An Interview with Dieter Fuss

Dieter Fuss

DF = Dieter Fuss
GAM = George Michael
February, 1994

GAM: I'm here recording the memories of Dieter Fuss. So, Dieter, why don't you start by telling us when you came to the Lab and so forth, and go on from there.

DF: Well, I came to the Lab, I think it was in 1966. I was hired by Joe Brady to do applications programming mostly for M Division at that time. When I got hired I didn't know programming. I got hired more or less because I had a physics background. And in those days I guess you didn't need to be a programmer to begin with�you learned it as you went along. So I started going to classes full time, and then on up, serving M Division for, I would say, six to seven years.

I remember one time getting a job from a professor at Berkeley named Ned Birdsall, who said he needed a programmer for two weeks. I wound up working for him for five years.

GAM: I remember in that interval you were working on a German-to-English interactive translator. I thought that was a really great thing.

DF: Well, I got to meet Hans Bruijnes along the way, and he'd heard I was dabbling in something. I was going to school, and I needed projects of sorts that were somewhat research-oriented, and so I did this mechanical translation thing, going from Russian to English and from German to English. Actually, I thought it was useful for, let's say, scientific translations where you didn't have to get all of the precise meanings involved but you wanted to understand enough to get some idea. If somebody really wanted an exact translation they could then look at the machine translation and determine if it was worth the time to have it translated by a human being. But it fell by the wayside because other activities took over. So that was an interesting project. It was a standard Hans thing�he had a lot of these little skunk works going on the side. That was my first introduction to Hans, but certainly not the last.

I spent a lot of time working with Ned Birdsall, who was one of the first people perhaps in the world to try to simulate (model) plasma physics phenomena. And eventually that became a big-time business. But Birdsall had the codes, and I was his programmer, to start that out. That was an exciting time. I was doing really useful work then, and had a lot of fun doing some of the more "play" things.

You know, at that time Computation was headed by Sid Fernbach, and Richard von Holdt. That was it as far as I remember about the administration in the 1966-67 time frame. Von Holdt was kind of my mentor. He taught me what it means to say "garbage in, garbage out." He was a believer in that. We're a little bit different now. Those were very interesting times to begin with; I enjoyed them a lot.

GAM: So you spent most of your time in the applications programming area?

DF: Yes.

GAM: And you did most of this stuff in FORTRAN?

DF: Yes, almost all of it in FORTRAN.

GAM: Almost all, okay. What was the first machine you worked on? It must have been the 6600 or something.

DF: Oh, actually I started working on the LARC, which was quite an interesting machine. I don't remember a lot about the LARC, but it was a base 10. I mean it was not a binary machine, and you could certainly hear what your code was doing to help you to figure out what it was doing. You'd just listen in. [1] Yes, that was an exciting machine. Eventually, then, I worked on 6600 and then a little bit of stuff with Bob Cralle on the 3600. It was an interesting side project as well. We did stereoscopic slides of Joe Brady's orbiting planets and his prediction about where Planet X was.

GAM: Where Planet X was�I remember.

DF: So, the 6600 was, I guess, my first introduction to CDC equipment, and from there on in I spent time on the 6600, 7600, Cray 1, Cray 2, and Cray X-MP. I've programmed on all of those machines.

GAM: A propos of that movie that you guys made, do have any idea what happened to it?

DF: Well, what we did was make stereoscopic slides.

GAM: Ah, slides.

DF: And I wouldn't be surprised if I still have some of those.

Dieter Fuss
GAM: Well, it would be interesting if you dredged them out. They're an important part of Computation history. We'd like to convince you to put them into the archive.

DF: Well, let me address that particular problem, not necessarily for history. When I was relieved of my duties as Deputy Associate Director, an awful lot of stuff that I had in the office had to be moved out rather quickly. And so I have some of the stuff in my garage, but I'm not too sure that I've kept all of those things that might be relevant to the historical end of things.

GAM: Well, we're interested all that sort of stuff, you know�manuals, memos, code listings, and things like that, and certainly film or slides. John Blunden and I have been cataloging all the old films we can find.

DF: I can make a couple of comments with respect to films�I mean, first of all, there's one stereoscopic film that Jim Foote and I made of the�

GAM: I've got that one�TIBRO. [2]

DF: Yes, right. And then there are lots of films that I made with Ned Birdsall with respect to looking at simulations of particles moving around and seeing how they form clusters. And those kinds of films were actually kind of standard in the days when I made them. So there's some pretty interesting stuff like that floating around. I think that the stereoscopic ones probably are fairly unique for that time period.

GAM: I think so, yes.

DF: Visualization was not the word in those days, but we did a lot of it, indeed.

GAM: Well, we had a lot of trouble with getting time enough to do all that stuff, and the PDP-1 never did pan out as the place where we'd do that kind of work. It turns out when you're on a machine, that's where you want to do the work and really produce it. It's an interesting piece of sociology. So, when did you drift away, so to say, from programming and tangle with other aspects of things like administration or whatever you want to call it?

DF: Well, Joe Brady at one point made me what you might call a Team Leader or Group Leader of the M Division programmers. By that time, the Computation Department had different application divisions, and that eventually led me more and more into that aspect of things. At one point Pat Gray was Division Leader of a whole bunch of application divisions, and I was his Deputy.

GAM: I see. I remember that, yes.

DF: I began to spend more and more time with the administration end of things. About 1972, I believe, there was a proposal to put together a computing facility for all of what at that time was called control thermonuclear research computational activities. And Sid Fernbach had Hans, Tad Kishi, me�and I can't remember whether there was somebody else on the engineering end�put together a proposal to try to get that facility to be built here at the Lab. And, as you know, it eventually did come here, but that took an awful lot of work and coordinating�going out to the various sites that were doing the big experiments, and talking to them about what they needed and how they saw this computation facility.

GAM: A minor anecdote: When I was in Washington many, many years ago, I forgot who told me this, but they said, "That adventure,"�the business of getting MFE here�" showed Sid to be more than just a skillful planner. He got to be a real arm-twister in Washington over that."

DF: Well, it was a difficult thing all around. You talk about anecdotes�some of the sites wanted to put a PDP-11 here at Livermore, and have 7600s at each one of the five major sites�which were the Princeton Plasma Physics Laboratory, Los Alamos, General Atomics, Oak Ridge, and Livermore. There was a tremendous amount of pressure that had to be put to bear to make this a central facility where not everybody was winding up with their own computer. And I think Sid had a tremendous amount of influence to make that happen, yes.

In those days, naturally, I again spent a lot of time with Hans, who I consider a real brilliant person who's done an awful lot of different things, many of which were ahead of his time. And I spent some time with Tad Kishi, who's an interesting person. So it was a real exciting time for me. And it actually turned out to be a plan that finally came to Livermore.

The original idea was to hire applications programmers for all of these magnetic fusion energy physicists, and that's probably the reason I was chosen to be the software manager of that center. It turned out we never had very many application programmers. We wound up developing systems, and networks, and the whole environment�the archival storage environment�and all of our manpower went into a duplicate of the Livermore Computer Center (LCC).

GAM: Well, it was said then, and I'd be interested in your comments on it, that one of the reasons the thing was successful was because you were able to leverage off the work that had been done in LCC vis-�-vis the time-sharing center. And what you had to add to it was the communications, and that was essentially unprecedented at the time we were doing it. Is that sort of correct?

DF: That's absolutely correct. I mean, we were not reinventing the wheel here, we were just picking up from where the Livermore Computer Center already was with respect to operating systems and archival storage, and things like that, and maybe here and there adding an improvement because we had an opportunity to do something new based on the experience that was already there. The communication end of things obviously was a long-haul network as opposed to a local area network, so there were some complications there that were fairly new. I think after a while it turned out to be pretty successful. I actually believe it was the forerunner of the NSF Centers, and a whole bunch of other centers like it.

GAM: Oh yes, and I think that it served the researchers in the magnetic fusion community quite well.

DF: I think it did, too. I never heard anybody complain about the fact that they couldn't compute here, although the network was somewhat limited.

DF: Well the networks today are now orders of magnitude faster than they were in those days. We have to remember that we used to work on 10-character-per-second terminals, and stuff like that.

Dieter Fuss
GAM: Yes�amazing.

DF: It probably isn't obvious to everybody that the first Cray 1 here at the Lab went to NERSC, not to the LCC.

GAM: Right.

DF: Developing software for the Cray 1 was done again, taking advantage of what already existed now in both centers, by moving that operating system to a 64-bit machine and making it run right, and making that first Cray 1 successful here at the Lab. I mean, it wasn't the first Cray 1, but it was the first Cray 1 at Livermore. And going through that transition where almost all of the users said, "Let's get a couple more 7600s." That was again one of Hans' foresights, and he had the right plan.

GAM: Well, I think that what you might call the monopoly that Livermore had on getting serial one of this, that, or another thing, was broken by virtue of the Star, call it "problem." The two-Star problem, I guess is what I'd call it.

DF: Yes.

GAM: So, we have not recovered from that in some respects yet.

DF: Yes, I suppose. I mean I wasn't involved with the Stars whatsoever, so I can't really speak from any authority. My opinion is that the Lab learned a lot more about vectorization in a hurry than any other place, and when I remember the number of 7600s that showed up when the Stars weren't working right, I actually think that the problem as it's made out to be was much more political than technical. And I think the technical things that came out of it helped us get to points where a whole bunch of other organizations even today haven't reached with respect to vectorization and understanding of some of those things.

GAM: Yes. Jim LeBlanc was quoted as saying, "Yes, we spent seven years trying to get these codes to work right on the Star, and I think that was too much." That's what he said. And I think that attitude percolated or permeated all over the Lab's weapon divisions.

DF: That's true. But even today I think most of what you call massive parallelism is vectorization on different kinds of architectures. And I think what we learned out of that�not that I'm saying that the Stars were the right thing�what we were forced to learn out of that, what we had to confront and come up with some solution for, I think is still paying dividends.

GAM: I think you're right. As I said, the problem was considered two Stars�what you learned, you know, could have been learned adequately on one. But we bailed Los Alamos out in some respects by taking the Star that they had rejected. That saved CDC in some sense, I think.

DF: For some time anyway.

GAM: Yes. Yes, that's a good way to put it.

DF: It's just kind of an interesting coincidence that John Ranelletti and I used to be even lab partners at USC together. We were not just in the same courses, we actually worked together on experiments. And then when I applied for a job here at the Laboratory, I ran into John in the cafeteria, and that's probably one of the reasons I wound up actually getting a job here, and in the end, working with John for Bob Borchers as Deputy Associate Director. It's sort of a strange but long coincidence, which I really enjoyed.

GAM: Well, it's a nice thing. So, you had a good time with Sid. You liked him, and you like Hans.

DF: Hans and I were�we enjoyed arguing with each other.

GAM: Well, I understand.

DF: But one of the things that I want you to get from Hans is he had ideas of transmitting pictures long before TMDS (television monitor display system), long before it was common. You know, these bit maps on Tektronix terminals and things like that that he was sending out from the G Machine (the unclassified CDC 6600)�I mean, he wasn't doing the work, but he was the one who was pushing these new projects all the time. There's a whole bunch of things like that that Hans did that now are fairly common, but the original experiments and the ideas were in his head long before other people were doing it.

GAM: Well, back in the early '60s, I worked rather closely with him, you know. He started as a hand calculator, and then he was in charge of the hand calculators, and they were doing a digit-by-digit comparison on the big code we had on the Univac. I worked closely with him then, and I had a lot of fun. All along, he had great ideas. Some of them were not implemented very well, but I don't think you can blame him for that.

DF: No, and also not all of his ideas were all that great, but, among other things, he had some really good foresight.

GAM: Yes. Well, I guess we'll say that's enough for now. Then, we'll get this typed and you'll get a chance to edit it.

[1] The LARC, among other computers from Remington Rand and CDC, had as standard equipment a radio-receiver-like device attached to the computer so it could detect the electromagnetic signals being generated at various parts of the machine circuitry. These signals were fed to an audio amplifier and speaker. Because this sound had a definite connection with what the machine was doing, programmers quickly learned to correlate the sounds with program flow. Much debugging was done in this way, and many special programs were written to play music, thereby making the computer the world's best and most expensive juke box.

[2] TIBRO was a code that calculated orbits of particles in magnetic fields. ("TIBRO" is "orbit" spelled backwards.)