Thanks for correcting me, BigEd.

It's an interesting question why there were more 68000 computers than there were 65816 computers.

Apple IIGS had a graphical user interface, so the reason can't be that the 65816 might be "too weak"
compared to the 68000 when it comes to "mouse pushing" while having colored windows on the screen.

Hmm... maybe there were more C compilers for the 68000 than for the 65816 in that time frame ?
Writing code for colored windows in C might be less fuss than writing code for colored windows in assembler...

Commodore never did a 65816 computer, and went for 68000 Amiga while axing the C65.

Yes, I would think C, and lots of nice wide registers, made the 68k a clear choice. (Having supervisor mode is good too. At that point proper operating systems were becoming normal.) The '816 only has the advantage of compatibility, and probably of price.

Recall Acorn's experience with the '816 and with everything else they tried: insufficient memory bandwidth to make full use of the speed of the CPU chip. So they invented ARM. The '816's byte-wide path to memory is an extreme case, like the 68008 and the 8088. It's only a good idea for a small cheap machine - which is a market.

The 68000 architecture (instruction set and registers) certainly was invented with being a good choice for compilers in mind,
while the 65816 was not: 65816 focused on backward compatibility to the 6502 which was invented for small embedded systems
using assembler code.

68000 is out of production now, but ARM and 6502\65816 still are big in the embedded sector.

Hmm... somewhere earlier in this thread, I had mentioned the 68k compatible Fido1100 from Innovasic.
There was a story, that Innovasic initially had plans to have a 200MHz ARM core inside the chip,
but customers insisted in having a 60MHz 68k compatible core instead.

68k had been big in 19" industry control applications.
For instance, the Deckel\Maho Dialog 12 CNC was 68000\68020 based (there was a story that the designers had used a Pascal compiler instead of C...).

I think ColdFire by NXP is still 68k inside, with a gcc port, but I could be wrong.
http://www.nxp.com/products/microcontro ... us:PC68KCF

68k also looked like PDP11, and well suited to hosting professional operating systems. Also, the 68k had a lot more upgrade potential than the 65816.

BTW, fans of the 68000 (or any other unusual CPU) might be interested in the anycpu forums, and in particular in this post about a well-documented ongoing build:
http://anycpu.org/forum/viewtopic.php?f=23&t=339

From what I heared, ColdFire isn't binary compatible to 68k, and Wikipedia says compared to 68k ColdFire lacks some instructions and addressing modes.

The product\price politics of Motorola sure was somewhat confusing in the time frame "after" 68k.

There was the ill fated 88k, the PPC (power performance chip), ColdFire...

Hmm... there is a story, that when the C64 was designed the designers didn't exactly meet the project deadline,
so they received a payment cut for this and then jumped ship.
And when the C128 was designed nobody at Commodore knew how the VIC2 from the C64 internally worked.
I just wonder if a similar thing had happened at Motorola with 68k.

Anyhow, development of the 68k family did end with the superscalar 68060, which internally translated 68k machine code
into something like 48 Bit RISC instructions.
Some less frequent 68k instructions were missing in the 68060... like 32*32 Bit multiplication.


Yep, PDP11 and VAX certainly had some influence on the 68k design.

The nice thing with 68k was, that (if we ignore "move from SR\CCR") the instruction set architecture
was pretty much the same for the small 68008 (8 Bit data bus), the 68000, and 68010...68040.
(I'm not sure about the 68060).
Stack format for responding to an exception did vary, of course.
The 68040 had a crippled FPU inside and featured a different bus system, but I'm getting off topic.

68k had upgrade potential: there was a story that in theory, one could add up to eight 68881\68882 FPUs (floating point coprocessors) to a 68020.
But I can't remember that there ever was something like a floating point coprocessor for the 65816,
despite that there is a COP instruction in the 65816 opcode table.

Interesting, thanks - so ColdFire is a kind of subsetted 68k. There are details here:
http://microapl.com/Porting/ColdFire/cf_68k_diffs.html

There are still a bunch of M6833x processors for sale, with CPU32 core. It looks like the CPU32 is much closer to 68000 than the ColdFire.

Fido1100 might be a better choice because unlike the 68332 it has SDRAM interface, unfortunately the documentation has some "potholes",
and it feels like the IC package might be ball grid array in the future.

Also, Fido1100 has five 68k register sets, and if I remember correctly one could switch to another register set if an interrupt hits
(that's one clock cycle interrupt response time... at 60MHz.)

Some years ago we had toyed with that chip at work, but because we were not pleased with the IDE and the documentation,
the eval board had ended up in the trash can.


CPU32, like in the 68332 and the Fido1100, basically is a 68020 subset, some of the more elaborate 68020 addressing modes are missing,
and I think that there are no FPU instructions (haven't checked).
But the instructions and addressing modes that are there in CPU32 seem to be 100% binary compatible to the 68020.

;---

When ColdFire came out, I took a look at the datasheets and to me it felt like ColdFire is something completely different from 68k.

I'd say, decoding the 68k instruction set ain't easy (sort of an "Origami trick"), so ColdFire was an attemt to have an instruction set
which decodes more nicely, maybe for reducing cost and improving CPU speed.

68000 uses microcode, plus three PLAs for decoding the instruction into a 10 Bit microcode address.

From the 65816 patents, it looks like the 65816 uses only one big PLA (similar to 6502 and 65C02).

;---

I just remembered the obsolete TMS320C31 DSP from TI.
Instruction set was very different from 68k, of course...
but it had 8 data registers (32 Bit), 8 address registers (24 Bit ?), and 8 floating point registers (40 Bit).
Another neat architecture that went forgotten.
...Long time ago I did a little bit assembly coding for that chip at work.
But it's been too long, and I don't remember much of the details.

Ha, the book he mentioned was my course textbook for Computer Systems Design at UKC in '87.

Looks like the CPU32 is a superset of 68000, but a subset of 68020. Or from Wikipedia:

The instruction set of the CPU32 core is similar to the 68020 without bitfield instructions, and with a few instructions unique to the CPU32 core, such as table lookup and interpolate instructions, and a low-power stop mode.

I used a 68360 at 3Com around '95-'97, as the control processor for Ethernet switches. Most of the work was in C++, but I did need to drop into assembler at one point and felt at home with it.

That was the last time I saw 68k in any non-nostalgic context.

If such a device exists it is a well-kept secret. COP, incidentally, is nothing more than a software interrupt with its own vector.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

That might be. The late '70s/early '80s experimented with a number of different CPUs (Z80, 6502, 6809, TMS9900 (on the TI-99/4A)). All the manufacturers were looking for an edge at the right price. By the early/mid 80s when the Atari ST and Commodore Amiga came out, mind share on CPUs had shifted to the 68000, which was the first affordable 32-bit microprocessor (16-bit bus, but 32-bit forward-looking ISA).

I have no nostalgia or attachment to the 65816, and looking at the datasheet I see no reason why anyone would have selected it over the 68000. Demanding compatibility with the 6502 would have been planting your feet firmly in the past while the industry was quickly moving forward. The 68k generation of home computers had big memory, big screens and big operating systems. They propelled a generation of hobbyists into a more powerful world.

If the 6502 is your hammer, and if feels comfortable in your hand, serves you well and it's the first thing you reach for... then maybe the 65816 is a version with a larger head that can drive in larger nails. It still fits your hand perfectly, has great balance but it can do more than your old hammer. It's your new favorite, and for the work you do it never lets you down.

The 68000 isn't that hammer. It's strange looking and uncomfortable. It's not clear if it's even worth getting used to. A hammer is a hammer... why should I spend time with this new one if I'm just pounding nails and I have the perfect hammer already?

I've learned to use a lot of hammers in my time, and to me the 68000 was a lovely hammer for its time, and so was the 6502.

Wikipedia says the 68000 was introduced in 1979, but that National's 32016 processor which became available in 1982 may have been the first 32-bit chip to reach mass production and sale, at least according to National Semiconductor marketing. I'm not sure that totally jibes, since the 68K was already going into laser printers by then. I don't see any price info on the 32016 though. Sophie Wilson, chief architect of the ARM processor, said, "an 8MHz 32016 was completely trounced in performance terms by a 4MHz 6502." (The 32016 was National's 32-bit processor, having 15 registers, including 8 general-purpose 32-bit registers.)


Interrupt performance.


Well put.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?