x'15' (21): Did MTS require a special version of the S/360-67 to run?

posted Dec 15, 2014, 3:52 AM by Jeff Ogden   [ updated Dec 15, 2014, 4:11 AM ]
Here is the backup e-mail on the topic. The topic came up on the Hercules 390 Yahoo Group e-mail list and later there was a series of private messages and an item posted in the Myths and Misconceptions section of the MTS Archive web site.

The cast of characters:
  • Richard (u4gh) is Richard Chycoski, a former systems programmer at Simon Fraser University (SFU).
  • Dave (G4UGM) is Dave Wade, a former MTS user at Newcastle Polytechnic (part of NUMAC in the UK).
  • Mike Alexander is a former Systems Research Programmer and Research Scientist at the University of Michigan Computing Center, Mike, together with Don Boettner, was chief software architect for the Michigan Terminal System (MTS).
  • Jeff Ogden (me), a former counselor, Systems Research Programmer, Associate Director, and Sr. Associate Director at the University of Michigan Computing Center.
  • Gavin Eadie, a former Systems Research Programmer and Associate Director at the U-M Computing Center.

From: Richard Chycoski
Sent: 13 December 2014 01:28
To: hercules-390@yahoogroups.com
Subject: RE: [hercules-390] System 360 Model 67 Reference Card

The MTS archive at <http://archive.michigan-terminal-system.org/documentation/documents> has a downloadable copy of the 360/67 Reference Card. MTS started on a special version of the 360/67, and University of Newcastle was one of the early MTS sites. The MTS site where I worked (Simon Fraser University, near Vancouver, BC) didn't start running MTS until the later '70s, so we didn't have 360/67 reference cards lying around. :-)

- Richard

On Dec 13, 2014, at 12:52 AM, Dave G4UGM wrote on the Hercules 390 Yahoo group:

Whilst MTS used the 360/67 in a special way it was just a standard 360/67 with DAT and 32-bit addressing. However the 360/67 did start life as a modified 360/65.

There is one feature of MTS that folk’s mistake for a hardware modification, as it is implemented via a Pseudo OP code in assembler.

It is explained in:-

http://en.wikipedia.org/wiki/MTS_system_architecture

in the section entitled “Branch on Program Interrupt” (BPI), but it DOES NOT require any hardware modification.

Dave
G4UGM

Sent: 13 December 2014 09:30
To: hercules-390@yahoogroups.com
Subject: Re: [hercules-390] BPI on modern machines (was System 360 Model 67 Reference Card)

Dave,

What a great instruction- a NOP without an index register (which is used
as mask for the expected/supported program interrupt).

Would it break anything if this would be implemented on current hardware?

I know it would help address-verification dramatically. Maybe there are other[s.]

From: "Dave G4UGM"
To: hercules-390@yahoogroups.com
Date: Sat, 13 Dec 2014 09:34:26 -0000
Subject: RE: [hercules-390] BPI on modern machines (was System 360 Model 67 Reference Card)

I am not sure about this. I know MTS ran on 370 hardware so the code works
on those "out of the box" as well. I know there were tweaks for 370 mode but
not sure if they were in this area...

Dave

From: Mike Alexander
To: hercules-390@yahoogroups.com
Date: Sat, 13 Dec 2014 13:53:50 -0500
Subject: Re: [hercules-390] BPI on modern machines (was System 360 Model 67 Reference Card)

Thanks, I always thought it was one of my better ideas. It was very useful in MTS, not just in the system, but in programs that ran in MTS.

You don’t really need hardware changes for this since it’s extremely easy to implement in the program interrupt handler. Implementing it in either way might break something since it’s possible that some code contains one of those special NOPs for other reasons. This still wouldn’t be an issue unless the immediately preceding instruction got a program interrupt. In 20 years or so of use of this I never heard of a case where it caused problems.

Mike

From: "Dave G4UGM"
To: <hercules-390@yahoogroups.com>
Date: Sun, 14 Dec 2014 08:22:49 -0000
Subject: RE: [hercules-390] BPI on modern machines (was System 360 Model 67 Reference Card)

So was it ever implemented in hardware?

From: Mike Alexander
To: hercules-390@yahoogroups.com
Date: Sun, 14 Dec 2014 14:16:26 -0500
Subject: Re: [hercules-390] BPI on modern machines (was System 360 Model 67 Reference Card)

The BPI instruction? Not that I know of. Since there was no significant advantage to having it in hardware we never asked anyone to do it.

Mike


On Dec 13, 2014, at 3:54 AM, Dave G4UGM wrote:

Folks,
 
Perhaps you could add something to say MTS used a “standard” 360/67. Many folks seem to thing BPI needs a hardware mod…
 
Dave
G4UGM

On Dec 13, 2014, at 5:48 AM, Jeff Ogden wrote:

We can add something.

Where are people expressing their opinion / confusion about BPI and hardware mods?

Is there an article or other description somewhere that is causing the confusion?

Where do you think that we should add something?
On the MTS Archive site? The Myths and Misconceptions page?
On Wikipedia? There is a description of BPI in the MTS System Architecture article.
Somewhere else?

   -Jeff

On Dec 13, 2014, at 9:51 AM, Dave G4UGM wrote:

They generally say “MTS needed a hardware mod” or “MTS ran on special hardware”, and when pressed they say “something to do with interrupt processing”.
 
Got one yesterday on the Hercules 390 group:-
 
“The MTS archive at <http://archive.michigan-terminal-system.org/documentation/documents> has a downloadable copy of the 360/67 Reference Card. MTS started on a special version of the 360/67, and University of Newcastle was one of the early MTS sites. The MTS site where I worked (Simon Fraser University, near Vancouver, BC) didn't start running MTS until the later '70s, so we didn't have 360/67 reference cards lying around. :-)”
 
So I am not sure if it’s that’s a typo, or they mean a special version of the 360/65, but the same thing cropped up in 2010, from Rob Tatum who is usually a pretty competent guy
 
From: rhtatum
To: hercules-390@yahoogroups.com
Sent: Tue, October 19, 2010 10:33:40 PM
Subject: Re: [hercules-390] Current status of MTS ?
  
No, MTS probably won't run under Hercules - I looked up MTS using a google  search, and it seems that there was an additional operation code that the folks that created MTS either did themselves or had IBM implement for switching  between pieces of MTS, user programs, etc. Someone would have to write the simulation code for that op-code for MTS to run under Hercules-390.
 
Is two enough to be “common”?
 
Dave Wade

On Dec 13, 2014, at 10:47 AM, Jeff Ogden wrote:

That should be enough to get started. Thanks.

   -Jeff

On Dec 13, 2014, at 12:25 PM, Jeff Ogden wrote:

Mike,

The S/360-67 at Michigan had an RPQ or perhaps several RPQs.  Were they required to run MTS or just performance enhancements?  At onetime I think the RPQ was described in MTS Volume 5, but the PDF that I have is a newer version which doesn't include that description. Was Search List (SLT) part of the RPQ? I know that UMMPS emulated the SLT instruction, but I'm not sure when that was added? And I know that UMMPS stopped using STL at some point (probably when S/370 support was added).

I added a new item to the Myths and Misconceptions section of the MTS Archive Web site about this. Could you take a look and see if I got it right?

   http://archive.michigan-terminal-system.org/myths#TOC-Myth:-MTS-requires-a-special-version-of-the-IBM-System-360-67-to-run

   -Jeff

On Dec 13, 2014, at 1:37 PM, Mike Alexander wrote:

I saw that message on the Hercules list and even had a reply typed but decided not to send it.  I guess I should have sent it.

There were a couple of RPQs on Michigan’s first 67, but none were required for MTS since other 67s that ran MTS didn’t have them. I’m not even sure our later duplex 67 had them.  One of them was the SLT instruction, but that was never used much and was never required.  All uses of it were under conditional assembly and were optional.  The emulation code was added later when we got machines that didn’t have the RPQ so that any user code that used SLT would still work.  The system never used the emulation code and I doubt if any user code did.  I had a counter on the emulation and never saw it non-zero except when I was testing it.  I can’t recall offhand what the other RPQs were, or even for sure if there were any.  I think that some things started out as RPQs, but became part of the standard machine.

Your new item in M&M looks fine.  You could add something to it about SLT.

           Mike

On Dec 13, 2014, at 4:20 PM, Jeff Ogden wrote:

OK, thanks.  I think there might have been an RPQ for extended floating-point, but I am always confused about what was done in hardware and what was emulated in UMMPS.

I wouldn't feel too bad about not sending your note. That was a few years ago now (4) and the original comment started out that MTS probably wouldn't run under Hercules.  I think subsequent events have made it quite clear that MTS runs just fine under Hercules.

  -Jeff

On Dec 13, 2014, at 5:07 PM, Mike Alexander wrote:

I think that's right that there were floating point related RPQs. There might have been others too.

The message I almost replied to was from yesterday.  It doesn't matter that I didn't reply since "Dave G4UGM" replied with the same comment I was going to make.

         Mike

On Dec 15, 2014, at 4:03 AM, Jeff Ogden wrote:

I found this in my e-mail archive:

On Aug 2, 2010, at 6:22 PM, among other things Mike Alexander wrote:

Michigan asked for a few things beyond just address translation hardware, including extended precision floating point.  There was concern that the base 16 floating point (which had never been used before) would have problems which extended precision might address. These fears turned out to be true (Len will certainly remember the guard digit fight).  The other features IBM put in for Michigan were the search list and swap register instructions.  When the 370 came out it also had 128 bit floating point, but not quite the same as on the 67.

On Dec 13, 2014, at 4:44 PM, Jeff Ogden wrote:

See: 

http://archive.michigan-terminal-system.org/myths#TOC-Myth:-MTS-requires-a-special-version-of-the-IBM-System-360-67-to-run

   -Jeff

On Dec 13, 2014, at 4:59 PM, Mike Alexander wrote:

That looks fine.

       Mike

On Dec 13, 2014, at 5:41 PM, Dave G4UGM wrote:

I think that’s fine Jeff, Thanks very much
 
Dave

From: Richard Chycoski
Sent: 14 December 2014 01:45
To: dave.g4ugm
Subject: Re: System 360 Model 67 Reference Card
 
Hi, Dave: I've read this in a number of sources, including <https://www.eecs.umich.edu/cse/publications/Publications/CSE_Booklet.pdf>:

Ultimately, IBM’s Model 360/67 was installed at
NUBS in January of 1967. Close cooperation between IBM people
and the Computing Center staff produced modifications of this
design, which became then the IBM 360/67M, the “M” standing
for “Michigan.” The two Computing Center people who were the
primary developers of all the details of MTS were Mike Alexander
and Don Boettner. One measure of their success is the fact that
within a year of the installation of this first IBM360/67M, IBM had
orders for forty more.

This is also how I heard it described by the MTS old-timers (now we're all old :-), including John Hogg (UBC) and Charlie Benet (UM/UQV/SFU).

So that would make it true that UM got 'vanilla' 360/67s, but were later modified and that version sold to others who wanted these added features?

- Richard

From: "Dave G4UGM"
Subject: RE: System 360 Model 67 Reference Card
Date: December 14, 2014 3:43:57 AM EST
To: "'Richard Chycoski'"
Cc: "Jeff Ogden"

Ok so if you look at the references in the end of the booklet, you will see it refers to this:-
 
“Susan Topol, “A History of MTS--30 Years of Computing Service,” published in the University of Michigan Information Technology Digest, May 13, 1996 Vol. 5, No. 5).”
 
You will find that article reproduced in its entirety here:-
 
http://web.archive.org/web/20060914022828/http://www.msu.edu/~mrr/mycomp/mts/others/feat02.htm
 
and its clearly been copied wrongly as that says they got a 65M (a model 65 with DAT) which became the 67….
 
Dave
G4UGM

On Dec 14, 2014, at 6:24 AM, Jeff Ogden wrote:

The Suzan Topol article that Dave mentioned in his reply is a good source for this.

I'm sure that UM got a "standard" S/360-67 in December 1966 / January 1967. It may have had some additional RPQs on it, but they were not related to the virtual memory features of the Model 67 and were not required to run MTS. 

At Dave's urging I added a new item about this to the Myths and Misconceptions section of the MTS Archive web site. Mike Alexander has been following this discussion (he is a member of the mts-comments@umich.edu e-mail group as well as the Hercules 390 e-mail group on Yahoo). Mike looked over the new Myths and Misconceptions item and said it looked fine.

There is in fact some uncertainty about what the actual designation of the one-off version of the Model 65 was before it became a standard product from IBM. It may have been the model 65M as stated in the Susan Topol article, but it may have been the Model 66 or 66M or perhaps a mix of all three at different times. But the uncertainty is about the model designation of the preliminary system and not about what Michigan finally received in December 1966 / January 1967.

There are a number of other sources for this and related topics:
  -Jeff

On Dec 14, 2014, at 11:24 AM, Richard Chycoski wrote:

Thanks, Jeff!

Given the muddiness of the remembrances, it's not too surprising that several different versions of the story got propagated through the community - sometimes even into print, where it became 'gospel' :-).

- Richard

On Dec 14, 2014, at 11:29 AM, Gavin Eadie wrote:

History is written by the winners?

On Dec 14, 2014, at 11:44 AM, Richard Chycoski wrote:

Or by the ones with the largest publishing budget - there's even research to suggest that some sections of the Bible in its current form came from copies-of-copies-of-copies of errant manuscripts, because so many copies of the flawed document got propagated, drowning out the earlier (and presumed to be more correct) copies that were later found. Repeat some untruth often enough and it can get to be believed by everyone - too many tyrants have made terrible use of this.

As I've gotten older, I've found how 'fluid' memory is - having memories of things that I saw as a kid, then looking at pictures from the actual time/event. Makes the whole idea of an 'eye witness' quite suspect!

- Richard


Comments