.youtube.com/watch?v=TDruBwysYoc
- kabdib 5y agoWe really didn't run GEM that way. And the ROM-based version booted without any commandline, in under a second. I miss 1980s boot times.
- pklausler 5y agoThere's some upside to having multiple register sets, though, and many quite successful ISAs have used them to good effect (CDC 6600, Cray-1/-2, &c.). It gives you more registers without having to use more bits for designators.