Dr. Culler Bedtime Stories

Ron Stoughton

When I was a toddler my parents made me a play desk with knobs and switches to entertain me and to keep me out of mischief. It is no surprise that 18 years later I would be standing in front of the window in North Hall looking in at the array of computers arranged neatly along the walls of the Computer Center. I didn't know what they were, or anything about them, but they had lots of lights and buttons, and I wanted in.

In 1965, during my fourth year of college, I demonstrated a total lack of wisdom by dropping varsity football and getting married. However, I made up for this by having the good sense to register for Numerical Analysis 104 taught by Glen Culler. Part of the curriculum required writing programs to solve numerical problems on the IBM 1620. I can vividly remember my first FORTRAN program that solved simultaneous linear equations using Smart's method (it is strange that I can remember this because I can't remember what I did yesterday). I didn't know it at the time, but this was the beginning of my career.

I don't know if I would have had the courage to ask Glen for a job, but it was unnecessary. He apparently took notice of me anchored to the console of the 1620 night and day, and asked if I was interested in a job. Clearly, I was. Glen's expression of confidence and willingness to take a chance on me is something I will never forget, and for which I will be eternally grateful. Being able to work for Glen, and learn from Glen, was singularly responsible for my success.

I don't remember specifically what my first task was. I believe it had something to do with the drum controller (block 300 in memory for the RW-400 aficionados among us). However, I do remember that I didn't have a clue about how to proceed. I asked Mike McCammon to explain, but I still didn't understand. I thought to myself that this was a hell of way to start my new job, and I certainly was not going to impress anyone. Fortunately, Glen had a better idea a few days later and I didn't have to complete the task.

At the time I had no idea that we were exploring the frontiers of computing. Looking back on it, I find it amazing what Glen was able to do with the RW-400. Imagine, a 1024-word machine that could support 16 users doing complex mathematical analysis with graphics display. He even got a multiprocessor version of this running. If one were to write a Book of Firsts, Glen's contributions to the field of computing would fill many pages.

Working with Glen was exciting, challenging and rewarding. There were even moments of absolute terror. I can remember sitting in my office during my first week on the job and suddenly hearing the sound of a machine-gun coming from the computer room. Shortly after that, the electricity went off and everyone went running down the hall towards the computer room. I thought for certain that Ray Bjorkman or Dennis Grubbs had electrocuted himself while fiddling with a power supply. It turned out that the sound was the big 1-inch Ampex tape drive running off the end of a reel of tape. I don't know what caused the power outage, but whenever this would happen, there would be a race to turn off three power supplies in every rack of equipment before power was restored. If this wasn't done, it would sometimes take days to get the RW-400 back on its feet.

I can think of many stories to tell about the good times we all had working with Glen. Here are a few. They are not in any particular sequence, and are just in the order I recall them.

Glen would often be invited to conferences to speak about the work he was doing at UCSB, and on occasion would give a live demonstration. Telecommunications was not what it is today, and Murphy's Law would often prevail. The Fall (maybe Spring) Joint Computer Conference in New York was one such occasion. Mike McCammon accompanied Glen to the conference, and it was fortuitous that he did. When it was time to present his "live" demonstration, they discovered they only had one-way communications. Glen could receive display output, but could not send keyboard input. Undaunted, Glen proceeded as if nothing was wrong. Glen described for the audience what he was doing as he pushed keys on the keyboard: "ID MULT 2 MULT 3.1415 COS DISPLAY." Mike was in the wings on the telephone with Curt Lemon who was in the Teleputer classroom sitting at a Culler-Fried work station. Mike would call out the name of the key pushes, Curt would press the keys, and the display would come out in New York. I don't think anyone in the audience suspected a thing. It reminds me of the movie "Network News" when Albert Brooks says, "I say it here and it comes out there."

Glen approached everything in a big way. At the end of the day we would sometimes go have pizza together at Shakey's on Hollister Avenue. Glen was obviously the center of attraction and would sit in the center of the table. That is also where the pizza was. I noticed when someone would replenish the pizza, it would disappear rather fast. I took particular note one day and discovered that Glen would take two pieces at a time and eat them like a sandwich. I guess you would call this the "Culler fast-start" algorithm.

Glen always saw room for improvement. The original keyboards used individual switches from Microswitch that were expensive and lacked the tactile feel that Glen desired. He would solve this by designing his own keyboard and developing a new switch. One of the problems with the old Microswitch was that you sometimes didn't know whether the key push was actually read by the computer. Glen's new switch would give both an auditory and tactile response as confirmation that the switch had closed. As always, Glen poured himself into this effort. The design required a printed circuit board with non-corrosive contacts and Glen determined that the best material was Rhodium -- good conductivity and non-corrosive. The only problem was that solder would not adhere to it on the printed circuit board. I don't know how this was resolved, but it was.

There is an addendum to this story that I am somewhat reluctant to tell, but I will because it is indicative of the enthusiasm that was part of Glen's approach to problem solving. Anyone who has been consumed by a project will be able to identify with this. So, if I embarrass you Glen, I apologize.

Glen's keyboard not only needed to be functionally perfect, but should be aesthetically pleasing. Ergonomics had not been invented yet, but Glen was in the thick of it. He took particular interest in the aluminum castings that would hold the upper and lower keyboards (and of course, there were lots of colored keys for added sex appeal). The castings needed to be sleek and appropriate for space-age technology. I suspect Glen went to bed one evening thinking about this and awoke early in the morning from a fitful sleep with an idea that just couldn't wait. He needed something to make a template. It needed to be sturdy, but something he could fashion with an Exacto knife. "Yes! Here is a canvas oil painting mounted on hard board which is just perfect!" The only problem was that it was a gift from Curt Lemon and was painted by his wife Joan. "No problem, Joan likes to paint and she can do another one."

I have this image of the painting hanging on the wall, Glen running to the wall and ripping the painting from its frame, and returning to the kitchen table where he proceeded to carve his templates. However, I am sure it was not like this at all. I suspect it was unframed and not hung. We all know that art is a personal matter, and one man's art is not necessarily another man's pleasure. In fact, I have several paintings down in the basement that my wife brought to our marriage and which might serve a similar purpose.

Thinking of how Glen would pour himself into a project reminds me of a story he once told me. Upon returning to UCSB after his sabbatical at TRW, the RW-400 remained in Los Angeles. After teaching his classes, he would jump in his car and drive to TRW to continue work on the Culler-Fried system, and then would make the long journey home. While driving he would ponder the days activities. This was a good time to think, and his mind was a caldron of new ideas. Once, after driving all the way back to Santa Barbara and having no recollection of the trip, Glen decided to press to have the RW-400 moved to UCSB. I think he had visions of driving off US 101 into the Pacific Ocean while visions of Fourier transforms danced in his head.

Is bigger better? Yes. Is faster more fun? Certainly. Glen bought himself a metallic blue ski boat with a powerful outboard motor (probably a Mercury, but I don't remember). Glen and Suzie invited people from the lab to accompany them to Lake Nacimiento for a little fun and water skiing. At full throttle (which is where Glen likes to be), his boat could pull a skier much faster than any sensible person would like to go. However, there were other boats there with big Chevy supercharged engines and chrome-plated headers that belched fire and broke your ear drums. Glen wanted one of those, real bad!.

Glen was a "can do" person. No obstacle would stand in his way. I didn't see this happen, but I saw the results and know it occurred. Apparently a few people were sitting around the fire drinking beer, and someone had a bottle with a twist-off cap that was being stubborn. "Here, gimme that" Glen purportedly said with an impatient look on his face. He proceeded to twist the neck of the bottle off and fillet the flesh of his finger right down to the bone.

Glen was a quick learner, even at golf. Ed Faeh, John White and I would often go play nine holes of golf at the University Golf Course (my ration of lost golf balls was one per hole). Glen came with us one afternoon and I remember it vividly. I don't know where he got his golf clubs, but his driver looked like it was carved from the limb of an oak tree during the Civil War. Never mind. By the third hole he had learned that the object of the game was to hit the ball as far as you can. In other words, he drove like a gorilla and putt like a gorilla.

It was a pleasure to watch Glen work with the RW-400. He was in his element. He knew every line of code and had written most of it. The IPL procedure was to stand at the RW-400 control panel and enter a series of machine instructions to read a block from drum. He could enter them about as fast as the computer could execute them. Having only 1024 words of processor memory was a little constraining. You can all imagine his response when he discovered an additional 1024 words of memory hiding in the Central Exchange, which he realized he could use as scratch-pad memory.

The RW-400 was eventually replaced by an IBM 360/50, and after that a 360/75. The 360 had all of the modern conveniences that the RW-400 didn't -- high-speed memory (512K bytes I think), bulk core storage (1.5M bytes), high-speed disk, assemblers, compilers, etc. However, it also had an IBM operating system call DOS. If Glen needed a special feature on the RW-400, he would just program it. There were no artificial barriers imposed by an operating system. As we were developing the Culler-Fried system on the 360/50, I could at times sense some frustration in Glen. It was hard for him to accept that we couldn't do something because DOS didn't permit it. No problem, change DOS!

DOS was designed around the typical thinking of IBM at the time -- read a job from the card reader, commit all of the resources to churn on it for a while, burp out the results on a line printer, and then go read the next job in the card reader (but not quite yet -- we need to wait for the print job to complete).

The first thing we did was modify DOS to give a program residing in bulk store a time-slice every so often, and then return to traditional batch processing. My recollection is that the time-slice occurred every 40 milliseconds. This was the first (and probably only) time-sharing system built on top of DOS. Curt Lemon then wrote a card and print spooler to speed up batch processing -- another first for DOS. An amusing side bar to this is that the DOS time-slice was not in harmony with the clutch cycle on the card reader. It sounded like a 50-caliber machine gun. IBM threatened to void the warranty if we persisted to do this. We told them to go pound sand, which they did. Curt eventually learned the secrets of channel I/O and chained CCW's, and the card reader began to purr.

IBM eventually woke up and said, "Oh! You want to time share and run more than one job at a time? Well, what you really need is MFT."

Roland Bryan, who was now working at BBN, designed us a 16-station graphics controller that connected to the IBM byte multiplexer channel. When it arrived, it was in a blue cabinet (what else) designed to match our new IBM computer. However, IBM decided to break with tradition and offer the 360 in something other than IBM blue. In our case it was IBM sort-of-red-or-maybe-salmon, which clashed with the BBN box -- that was its name, the "BBN box."

Well, we just couldn't stand for this. You don't spend millions of dollars on a sort-of-red-or-maybe-salmon computer and connect it to a blue BBN box. Over the weekend when Roland was not around, the blue box changed color like a chameleon. A brush and bucket of paint from the local hardware store did the trick (after all, you wouldn't want to use spray paint and coat all of the flip-flops and registers inside). I don't recall for sure who did this. I think it may have been Chuck Loepkey, but Glen's finger prints were all over it. Roland was not happy, and threatened not to fix the BBN box if it failed (but he did). However, we had to be careful after that because if we blamed the BBN box, Roland could counter by blaming the red paint.

The RW-400 computer module had a 26-bit word consisting of a 6-bit opcode and two 10-bit address fields. Each word was divided into two 13-bit half-words with its own parity bit. It had seven flip-flop registers including the instruction (X) register. Most computation was carried out using the accumulator (A) and extension (B) registers. The fastest instruction was a "LOAD A" that zipped to completion in 33 microseconds. The slowest instruction was a "REPLACE SQUARE ROOT" (yes, it had a square root instruction) that took 274 microseconds. Data was represented as a signed binary fraction with no exponent (that is, the most significant bit after the sign bit represented 1/2). A FLOAT instruction was used to determine exponents. In vector arithmetic, Glen would normalize the data by floating each word to determine a common exponent. Therefore, only one exponent needed to be stored for the entire vector. (I think this is the way it worked, but I am not sure. Glen and Mike McCammon wrote most of the computational code, and I never quite understood how it worked. I was more interested in executive aspects, and was content to leave the hard stuff to the mathematical wizards.)

When the 360/50 arrived, we all participated in a week of training classes. I am going to take this opportunity to admit that I was totally wrapped around the axle with this base/index addressing mumble-jumble. Once again I had visions of being a total bone-head, and wondered how I was going to cope with all of these brilliant people around me. Somehow I stumbled through.

Glen would often need to be out of town to attend conferences or deal with government funding issues. Before Chuck Loepkey came along, he would leave Kathy (last name?) in charge of administrative issues. The RW-400 came with a device known as a Display Analysis Console, or DAC. Think of it as a two-ton terminal. Apparently in the early days it was used as the Culler-Fried terminal until Tektronix storage displays were developed. During my time it just sat in the center of the computer room and kept the floor tiles from coming loose. While Glen was on an extended trip, someone asked why we needed to keep this thing. Kathy checked with the contracting office and they authorized us to "surplus" it. To them, that meant selling it to the highest bidder. To us, that meant bringing in our hack saws and bolt cutters and scavenging anything of value.

Dennis got some nice drawer slides and the ground buss made nifty battery cables. I whined so much about Dennis' drawer slides that he finally gave me a pair. I had to give him something in return, but I don't remember what it was. The drawer slides were so nice that I wanted to save them for "that special project." Every candidate project that came along just was not quite good enough, and as a result, I still have them in the basement. Some of the less-nice drawer slides ended up in my work bench supporting a power-tool drawer, and are in use today. Every time I go out to the garage to get a power tool I think of the telephone call Kathy received informing us they found a buyer for the DAC. I made it a point to not be around the day Glen came back and Kathy informed him of some of the administrative decisions made in his absence.

It surprises me how much I can remember from these wonderful days. When I am old and senile (not far away since I am already old and partially senile), and I am sitting in my chair uttering incomprehensible words, it will probably be CRL-ese such as:

"spring-gate" "item-structure" "opcon" "button gatherer" "ice-box" "bug-watcher" "pred-operator" "user-list"

"pred-operator" reminds me that there are many things Glen has done that directly benefited me, and for which I am very grateful. Glen asked me to develop the PRED operator for user-list programs. I really got interested in this project, and began working at the computer center around the clock, as was my habit. However, when he learned from one of my math professors that I was not doing well in his class, he forbid me to come to work until finals were over, but continued to pay me nonetheless. My reaction to this was to secret myself in the bedroom and continue to work on the PRED operator under the pretense I was studying. When my wife discovered this, she called Glen, who promptly informed me that I would no longer have a job if I didn't pass all of my classes. So you see, Glen not only served as a mentor and role model, but would also dispense some sound, fatherly advice.

Glen is a brilliant man who has not been sufficiently recognized for his accomplishments. Most of us are lucky to have had one or two truly good ideas during our lifetime. Glen had an abundance of ideas, many of which were probably never acted upon because he just didn't have enough time.

I can remember standing in the job-submission area of the computer center talking with co-workers about how amazing Glen Culler was. Someone suggested that the best use of Glen was to lock him in a room for a week with a tape recorder, record all of his ideas, pay him a million dollars and then send him on his way. You would then hire all of the programmers and engineers you could afford and spend the next several years implementing his ideas. That was an interesting concept, but in retrospect, I would have never wanted to miss the opportunity to work along side of Glen, watching him attack tough problems, experiencing his enthusiasm, and sharing in the pride of his accomplishments.

I am looking forward to seeing Glen and everyone else at the big event in September. It is long over due.

P.S. One last story comes to mind. Last year BBN sponsored an event to honor the ARPAnet pioneers on the 25th anniversary of the ARPA Network. Glen was one of the honored guests. On the day of the event, Glen had an accident and broke is arm, but that did not stop him from participating in the evening's festivities. This is vintage Glen.


Previous
Tribute
Tribute
Contents
Next
Tribute
Back to
Glen Culler Honorary Lecture
Home Page
Back to
College of Engineering

kk October 2, 1995