That was beautiful. Nicely done, and congratulations on your restomod A1000!
I’ve heard people describe it as a 16 bit computer a few times lately though and it always catches me off guard. It was usually referred to as a 32 bit computer at the time, and defensibly so.
Some of its implementation details were 16 bit, like the data bus and ALU, but they were largely invisible to users and programmers. Assembly code used 32 bit math instructions, even if the CPU executed them in 2 steps. It had a flat 32 bit address space, although only 24 address lines were implemented (kind of like how not all 64 address lines are available on a 64 bit CPU today). Registers were 32 bits wide. And later 68K CPUs could run A1000 software on pure 32 bit CPUs natively with no emulation or trickery.
Credit it with those extra 16 bits. It earned them.
It was pretty commonly accepted at the time that the 68000 was a "16-bit CPU" because of its ALU and data bus width, despite the 32-bit being clearly ISA. The Sega Mega Drive/Genesis, for instance, is always referred to as a 16-bit console, even having it embossed right on its shell.
The technical truth is that using a single number to describe the "bits" of a CPU is going to be inadequate in many cases (including this) so this was a choice by marketing.
I'm not sure why that was the case; one would think marketers would have rushed to use the biggest, most impressive sounding numbers available. Maybe they wanted to keep something in reserve for when the 68020 came down the pipe? Maybe they were scared of being criticized or even sued for false advertising since the chip wasn't "truly" 32-bit?
Well, I don't think there was any consensus. Atari played on this with the "ST" (for sixteen/thirty-two) product label. From the software side it was 32bit. The flat, huge (at that time) address space made things so much easier (compared to the awful segmentation of the Intel CPUs then). That the ALU was just 16bits meant that some operations took longer than one would have liked, but was otherwise of no concern. No "far pointers" or any such crutches.
That's how I see it. As a developer, you still used 32-bit math operations in assembly and the CPU executed them as a series of 16-bit ops. The details are abstracted away. IMO, that made it 32 bits.
By analogy, SATA is a serial protocol, but you wouldn't ordinarily refer to it as a 1-bit bus.
> That was beautiful. Nicely done, and congratulations on your restomod A1000!
Author here. Thanks, had a lot of fun! Not sure why the original post from a few days ago didn't merge with this one, it's the same url ¯\_(ツ)_/¯ https://news.ycombinator.com/item?id=42832032
Yeah, you posted it a few hours before I did (https://news.ycombinator.com/item?id=42839928), and neither got voted up (and, mysteriously, mine was not detected as a dupe inside a very short time window, as it seems like it should happen).
I'm going to blame the first part on us being in Europe and most of the HN crowd living in US timezones :)
I’ve heard people describe it as a 16 bit computer a few times lately though and it always catches me off guard. It was usually referred to as a 32 bit computer at the time, and defensibly so.
Some of its implementation details were 16 bit, like the data bus and ALU, but they were largely invisible to users and programmers. Assembly code used 32 bit math instructions, even if the CPU executed them in 2 steps. It had a flat 32 bit address space, although only 24 address lines were implemented (kind of like how not all 64 address lines are available on a 64 bit CPU today). Registers were 32 bits wide. And later 68K CPUs could run A1000 software on pure 32 bit CPUs natively with no emulation or trickery.
Credit it with those extra 16 bits. It earned them.