lizby said  

  bfwolf said  

... omitted the slow CISC 68k instructions (which require multiple clock cycles

I don't remember multi-clock instructions on the 68000. It was around 45 years ago, but I wire-wrapped a 68000 single-stepper which, as I recall, depended on the one-instruction-per-cycle characteristic. Or maybe it was mostly one cycle per instruction, with some requiring multiple button-pushes. As I say, a long time ago.

Yes, you're right: It was really easy to single-steep it from the outside using buttons - this worked both on a memory-cycle basis via the nice asynchronous bus interface and via the HALT pin (I even mean on an instruction basis).

In fact, there were some instructions that required quite a few clock cycles — for example, 'MULS' (16x16 to 32 bits) required 70 clock cycles. But that was still much faster than running it in software on a Z80.
Or "fine things" like 'MOV.L (An)+,(Am)+' used 20 clock cycles.
I confess I just looked this up in the MC68000UM.PDF ...  

  lizby said  

A great chip, though, with a flat 1MB memory architecture. It was a disappointment to me that the IBM PC didn't use the 68000.

It could even physically address 16 MB (23 address lines + UBS/LBS) and virtually address 4 x 16 MB (16 MB per user/supervisor/code/data signaled by the FCn lines) using an MMU 68451, but I think I recently read that this didn't work correctly due to a design bug that was then fixed in the M68010. The M68010 also received a vector base register and a minimal instruction cache. The address registers, including the PC, were already 32 bits wide, but I have no idea what happened when addressing anything above 16 MB?

Yes, indeed: Long long ago..

An actual 1MB of RAM on a micro was only a dream then. The U.S. Senate's IBM370 mainframe that I had access to then could only run a 1MB Fortran text-processing program of mine in the middle of the night when there were no other users--each Senator's vote on everything voted on in their careers with vote description and ordered by category ("Defense", "Agriculture", "Social Security", etc.), a one hundred-page report per Senator, double-sided, for those who had served longest.

  lizby said  

An actual 1MB of RAM on a micro was only a dream then. The U.S. Senate's IBM370 mainframe that I had access to then could only run a 1MB Fortran ...

When I started my studies, the VAX11/780 at the university, I believe, only had 4 MB of RAM.

Later in my studies (1986?), I saved up for an Amiga 2000 for over 2000 DM (I worked every semester break) – with 1 MB of RAM. That quickly became insufficient, so I treated myself to a 2 MB expansion card (expandable to 8 MB) for 500 DM, which I found on special offer. At least I could then use it as a RAM disk, which made the C compiler fun to use, as well as word processing and a CAD program for schematics and PCBs.  I couldn't afford a hard drive back then; they were far too expensive! It wasn't until much later that I indulged in that luxury: a GVP SCSI-II controller card and a Quantum LPS105 together cost 1100 DM.