3 ms·
The 68020 is probably the last great CPU to write pure assembly in, too. The instruction set is rich enough to make it straightforward to write by hand - unlike
by wk_end 5y ago
The 68020 is probably the last great CPU to write pure assembly in, too. The instruction set is rich enough to make it straightforward to write by hand - unlike RISC-ier chips, C is almost superfluous for many tasks - but simple and clean enough to keep in your head (unlike x86). It cleans up some of the rough edges from the 68000, and adds some great features and a ton of speed. But: no MMU, no complex pipeline or branch prediction, and the cache is instruction-only, meaning you rarely need to think about it to get optimum performance.
- zozbot234 5y agoI disagree that RISC chips are especially hard to write for by hand - plenty of people did exactly that for the Acorn Archimedes. (And writing by hand can still be worthwhile even when you're lacking fancy CISC instructions, because you can design bespoke, custom-use "calling conventions" that make way better use of the register file. This is hard to do in C even when theoretically possible-- via fancy, non-standard asm intrinsics-- because you don't know beforehand how the compiler will choose to allocate registers.) The MC68k architecture proved a dead end, and this is ultimately what would have doomed the Amiga installed base, even if Commodore hadn't been mismanaged to the extent it was.
- wk_end 5y agoThe ARM chips were brilliant and definitely the most pleasant of the RISC chips to handwrite for, particularly with a nice assembler that handles the quirks of dealing with constants for you. While the 68k ISA was a dead-end, that's mostly because Motorola was all-in on PowerPC; AFAIK there's no fundamental reason they couldn't have pulled an Intel and thrown transistors at it until it could keep up despite the decode complexity.
- ndiddy 5y agoThe biggest problem I have with RISC assembly is that only being able to access memory through dedicated load/store instructions tends to make the code more complex with additional register shuffling.
- b112 5y agoFYI, the 68060 was fast, providing a good upgrade path from the 68040s found in a4000s. Near the end, after the initial bankruptcy, PowerPC boards, some with 060s and PPC chips both, hit the scene. Entire SDKs were introduced, and many bits of software were ported. This was a good bridge, and as Apple made this sort of jump, eg, they didn't die when switching from 68k to ppc, I don't think a healthy Commodore would have either. So I disagree. The Amiga could have transitioned too.
- msh 5y agoI think its too much to lump all RISC chips in the same bucket. I have written some assembler for MIPS and Alpha AXP back in the day. The mips assembler was quite straight forward and reasonable but I found the alpha to be very hard.