1.0: The early days at UM

posted Nov 21, 2010, 12:11 PM by Jeff Ogden   [ updated May 3, 2014, 3:02 PM ]
ACG: Andy Goodrich
BSC: Brian Cashman
DWB:  Don Boettner
GCP:  Gary Pirkola
JCO:  Jeff Ogden
JS:     John Sanguinetti
MTA:  Mike Alexander
WSG: Scott Gerstenberger

On Nov 7, 2010, at 1:04 PM, Jeff Ogden wrote to Don Boettner, Gary Pirkola, Mike Alexander, and Bruce Arden:

I know that Don and Mike were the first MTS developers. I've always assumed, but I'm not sure if anyone actually told me, that Mike mostly did UMMPS and Don mostly did the command language.  Is that fair?

MTA: Yes, that's correct.

I don't suppose either of you keep that first bit of code "written out on a kitchen table" in the house on Church Street (it was on Church St. wasn't it or was it on Oakland?)?

MTA: It was on Sylvan Street.  I moved to Church Street much later and Don never lived there.

Who else was involved from say 1966 until November 1967 when VM support was turned on?  I know some of the names, but I'm unclear when individuals actually started and who worked on what.

MTA: I don't recall all the names, Don will do better I'm sure, but Kip Moore did the file system.  Dave Mills did the Data Concentrator in that period too.  The original Data Concentrator was first attached to the model 50, even before we got the 67.  Did you do the first DSR for that, Scott, or did someone else?

Al Emery worked on Memorex 1270 support. When was that?  Did he work on IBM 2702 or 2703 support too?

MTA: Yes, Al did all the IBM terminal controllers.
JCO: Al was also an Associate Director and later the Deputy Director of the Computing Center. The operations staff and the business office staff reported to Al.

Scott, you worked with Dave Mills, Jack, and Dave Flower on Data Concentrator support. Who else was involved?

WSG: As I recall, I wrote the DC DSR (PDP8RTN) somewhat basing it on a 1050 DSR which I think Don had written. But the 1050 routines used PCIs for processing input which made the DC code quite different. Dave Mills, Jack and I (and others) all worked on RAMP. I don't think that either Jack or Dave ended up writing any parts of MTS but maybe I've forgotten. I also worked a lot on the graphics version of RAMP that we used for the CONCOMP project. The fundamental parts of RAMP were the same in both versions but each system supported a different set of I/O hardware, e.g., the 360 channel interface, the 338 display, a hard drive disk, paper tape hardware, all kinds of serial communications devices, the Grafacon (sp?) tablet, etc.

WSG: The PDP8RTN development (by me) paralleled (but lagged somewhat behind) the PDP8/360 channel interface hardware development and support for same in RAMP, both done by Dave Mills. I think Dave also wrote the PDP8 assemblers both for the 7090 and MTS.

WSG: Jack [DiGiuseppe] wrote various parts of RAMP as did a few other people.
ACG: Jack D. worked on PDP-11 terminal driver development, some DSR work, RDC deployment and other network initiatives, took phone calls on the data concentrators.
JCO: Jack D. together with Gary Pirkola hired me!!

You (Scott) also did the tape routines. Were you the original author? Did you write *MOUNT?

WSG: Yes, but I'm pretty sure there was an earlier version of the tape routines and probably *MOUNT but there was no support for blocking or labels. I think earlier work was probably done by Jay Jonekait (who was at the Bartels award ceremony I think).

Gary, when did you get started?  Was your first work on the file system?

GCP: As Mike indicated, Kip Moore wrote the initial version of the MTS file system (including support for line files) while he was a graduate student at UM in late 1966 or early 1967. I initially worked part time for the computing center as a counselor from  summer/66 to summer/67 while a grad student myself, but I didn't start working for the computing center full time till the summer of 1967. At that time I took over support of the file system from Kip (presumably because he needed to get serious about his PhD). Some time after that I put in sequential files and data cell support and support for really shared files. At some other point Jim Hamilton put in major enhancements to line files to allow lines longer than 255 characters and file sizes bigger than ? ( some initial size that was too limiting). As a historical note, I've always assumed that the initial MTS file system architecture was patterned after the DTS (Dartmouth Time Sharing) file system, primarily because that's where Kip went to undergraduate school, and I believe he was at least familiar with the DTS file system. But I don't know that I ever explicitly asked him. Mike may be able to confirm (or refute) my assumption.
MTA: You're right that Kip came from Dartmouth and was familiar with the DTS file system.  I don't think he copied much of it for MTS, but the idea of line numbers came from there.
ACG: Kip told me at some function that he implemented line files as the architecture of the file system because that was how he thought file systems worked given his experience with Dartmouth Basic.

Bruce [Arden], I assume that you didn't write any code by this time.  Is that correct?
JS: When I was a sophomore (Fall 1967), I went to Bruce's office and asked him if he would supervise an independent study course for me. He agreed, and gave me a simple project. Every time I went into his office, there were notes and drawings on his white (black?) board about MTS. I knew nothing about it at the time, but I was aware that he was involved in the project, at least at the management level.

Did Frank [Westerveldt] or Bernie [Galler] write any code for MTS?

MTA: Neither did any core work.

When did these folks start and what did they work on? (I don't need exact dates, just an indication if they were involved in the first round of MTS development or if their work came later)

    Len Harding? IHC? Elementary function library? FORTRAN compiler?

MTA: Len was already working at the Center when I started.  He heavily involved in the Fortran library, but didn't have a lot to do with the compiler itself.

    Doug Smith?

DWB: Doug Smith wrote *SORT, and the second billing program, and probably installed the COBOL compiler. Doug Smith, who knew COBOL, wrote the next billing program [next after the 7090 billing program].

    John Sanguinetti?

ACG: Supported Fortran compiler, developed shared file system with Gary Pirkola, miscellaneous MTS development, external storage support for the PDP.
JS: I supported Fortran-H, my first introduction to a large-scale program. There wasn't much to do to it, fortunately, since just recompiling it was non-trivial. It was written in Fortran-H, which had a number of non-standard features just for the compiler. Fortran-H (written ca. 1965) had a LOT of optimizations which became standard over the years. It also had a lot of bugs. Larry Flanagan's best quote about it was "Fortran-H is a great compiler, if almost is good enough".
I also took over XPL from Jim Henriksen, and added separately compiled program units (procedures). This was fun, but not terribly significant.
I installed the security gateway that was written at Wayne State (and maybe UBC?), and changed it to use the monitor instruction of the 370. When I got to Amdahl, I was told that the paper I wrote about it for a Sigmetrics workshop gave them serious heartburn, since the monitor instruction was implemented in macro-code on the 580, and was particularly slow.
The only thing I did with the PDP was to add support for the Intel and STC solid state storage devices, around 1981.
JCO: John did a little more than add the Intel and STC support to the PDP.  He did a fair bit of data collection and analysis and even wrote a paper.

John Sanguinetti 2005 U-M Alumni Society Merit Award Computer Science and Engineering:  http://www.cse.umich.edu/eecs/alumni/honoring.html

"Interview with John Sanguinetti by Sean Murphy, May 2009,  http://www.skmurphy.com/blog/2009/05/19/interview-with-john-sanguinetti/
Co-founder and chief technical officer of Forte Design Systems, John Sanguinetti talks about his experience of turning an idea into a business. He was the principal architect of VCS, the Verilog Compiled Simulator, and was a major contributor to the Verilog’s resurgence in the design community. He has 15 publications and one patent. He also developed the Verilog Online Training course. He holds a PhD in computer and communication sciences from the University of Michigan, 1977.

Oral History of John Sanguinetti February 2009: http://archive.computerhistory.org/resources/access/text/Oral_History/102702042.05.01.acc.pdf

From http://www.forteds.com/company/team.asp:
John Sanguinetti
Chief Technology Officer 
Dr. Sanguinetti has been active in computer architecture, performance analysis, and design verification for 20 years. After working for DEC, Amdahl, ELXSI, Ardent, and NeXT computer manufacturers, he founded Chronologic Simulation in 1991 and was President until 1995. He was the principal architect of VCS, the Verilog Compiled Simulator, and was a major contributor to the resurgence in the use of Verilog in the design community. Dr. Sanguinetti served on the Open Verilog International Board of Directors from 1992 to 1995 and was a major contributor to the working group which drafted the specification for the IEEE 1364 Verilog standard. He was a co-founder of CynApps. He has 15 publications and one patent. Dr. Sanguinetti's Ph.D. is in Computer and Communication Sciences from the University of Michigan.

From http://www.aycinena.com/index2/index3/archive/john%20sanguinetti.html
John Sanguinetti – A Profile by Peggy Aycinena

John Sanguinetti is a household name in the EDA industry and this is his profile. Currently, John is serving as the industry's poster child for Multiple Myeloma, a rare blood cancer and the subject of a lot of intense medical research. As such, John will be the guest of honor at an industry fund-raiser in San Jose, California, on September 15, 2004.

However, that's not what makes John's story here compelling. The reason you should rise to the challenge and read this thing in its entirety is because John Sanguinetti is more appropriately the poster child for EDA – a Ph.D. technologist, an entrepreneur, a long-time player in Silicon Valley, and someone who has influenced and been influenced by the trends and characteristics unique to the EDA industry.

If you're involved in the EDA industry, this is more than John Sanguinetti's story. This is your story.

Editor's Note: An edited version of this article appeared in EDA Nation in September 2004.

John W Sanguinetti named ACM Fellow in 2011 for contributions to hardware simulation.

SAN JOSE, CALIF. -- January 10, 2012 -- Dr. John W. Sanguinetti, chief technology officer at Forte Design Systems™, has been named a 2011 Association of Computing Machinery (ACM) Fellow for contributions to hardware simulation.

Dr. Sanguinetti has been active in computer architecture, performance analysis, and design verification for 20 years; he is being recognized by ACM for helping to drive innovations that will sustain competitiveness in the digital age.

After working for DEC, Amdahl, ELXSI, Ardent, and NeXT computer manufacturers, he founded Chronologic Simulation in 1991 and was president until 1995. Dr. Sanguinetti was the principal architect of the Verilog Compiled Simulator (VCS), and was a major contributor to the resurgence in the use of the Verilog hardware design language (HDL) in the design community. He served on the Open Verilog International Board of Directors from 1992 to 1995 and was a major contributor to the working group which drafted the specification for the IEEE 1364 Verilog standard.

After Chronologic, he co-founded CynApps™, Inc., in 1997. It merged with Chronology™ Corporation in 2000 to form Forte Design Systems. Dr. Sanguinetti holds a Ph.D. in Computer and Communication Sciences from the University of Michigan, Ann Arbor.

ACM will formally recognize Dr. Sanguinetti and the other 2011 Fellows at its annual Awards Banquet in June in San Francisco.



    Jim Henriksen?

ACG: Worked with Ken Dejong wrote the link editor in PL/I, maintained the PL/C compiler, maintained GPSS, wrote GPSS/H.
JCO: Jim also hired and managed the Computing Center Counselors (aka consultants).
JS: Jim supported XPL. This was a "compiler-generating system" for writing compilers, written by McKeeman at Stanford. It was a one-pass compiler written in itself and was very slick. The parser generator pre-dated yacc/lex by at least 7 or 8 years (it was published in 1968). This was used for CCS 575, which Jim was one of the first TAs for, if not the first.
I think Jim also supported Simscript. [JO: He did.]

    Jim Hamilton?

GCP: Jim Hamilton put in major enhancements to line files to allow lines longer than 255 characters and file sizes bigger than ? (some initial size that was too limiting). [255 disk pages]
ACG: Worked on the Paging Drum Processor (I don't know who wrote the original paging support, MTA?), various pieces of disk support, e.g., 3330 support, locking support for the shared file system.
JS: Jim wrote the adaptive page-replacement algorithm (I don't remember what it was called, but it had a name). It was table-driven, and you could watch the performance of the system change as you modified the various parameters in the table. I remember standing around watching as Jim and Mike experimented with different settings. At least that's my recollection. This was the subject of his PhD dissertation. I thought this was one of MTS' most significant features, something which Unix systems never had.

    Fred Swartz?

MTA: Fred started in 1964 or 1965.  I don't recall what he did early on, but he was mostly involved in languages and compilers, including MAD/I.

    Marty Raim?

MTA: He started early too, but not as early as Fred.
DWB: Marty Raim wrote *SDS

    Ed Fronczak?

DWB: Ed Fronczak did the Calcomp (plot) routines, Gail Lift took over the plot routines from Ed later
JCO: Ed was involved in the development of MTS BASIC, support for the Votrax based Audio Response Unit, possibly the IBM Audio Response Unit, the $NET command and in particular .copy.

    Clark Lubbers?

ACG: Wrote PDP/11 based data concentrator software, including operating system, supported RAMP after Dave Mills left, wrote a macro language that was used when the Computing Center was contracted to track election results, maintained the PDP/11 cross assembler (he may have also written it).
JCO: Together with Joe Gersch wrote operating system and support for the Ontel Terminal (OP/1). Was involved in the development of MTS BASIC.
BSC: Joe and Clark created the RDCs where Joe did the hardware and Clark the software. Joe and Clark did write the original Ontel code which I later maintained (I added support for the screen support routines in the editor). Clark, I believe, wrote the original MM16 microcode and I worked on maintaining that later.

    Joe Gersch?

JCO: Together with Clark Lubbers wrote operating system and support for the Ontel Terminal (OP/1).
ACG: Supported *wirewrap, supported PDP8- and PDP11-based data concentrator hardware, including Unibus to 360 channel adaptor, RDC (LSI-11) hardware.
WSG: I think Joe's work was entirely related to development/maintenance in "The Shop". I think it may have been entirely hardware and it likely included the RDCs [LSI-11 based Remote Data Concentrators].
BSC: Joe and Clark created the RDCs where Joe did the hardware and Clark the software. Joe and Clark did write the original Ontel code which I later maintained (I added support for the screen support routines in the editor).

    Andy Goodrich?

ACG: Jim Blinn had worked on IG for a year when I joined him. At that point he had done 2D support, when I joined we added 3D support, and after he left I added 3D perspective support.
JCO: Later Andy wrote the Disk Manager (DMGR), and took over $Edit after Viktors Berstis rewrote it. Andy added SNOBOL-like patterns.

    Jim Blinn?

DWB: Jim Blinn did *IG
ACG: Jim had worked on IG for a year when I joined him. At that point he had done 2D support, when I joined we added 3D support, and after he left I added 3D perspective support.
ACG: Maintained LISP ( Andy took over that from him), maintained Watfor, maintained / developed operating system / music software for the PDP/9, wrote the DEC/GT40 operating system used for GT40s in the Architecture School.

    Gail Lift?

DWB: Gail Lift took over the plot routines from Ed [Fronczak] later

    Charles Engle?  Accounting I assume.

MTA: Charlie is the only one to have ever worked in MTS accounting.
DWB: Doug Smith, who knew COBOL, wrote the next billing program [next after the 7090 billing program and the 1st MTS billing program]. So the next [second MTS] billing program was written by Charles Engle in (I think) PL1.

    Ken Dejong?

ACG: With Jim Henriksen wrote the link editor in PL/I.

    Lynn Leader?
    Jim Foley?

    Tad Pinkerton?

MTA: I always associate Tad Pinkerton with the 2250 [in addition to Bill Ripperger(sp?)] also.  I don't recall if he worked on the support for it or just used it.
JCO: Tad says that he was just a 2250 user. Tad wrote the analysis program for the Data Collection (trace tape) Facility. Mike wrote the Data Collection Facility itself.

    Larry Flanigan?

MTA: He [Larry] was here in the 7090 days.  I don't think he did much on MTS.
ACG: Worked on / maintained GASP, a Fortran simulation package

    Bernard Tiffany?

DWB: Bernard Tiffany got *PL1 going on MTS, and wrote *USERDIRECTORY
JCO: Bernard didn't write *USERDIRECTORY or at least not the initial version.

    Howard Young?

ACG: Started as a machine operator and later maintained OS/360 with Chuck Gray
JCO: Wrote the MTS time conversion routines.

On Jul 7, 2011, at 10:42 PM, Brian Cashman wrote:
July 19, 1949 - July 2, 2011
I'm sorry to say that Howard Young passed away on Saturday. See:
    http://wujekcalcaterra.tributes.com/show/Howard-Bruce-Young-94104737

    Jim Knox?
Jim Knox, Director of the University of Michigan Adaptive Technology Computing Site and one of the first consultants to help those with disabilities use computers, died Sunday [July 4, 2010] at his home in Ann Arbor. He was 66. http://obits.mlive.com/obituaries/annarbor/obituary.aspx?n=jim-knox&pid=143971626
The James Edward Knox Center at U-M is named for Jim Knox (1944-2010), the first Director of the University of Michigan Adaptive Technology Computing Site.  http://www.itcs.umich.edu/atcs/jim-knox.php
A Man of Michigan: http://www-personal.umich.edu/~bing/jimknox/Jim_Knox.html

    Who else?

GCP: A couple of other people, not mentioned, that I remember from the early days of MTS:

GCP: Ron Srodawa, who I seem to remember working on or writing the loader initially, but I could be wrong, He definitely was around as a grad student and worked on some part of early MTS. Mike or Don, what did Ron work on?
MTA: Yes, Ron did the loader.  The original version had two unusual characteristics. One (not so unusual actually) was that it fit in one page.  The other was that it was location independent.  You could put it anywhere and call it and it would work.  There was nothing in it that had to be relocated when it was moved to a new location.

GCP: Bill Ripperger(sp?), who I remember worked on. or wrote the 2250 device support (the 2250 was an early/expensive IBM graphics terminal). I think the reason I remember Bill is because he "gave" the 2250 support routines to me when he left, but I never did anything with them, except to try to understand how they worked.
MTA: I always associate Tad Pinkerton with the 2250 also.  I don't recall if he worked on the support for it or just used it.

JCO: Later Don Boettner wrote GOM (Good Old MAD) and with some help from Jim Sterken wrote the MTS Macro Processor.

JCO: Later Jim Sterken wrote $Message. Gavin Eadie and Jim added remote mail support.

JCO: Later Jim Sterken and Steve Burling wrote the $Log command.

BSC: The MM16 hardware, as I recall, was Dennis Tokarski's and possibly Bob Husak's work. I worked with Dennis on that, debugging problems with it, but the hardware design was his. Clark, I believe, wrote the original microcode and I worked on maintaining that later.

Who wrote *EDIT? (I know this didn't exist in Dec. 1967, but did in 1971)

ACG: I actually think that Don Boettner wrote the original version and did the conversion to CLS, but I am not sure.
JCO: Viktors Berstis rewote $Edit and Andy took it over and added SNOBOL-like patterns among other things.

Who wrote *SDS?

DWB: Marty Raim wrote *SDS

Who wrote *SORT? (this seems to have come after Dec. 1967)

MTA: Charlie [Engle], I think.
DWB: Doug Smith wrote *SORT
JCO: Charlie took over support for *SORT later.

Who installed HASP? *BATCH? *HBQ?

GCP: I remember at some point we hired Jim Hansen/Hensen (sp?) (NOT the one who created the muppets) to take over HASP support (from Don who installed it?)
MTA: Don installed HASP.  Jim Hanson took it over later.

-Jeff


From: Donald Boettner
Date: November 15, 2010 11:46:24 PM EST
To: Jeff Ogden
Subject: Re: so who other than Don and Mike were the initial MTS developers?

some additions to this information:

Doug Smith wrote *SORT, and the second billing program, and probably installed the COBOL compiler.
Marty Raim wrote *SDS
Ed Fronczak did the Calcomp (plot) routines
Jim Blinn did *IG
Gail Lift took over the plot routines from Ed later
Charles Engle did a lot of time/date conversion routines and the third billing progam
Bernard Tiffany got *PL1 going on MTS, and wrote *USERDIRECTORY
[JCO: Bernard didn't write *USERDIRECTORY or at least not the initial version]

A note on the billing programs:

The 704/709/7090 machines had no clock.  We had a standalone box (built by GM Research (a copy of the one they used on their own machine)) to keep the time of day and it was hooked into the console printer in such a way that a couple of peculiar sense instructions to the printer would extract the time.   The time for each job was written on the punch output tape with commands to stack the card that was punched in the third stacker (on the 2540 this was the middle stacker of five).   The cards were then read by a Fortran program written by Len Harding which produced output on multipart IDR paper.  I remember coming in late one night the day the billing had been run that month, and finding Sally Brando (our one and only secretary at that time) with an adding machine and the stack of multipart output, and she was adding up the columns of numbers and correcting the totals when they didn't add up because of Fortran's floating point rounding errors.

When it came time to replace this program, we said that billing and COBOL deserve each other, so Doug Smith, who knew COBOL, wrote the next billing program.   It was soon figured out that this was a Big Mistake, as we had to keep maintaining COBOL just for the sake of this one program.   So the next billing program was written by Charles Engle in (I think) PL1.  I'm a little hazy on this area from this point on.

   -don



From: Andy Goodrich
Date: November 21, 2010 7:12:27 PM EST
To: Jeff Ogden
Subject: Re: more memory tests

Some additional observations:

Clark Lubbers
  • Wrote PDP/11 based data concentrator software, including operating system.
  • Supported RAMP after Dave Mills left.
  • Wrote a macro language that was used when the Computing Center was contracted to track election results.
  • Maintained the PDP/11 cross assembler. He may have also written it.
Kip Moore
  • Kip told me at some function that he implemented line files as the architecture of the file system because that was how he thought file systems worked given his experience with Dartmouth Basic.
Ken Dejong
  • With Jim Henriksen wrote the link editor in PL/I.
John Sanguinetti
  • Supported Fortran compiler.
  • Developed shared file system with Gary Pirkola.
  • Miscellaneous MTS development.
  • External storage support for the PDP.
Jim Hamilton
  • Worked on the Paging Drum Processor, I don't know who wrote the original paging support, MTA?
  • Various pieces of disk support, e.g., 3330 support.
  • Locking support for the shared file system.
Jim Blinn
  • Maintained LISP ( I took over that from him.)
  • Maintained Watfor.
  • Maintained / developed operating system / music software for the PDP/9.
  • Wrote the DEC/GT40 operating system used for GT40s in the Architecture School.
Jim Henriksen
  • Worked with Ken Dejong wrote the link editor in PL/I.
  • Maintained the PL/C compiler.
  • Maintained GPSS
  • Wrote GPSS/H
Bernard Tiffany
  • Maintained the PL/I compiler.  "Enable string-range, subscript-range"
Larry Flanigan
  • Worked on / maintained GASP, a Fortran simulation package.
Howard Young
  • Started as a machine operator and later maintained OS/360 with Chuck Gray
Jack Diguiseppe
  • PDP-11 terminal driver development.
  • Some DSR work (vet with Scott).
  • SCP deployment and other network initiatives.
  • Took phone calls on the concentrators.
Joe Gersch
  • Supported *wirewrap.
  • PDP8- and PDP11-based concentrator hardware, including Unibus to 360 channel adaptor.
  • Ontel support with Clark.
  • SCP [RDC?] hardware (vet that with Scott).



From: Scott Gerstenberger
Date: November 22, 2010 9:30:13 AM EST
To: Jeff Ogden
Subject: Re: more memory tests

I don't remember much about what Joe worked on although I think
it was entirely related to development/maintenance in "The Shop".
I think it may have been entirely hardware and it likely included the
RDCs. I don't remember much about the transistion from RDCs to
SCPs (or the DCs to PCPs for that matter). Possible sources of info
on what Joe worked on are Jack DiGiuseppe and Brian Cashman
since I'm quite sure they both worked with Joe (as did Clark Lubbers).
But I think Joe and Brain worked together on various hardware
project probably including the MM16-based serial adapters.

I didn't work on any PDP-11 based systems.

Scott



From: Brian Cashman
Date: December 15, 2010 1:03:41 PM EST
To: Jeff Ogden
Subject: Re: Fwd: more memory tests

Sorry about the late response. Joe left not long after I started so I don't have much recollection of what he worked on there although I do recall that it was mostly hardware related. I believe he and Clark created the RDCs where Joe did the hardware and Clark the software. I think that after Joe left and Merit began to work on the SCPs that Clark decided to end the RDC work and the two efforts then combined.

However, Joe and Clark did write the original Ontel code which I later maintained (I added support for the screen support routines in the editor). Kenobi was a program that ran on the Ontel and was used to debug other Ontel programs (e.g., the OP-1 MCP terminal program). So, Kenobi allowed you to display register values, etc. We burned Kenobi into EEPROM and the Ontel would run it on start up. Then, using Kenobi, you could download the terminal program (whose name, escapes me) and run it or run it out of adjacent EEPROMs.

The MM16 hardware, as I recall, was Dennis Tokarski's and possibly Bob Husak's work. I worked with Dennis on that, debugging problems with it, but the hardware design was his. Clark, I believe, wrote the original microcode and I worked on maintaining that later.

Brings back memories...

     Brian



From: Tad Pinkerton
Date: December 13, 2010 5:35:14 PM EST
To: Jeff Ogden
Subject: Re: how good is your memory?

Hi Jeff,

I'm not at all sure that I ever wrote a line of code that ended up in MTS.  I worked very closely with Mike on what to collect and how to interpret/understand what was collected, but he did the coding.  I ran analyses of data from the system, but I didn't write the code that put the data out on tape to begin with.  I probably wrote the data collection facility, but I'm not sure of that.

I was a heavy user of the 2250, putting a lot of my thesis text into the computer using it, but did not write any of the DSR  I was also a very early user of the line printer (I've forgotten its number) with a Mylar ribbon.  I produced the first thesis at Michigan that way, using the IBM document preparation system, for which I needed a waiver from the Grad School on formatting issues.  So I was a good client of Bill Ripperger's but didn't do any of the code.

Tad



From: John Sanguinetti
Date: January 11, 2011 5:44:28 PM EST
To: jeff.ogden
Subject: stuff for the archive

Jeff,

Andy sent me a link to the Google groups web site, and I was going to add some things, but it looks like it is easier to just send it to you.

Here are some other things various people worked on:

Jim Henriksen
Jim supported XPL. This was a "compiler-generating system" for writing compilers, written by McKeeman at Stanford. It was a one-pass compiler written in itself and was very slick. The parser generator pre-dated yacc/lex by at least 7 or 8 years (it was published in 1968). This was used for CCS 575, which Jim was one of the first TAs for, if not the first.
I think Jim also supported Simscript.

Jim Hamilton
Jim wrote the adaptive page-replacement algorithm (I don't remember what it was called, but it had a name). It was table-driven, and you could watch the performance of the system change as you modified the various parameters in the table. I remember standing around watching as Jim and Mike experimented with different settings. At least that's my recollection. This was the subject of his PhD dissertation. I thought this was one of MTS' most significant features, something which Unix systems never had.

John Sanguinetti
I supported Fortran-H, my first introduction to a large-scale program. There wasn't much to do to it, fortunately, since just recompiling it was non-trivial. It was written in Fortran-H, which had a number of non-standard features just for the compiler. Fortran-H (written ca. 1965) had a LOT of optimizations which became standard over the years. It also had a lot of bugs. Larry Flanagan's best quote about it was "Fortran-H is a great compiler, if almost is good enough".
I also took over XPL from Jim Henriksen, and added separately compiled program units (procedures). This was fun, but not terribly significant.
I installed the security gateway that was written at Wayne State (and maybe UBC?), and changed it to use the monitor instruction of the 370. When I got to Amdahl, I was told that the paper I wrote about it for a Sigmetrics workshop gave them serious heartburn, since the monitor instruction was implemented in macro-code on the 580, and was particularly slow.
The only thing I did with the PDP was to add support for the Intel and STC solid state storage devices, around 1981.

Bruce Arden
When I was a sophomore (Fall 1967), I went to Bruce's office and asked him if he would supervise an independent study course for me. He agreed, and gave me a simple project. Every time I went into his office, there were notes and drawings on his white (black?) board about MTS. I knew nothing about it at the time, but I was aware that he was involved in the project, at least at the management level.

I hope these recollections are right.

John



On Jan 11, 2011, at 5:23 PM, Jeff Ogden wrote:

Thanks John.

Sending things to me is a fine way to add them to the Archive.

Was Jim Hamilton's "adaptive page-replacement algorithm" controlled by something called BJPMOD?  I took that over at some point. It was called every 20 seconds from the System Status Routine (the JOBS program) and set two values, BJT (big job threshold) and NBJ (number of big jobs) that UMMPS used to determine which jobs were "big" and so should get a big time slice and how many big jobs were eligible to run simultaneously. Jobs were either Big (eligible to run with a extended time slice), Ineligible (big jobs that are not able to run because there are too many eligible Big jobs), or Neutral (eligible to run with a regular time slice or perhaps to be made Big or Ineligible depending on the settings of BJT and NBJ). While Big jobs would get an extended time slice, they would also eventually get a long time off the CPU after their time was up and they became ineligible. The whole thing (BJPMOD) was a sort of table driven state machine.

I used XPL in one of the CCS classes I took, probably 575, but Jim wasn't the TA when I took it.

You are right about JimH working on Simscript. 

You did a little more than add the Intel and STC support to the PDP.  You did a fair bit of data collection and analysis and even wrote a paper.

So did the monitor call instruction turn out to be a real performance problem or just a cause for concern?  I remember some performance issues with that first 5860, although I think you also pointed out to us at that time that our benchmarks weren't very realistic in terms of a real workload.

  -Jeff



On Jan 12, 2011, at 2:35 AM, John Sanguinetti wrote:

Jeff,

Thanks for refreshing my memory about bjpmod. Other operating systems had crude mechanisms for swapping jobs in and out of real memory, but this was a very elegant way of managing that process. My recollection was that there were something like 10 states that jobs could be in. It would be interesting to look up Jim's thesis.

The paper I wrote about the solid state paging devices got me introduced to Mac MacDougal, who eventually hired me at Amdahl. Speaking of Amdahl, I went to the Computer History Museum tonight and they have a piece of a 470 on display. It actually looks old.

The monitor call instruction probably took on the order of 100 cycles, instead of 10 for an SVC. I doubt that it was a real problem, but it was definitely not very efficient. The original 580 was almost a factor of 2 slower than had been promised. This was mostly due to long cache miss times, and amdahl's lack of understanding of cache miss rates. In the first 6 months that I was at amdahl, the machine improved by nearly that factor of 2, so it did hit its original performance target. There were something like 100 performance improvements, the largest of which was less than 5%.

John 
Sent from my iPad


Comments