Thanks for the additional details, Don. You definitely know more, and
were more involved than I ever was.
A couple of additional comments...
On 2020-07-10 20:52, Don North wrote:
On 7/10/2020 7:20 AM, Johnny Billquist wrote:
On 2020-07-10 14:19, Paul Koning wrote:
On Jul 9, 2020, at 10:40 PM, Johnny Billquist <b...@softjar.se> wrote:
On 2020-07-10 04:37, Don North wrote:
On 7/9/2020 6:25 PM, Bob Supnik wrote:
Yes, the PDP11 Architecture Handbook was a post-facto effort. The
J11 was finished; DEC did not intend to do another PDP11
processor. (I wrote a spec for one, primarily as an exercise in
trying to do a different microcode structure than the PLA/ROM of
the LSI11/F11/J11, but I lost it.) The only formal part of the
PDP11 architecture was the Commercial Instruction Set extension,
DEC STD 168, which was only implemented by the F11 and the 11/44.
AND the PDP-11/74 CIS option, I might add. Fully implemented, never
sold.
Not to mention that the 11/74 in itself was fully implemented, but
never sold... Not even the 11/70MP...
Were there actually two prototypes called 11/74? I know the MP
machine, which the RSX-11 development team owned. And in Merrimack
(home of RSTS and some of the compiler teams) there was an 11/74 with
CIS, for COBOL. But that one wasn't an MP machine. Perhaps a
coincidence, I don't have a real memory either way.
Well, yes and no...
The PDP-11/70MP is a modified KB11-C. In the end it's really just a
modified 11/70. There are changes to the microcode, and if I remember,
a couple of boards were changed. (See
http://www.bitsavers.org/pdf/dec/pdp11/1174/EK-70MP-TM_PRE_1170mp_Prelim_Technical_Manual_1977.pdf).
The PDP-11/74 is the KB11-E. This was a more major redesign, which
then allowed for the addition of CIS. (See
http://www.bitsavers.org/pdf/dec/pdp11/1174/Prelim_KB11-E_Diffs_Aug78.pdf).
But depending on what documentation you read, they might both be
called 11/74.
Johnny
What Johnny said is basically correct. Way back then there were two
separate programs going on:
1) a research program led by Verell Boaen (IIRC) on turning a standard
11/70 in an MP config. As indicated this required minor updates to the
11/70 microcode (to make the ASRB instruction act as a hardware
interlock) and the IIST board for interprocessor communication.
Multiport memory boxes also as I recall. This project was termed
'11/70-MP'.
It also required adding the cache bypass to the PDR register, the cache
flush functionality, and enabling and disabling of the cache. So a few
more changes, but nothing really surprising. Also yes, it required the
MKA11 memory boxes, which had four ports. (And obviously the IIST.)
2) a hardware program led by Dick Brown (IIRC) in PDP-11 engineering to
add CIS functionality to the 11/70 and update the CPU to what was
called, then, the 11/74. Some changes to CPU microcode to support CIS
instruction dispatch and communication with the four board CIS
processor. Changes to 11/70 backplane. Changes to a few CPU modules (not
many and not complex).
Sadly the document I linked to, about the KB11-E differences, would
appear to be incomplete. It lists some of the basics, but later
chapters, which described the backplane and module list is missing. It
would be interesting to see that one.
I know I have seen lists of which modules were changed/different for the
KB11-CM, and I think I've even read that some of those got around
because they could be used in any 11/70.
At some point the CPU microcode update for ASRB was folded into the
11/74 CPU also, making it MP compatible.
Makes sense. The same is also true of the J11.
I even have a plex of the updated 11/74 front panel I 'appropriated'
when the 11/74 was canceled, Here is a scan:
https://www.ak6dn.com/stuff/1174.jpg
Nice. I think I've seen it before (you probably posted a link in the
past). It would have been a really nice machine, if DEC only had
released it.
Ultimately DEC marketing canceled the 11/74 CIS. There have been
arguments and justifications over the years as to exactly why, but my
communications with the marketing folks at the time was that when
benchmarking Cobol performance the 11/74 CIS significantly outperformed
the 11/780 (which had just been released). So marketing had a
positioning problem and the solution was to kill the 11/74 in favor of
promoting the 11/780 family to eliminate the issue. Inarguably a wise
decision at the time as the VAX was to be the future of DEC. At least
until Alpha...
Interesting, and a bit of a parallel story to why the 11/74 got
cancelled in general. Legend has it that in the MP setup, it
outperformed the 11/780. Of course, another rumor was that it was too
complex to deal with, especially with all the flat cables going around.
And yet another story was that it would have been too much special
tailoring on a per customer basis, which was contrary to how most of the
PDP-11 business was going, and was more into PDP-10 land...
Who knows. Either way, the machine would have been really nice to have a
few surviving examples of, but I suspect the chances of that happening
now are low. As far as I know, the last working machine stopped
operating in 2001 or so. Seems unlikely that it would still be around.
The 11/44 had a parallel implementation of the same CIS architecture (in
fact I have an 11/44 with the hardware CIS option in my garage that is
fully functional. Sadly no 11/74 ...) but it was much lower in
performance (and cost) than the 11/74 CIS.
I was never particularly charmed by the 11/44. Too much just "business"
feel about it.
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: b...@softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
_______________________________________________
Simh mailing list
Simh@trailing-edge.com
http://mailman.trailing-edge.com/mailman/listinfo/simh
