Subject: Re: FPGA VAX update
From: Johnny Billquist <bqt at Update.UU.SE>
Date: Tue, 25 Oct 2005 15:38:27 +0200 (CEST)
To: cctalk at
classiccmp.org
On Tue, 25 Oct 2005, Ken Seefried <ken at seefried.com> wrote:
From: Johnny Billquist <bqt at
Update.UU.SE>
I think I have the documentation for the
microcode for
the different engines, and I also have the binary microcode
files...
Do you have the microcode files for the 11/730?
Might still have the TU58 tapes around, yes. No TU58 drive to read them,
though.
I must be thee rare one that maintaines TU58s for use.
The Am2901 VHDL (and other formats) is freely
availible and well
understood and fits several to an FPGA (even small ones covered by free
tools). The rest of the logic on an 11/730 CPU isn't that terrible
complicated. Given access to the microcode, I'd imagine it would be a
reasonable project to whip up an ersatz 11/730, validate that it works
correctly, then proceed to optimize and otherwise improve the design
(add in pipelining, test, add in FPU, test, add in cache, test, migrate
to faster FPGA family, test). Iterative, rather than shooting the moon
first run.
Maybe. I think that there is probably a lot more logic outside the 2901
than inside... We're still talking several boards of logic here... I doubt
all of that is microcode memory, 2901s and 74xx series stuff. :-)
Bet on it as the 2901 only provides 16 registers and I think VAX had more.
It's also a slow device.
You could even
short-cut some of the FPGA work. I don't recall what
the microsequencer looks like on the 11/730, but if it's the usual
Am29{09,10,11}, there is at least one shop that still (as of at least
last year) makes an 8 x Am2901 + sequencer single chip ASIC that runs at
something like 20MHz. Or you can find NOS for the IDT 49c402 that is 4
x Am2901s.
This way of thinking won't result in the fastest VAX possible, but I
venture it has a better chance of resulting in an actual working VAX.
It might be a possiblity. Not that I'll do it. But if someone really
wanted this to happen, it's one possibility. Might be easier to talk with
someone who still have an 11/730 running, to read his tapes though.
Not I. I must ahve at least 16 2901C and 2911 and they are not that pretty
without a lot of glue. More correctly, for some archetectures it's ugly
and for some a very easy set of peices to use. The biggest problem is
multiplexed external busses and that does impact design and speed.
Allison