Enable/34

[utzoo!decvax!duke!chico!teklabs!tekmdp!teklabs!clemc]

	People have been on me to break silence so here goes.  First
The ground rules:
	If you are an Arpa person who sends me mail, send me a PHONE
NUMBER and a time in Pacfic Time to call you back.  (Before 10 Pac Time
is unacceptable, I'm a Computer Scientist and like most, hack in the
evening).
	If you want copies of the Paper send me a letter at this
address and I will mail you a copy.  Please include a xerox of your V7
license agreement.  I include in the appendix the code, so we need to
hang on to such.  If you do not send the license, my secretary will
remove the apendix.
	Clement T. Cole
	Computer Research Laboratory
	Tektronix Laboratories
	Tektronix Inc.     M/S 50-384         <<-- Very important
	PO Box 500
	Beaverton, Or 97077


	Mike O'Dell at LBL-unix is trying to set up a place from which
to FTP the paper and the code.  Once I know how to do it (and I get it
to his machine, I will send a note.  All of this has been submitted to
USENIX, but I understand people want it now.


	The ENABLE/34 should be the "shot in th arm" that Mike
mentioned for not only 40 but I beleive 45's.  Electrically, it works
on an 11/45.  If you want to try to use it that way, the ground rules
are simple:
	1st) No Fastbus memory   (sorry it's not on the Unibus so life
is rough)
	2nd) No supervisor map  ( a minor misstake.  No body though
about it I guess)    This means copyseg will have to be rewritten.

	What you get in return:

	4 Meg's of addressability, without losing your old 18 bit
		Unibus Memory!!!
	An I/O Map like the 11/70 class processors.
	Increased TOTAL system proformance.

	The ENABLE will slow down fetch times.  It is just
a glorified bus repeater.  But remember you were swapping before,
so you are trading in access time for swapping time.  (Seems
like a fair trade to me).

	How it works:

	The ENABLE terminates the UNIBUS and creates two new bus's out
its back.  One is a werido bus for the old 18's memory and to use it
will take MODING an old Unibus cable and a finding a small backplane(s)
to put the old memory in.  The other bus is the new 22 bit 11/44 memory
Unibus (so called MUD bus).  On this bus is the new memory (get it from
who ever.  I have used both DEC 11/44 memory by raping my 11/44 and
MOSTEK memory from a loan).
	If we remember that adding zero to any number is the identity
element, we can understand how the ENABLE works.  The ENABLE runs while
DEC Memory Mapping is turned on!!  The DEC MMU will be set ONCE to
effectively add zero to the 16 bit address produced by the running
program.  The extra bits supplied by the DEC MMU are used as indexs for
the ENABLE MMU. (See our paper for details)

	Further Adventures:
		Once I get an 11/45 to play with, we will put this on
an 11/45 and rerun all benchmarks.   We currently have the new DEC
distribution, and will make changes for their overlayed kernel.  I'm
told the UCB 11/34 kernel showed up today, so we run them in it.   When
we have a new set of numbers and a better benchmark, we will let people
know.

	FYI:
	I talked to ABLE yesterday, an they are quoting about 30 days
for ENABLE boards.  This is just to give you an idea.  about
deliveries.  I suspect getting memory will be more of a problem.


	Clem Cole  teklabs!clemc
		Leave messages for me at (503) 627 5465 with
			my Sec. (Marcy Briggs)