OK, can't resist the urge to swap old computer stories any longer.

  I used one of the same 1620 machines that Dave referred to. The 1620 
could add in hardware but not multiply; that was in software. Depending 
on the language (no resident OS on these things...), this could very 
well be a lookup table in memory. An errant program could easily 
overwrite this table, temporarily altering the local space-time 
continuum. 6 X 9 = 42 was well within the realm of possibility on these 
occasions.

  Also worked on the System/36. Each machine had two processors, of 
differing architecture and instruction sets. The CSP (Control Store 
Processor) did most of the I/O and low-level stuff. The MSP (Main Store 
Processor) ran the System/3 instruction set.

  The IBM 5100 came with BASIC and APL built in. During development it 
came to pass that there was not enough resource to write an APL 
interpreter for the machine. The underlying CPU was also System/3 
architecture and there was existing IBM/360 emulation code for the 
System/3 as well as APL/360. So the 5100 ran APL/360 on top of a S/360 
Emulator all on top of the 16 bit CPU. Performance in APL mode was less 
than mind-bending.

  My first 'PC' was a wire-wrapped 8080A (2MHz!) based machine of my own 
design. (I still have it in a box in the basement.) It had a whopping 4K 
of RAM (the premium 650ns stuff) and unlike the Altair/IMSAI kits of the 
day, it had seven-segment LEDs and a Hex keypad for I/O rather than the 
usual discrete LEDs and bit switches. I know it sounds weird but I don't 
recall ever running out of memory; keying in code in hex made you 
appreciate efficient design I guess...



					-=[ Steve ]=-




David Dyer-Bennet wrote:


> 
> That's too old for me.  386?  I know of the 360 and 370, of course.
> Descendants, a couple of generations later, of the 1401 I programmed
> in my first paid programming job.
>