On 10/25/2018 02:24 PM, allison via cctalk wrote:
Likely make a fortune off my stockpile of 2901s. Building
machine from the earth up is not that hard, software to
make them useful is a big deal.
Yes, and that's where my 32-bit 2903 project started to bog
down. I knew some people, OS security was a total joke, so
I COULD have just stolen OS 360 MVT, but REALLY, who would
do that to themselves?
I had a few more bits of logic to wire in, to make a 256-way
branch from the OP-code field of the instruction register to
decode instructions, and from the register fields of the
instruction register to OR into the register address. Then,
I had to write the microcode. I'd done some small test bits
of microcode, including the multiply, and that worked.
(IIRC, the 2903 has an extra shift register, so it can do
the multiply step in one CPU cycle, the 2901 takes 2.)
Well, after that, I had a big decision to make. Should the
memory be on the system bus, like PDP-11 and VAX, or part of
the CPU, like IBM-360 and PDP-10? Then, I had to get memory
wired to the bit slice system, and then build peripheral
controllers. I had a very rough concept scratched up, about
30 chips to make a microcoded 16-bit machine, using fast
EPROMS for the control store. A SCSI interface would be
pretty trivial, but a read-after-write mag tape control and
an 8-channel serial multiplexer would be much more
complicated project. THEN, the big stuff would come, I'd
need an OS and language compilers. I could probably whip up
a version of CP/M with hierarchical directories and
time/date stamps, and maybe a simple editor, but the WHOLE
REASON for this project was to move up to modern high-level
languages. And, I had badly underestimated how difficult
that might become. One scheme might be to start with my
CP/M-like OS, and build a wrapper program that would allow
me to run OS-360 compilers and linkers with whatever object
libraries they needed, and then use them to compile
something more to my liking like Pascal. But, it was all
looking like a LOT of work.
So, I managed to clone a Nat. Semi 32016 system and got it
running, but it was amazingly slow.
I suspect that my kluged memory interface was not fully
optimal, but even the original that I copied was pretty
slow. Then, I spent BIG BUCKS to buy a uVAX-II CPU board
from a broker, and was finally in HOG HEAVEN! It was
certainly fast, almost the speed of the VAX-780's I used at
work, and ALL MINE!
So, that's my story.
Jon
