I built a 32-bit micro-engine for a project that was
eventually going to be an IBM 360-line CPU.
I pieced the 360, not because it was the greatest design,
but it was VERY well laid-out and would be easy to write
efficient microcode for. I used the 2903 with 2910
controller. I was able to get it to run at 8 MHz, with
3-address operations running at 6 MHz.
But, the project got bogged down, as at a certain point, I
realized HOW MUCH more work lay ahead of me to get a working
system. I had to add 2 more features to the micro-engine -
a 256-way branch from the op-code, and some OR gates to OR
in the register address fields. Then, I had to build a
system bus and memory interface.
(I was going to make the I/O architecture much more like a
PDP-11 than the 360 channel architecture.) Then, I had to
design a general-purpose peripheral controller. I had a
VERY rough sketch for about a 20-chip micro-machine using
(probably) 3 byte-wide EPROMS for instructions) that would
hopefully run at 4 MHz. Then, I had to build a SCSI
controller (I already had a SASI disk on my S-100 system), a
serial mux and a tape controller.
Finally, I had to write at least a primitive OS and figure
out how to come up with compilers for it. Had I known that
UNIX-360 existed, I might have tried to make some kind of
port of that. But, obviously, YEARS of work would have been
needed to make it usable.
See http://pico-systems.com/stories/1982.html for some
pics and description of it.
Jon
