An Interview with Glen Culler

Glen Culler


GC = Glen Culler
GAM = George Michael



GAM: Today is 12/7/97 and we are interviewing Dr. Glen Culler, one of the first mathematicians to arrive at Livermore. Glen, please start, and tell us when and how you got to the Lab.

GC: I started at LBL in 1951, and later Sid Fernbach talked me into coming to Livermore, in late 1952 or early 1953. Tom Wilder hired me and, after I got my clearance, told me in some detail what was being worked on. One of the first things I did was take over the group of people doing hand calculations. Some of the people in that group were Kaye Purdom, Leona Schloss, who later married Sid, Edna Stevens, Joe Wirsching, and Bob Pexton. Later, I hired Hans Bruijnes but, now, I don't recall any other names. As a group, we validated many of the computer calculations, giving every one confidence that the programs were executing correctly.

Another thing that we started was our informal mathematics discussion group.

GAM: I remember that. I gave a talk about Elliptic Integrals, and I learned a tremendous amount of math from Bob Pexton and Glen.

GC: We looked very carefully how various algorithms would perform on those new things called Digital Computers. It was a shock for some in the group to learn that (A*C)/B is not necessarily the same as (A/B)*C. But the joint discovery of these things was important and educational.

There really isn't much more to tell. I wanted to go back to school and get my Ph.D. I talked this over with Sid, who agreed with me. Somewhere between mid-1954 and 1956, I left the Lab. I got the degree and went to UC Santa Barbara and other places. Somewhere along the way, I met Prof. Burton Fried from UCLA, and we designed a computing laboratory around a storage display from Tecktronics. As far as I know, that facility, which allowed symbolic Mathematics, was the first attempt in this country to do such stuff. Later, Macsyma and, still later, Mathematica became available as very effective math tools. I'll assume we can stop here; talking about my later work on ARPAnet and such things as the Culler-Harris computer system are outside the window of your survey.

GAM: I suppose you're right, even though those are fascinating stories. So, we'll stop here. I'd like to thank you and Susie for taking time to remember things from when we started.



Editor's Note:

This interview is shorter than most in this collection. Glen Culler suffered a severe stroke that, among other things, left him with a speech impairment. His wonderful wife Susie, since deceased, assisted in interpreting Glen's remarks.

Glen was awarded the 1998 President's Science Medal. He was also the recipient of the 1998 Seymour Cray Computer Engineering Award in recognition for his pioneering work. Professor David Culler, of the University of California in Berkeley, accepted the Cray Medal on behalf of his father during the 1998 Supercomputer Conference in Portland, Oregon. Professor Culler's remarks in accepting the award, reproduced below, described Glen's career in more detail.

Professor David Culler's remarks:

Looking back from our perspective as users of modern high-performance personal computers with graphical user interfaces and multimedia capabilities, Glen J. Culler's work stands out clearly as a vision of the future, twenty five years ahead of its time and many years ahead of the technology. His work helped put UCSB in the forefront of what would become the field of Computer Science and his influence gave rise to much of the computing industry in the Santa Barbara area.

Culler joined the Mathematics faculty at UCSB in 1959. Trained as a pure mathematician at Berkeley and UCLA, he nevertheless brought with him a lot of working experience with the computational aspects of applied mathematics and physics. While many engineers, scientists and mathematicians were attracted to computers for their computational power; Culler wanted the computer to be a problem-solving tool in an real-time loop of man-machine interaction.

For a scientist in 1965, it was impressive and satisfying to be able to punch a program on a deck of cards, run these through the computer and, in a matter of hours, receive the results of a very complicated computation that would have required a room full of people operating mechanical calculators and recording their results by hand. Glen Culler, however, had a radically different vision of computing. He imagined that people would be sitting at a computer and solving problems in real time, by interacting directly with the computer. He saw the computer as an extension of the human intuition and intellect.

Indeed, the way in which we see and use computers today owes much to Glen Culler's persistent efforts to realize his vision and to encourage his colleagues to make choices that would keep the future open to a much broader use of computers. This involved more than building systems. The vision had to be kept clearly in mind when dealing with every technical detail of the emerging technology. A small illustrative example involved the determination of the ASCII character code standard. In the early 1960s, Culler insisted that it was essential to include both lower and upper case letters in the character set. Since this view was in opposition to most of the computer industry of that time, he traveled to Washington to argue his case before ANSI; one can imagine the effect on word-processing if he had not been successful.

In 1961 Culler went on leave from his appointment at UCSB to become Assistant Director of the Computer Research Laboratory at Ramo-Wooldridge (now TRW). In that role, and in collaboration with Professor Burton Fried of UCLA, he designed and directed the implementation of the first interactive, mathematically based, on-line graphic system. The Culler-Fried System was unique in that the user interacted with the system through graphics and in the summer of 1962 a group of physicists, including two Nobel Laureates, Feynman and Schrieffer, were invited to bring their own physics problems to Ramo-Woolridge in order to use this new computing concept in their work. Now, thirty-five years later, when the world's software giants have come to accept the graphical user interface as the most natural way to communicate with a computer, and when many children are involved in interactive computing before they go to school, it is difficult to imagine what a bold and innovative idea this once was.

When Culler returned to UCSB, as Director of the UCSB Computer Center and as Professor in the College of Engineering, he extended his revolutionary view of he role of computers to include their use in the classroom. He persuaded Ramo-Wooldridge to donate RW-400 computer to USCB so that he could continue the development of the Culler-Fried system. His plan was to set up the first computer classroom in which students sat at two workstations, each networked to the RW-400, which managed the input and output for all the stations. The technology for the display did not exist at the time and the size and cost of magnetic core memory made it impractical to use video displays. (In fact, the RW-400 itself had a grand total of 1024 26-bit words of magnetic core memory). Culler's group, therefore, created display screens by adapting Tektronix oscilloscopes.

To manage the multiple workstations they also had to invent one of the first multi-programming multi-tasking operating systems. Moreover, to draw the characters and graphs efficiently, Culler invented a method of vector graphics that is still used in high-performance graphics systems today. Yet another breakthrough and unique feature of the system was the concept and development of function keys, allowing the user to specify high-level operations, such as integration and exponents, and store them under a single key. To accommodate these tasks, Culler and his group had to build the keyboards for the stations. When completed, the system supported 16 workstations with response times comparable to those of modern computers � and with less than 4KB of processor memory. The UCSB classroom was used by a number of UCSB professors to teach subjects such as complex variables, network theory, controls, and hydrodynamics. These courses were well ahead of their time.

The UCSB On-Line System concepts are ubiquitous today. The idea of having separate keys for the elementary functions, of being able to interactively graph functions, and the ability to write programs by recording a sequence of keystrokes, were all present in the On-Line System in almost exactly the form that they appear in a modern graphing calculator. (Of course a modern TI-82 does have ten times the memory of the RW-400 and only supports one user). Finally, after thirty years, the idea of using computers in education has taken off. Specifically, Mathematics departments across the country, including at UCSB, are working to incorporate technology, in the form of graphing calculators, into the calculus curriculum. The modern equivalent of the On-Line System is being used to give students intuition about mathematical functions in very much the way that Culler first articulated in the mid-1960s. When the RW-400 was replaced with an IBM mainframe, the On-Line System was expanded to include matrix operations, making it a precursor of today's MATLAB system. It was also expanded physically throughout the UCSB campus.

It was Culler's work, to develop one of the world's first networked computer systems, and his group's expertise in networking, that led to the selection of UCSB as one of the four original sites of the ARPANET. The other sites, UCLA, Utah and SRI, were much larger, better known computer science programs. Nevertheless UCSB was, in fact, the only one of the original sites to implement the ARPANET interim protocol and was the first ready with the final version. The ARPANET ultimately evolved into the INTERNET which, as we all know, is poised to cause dramatic and revolutionary changes in the way we live. Several companies in Santa Barbara, including ACC Systems, EDI, CMC (now Rockwell Network Systems), and ComDesign, formed as a result of this work at UCSB.

By the late 1960's Culler realized that digital technology had advanced to the point where it had become feasible to use computers to synthesize and analyze sound. Together, with his group of engineers at UCSB, he designed analog-to-digital converters for his on-line systems and began to experiment with digitized sound. Some of the world's first digital recordings were made at UCSB and at Culler-Harrison Inc., a Goleta-based company Culler formed in 1971. Today, of course, digital recording of music has completely replaced analog methods and revolutionized the recording industry. The notions of interactive computing were extended in the development of a system for on on-line digital signal processing. Culler did a great deal of work in the algorithmic as well as the mathematical aspects of signal-processing, including work in what would now be called the theory of wavelets. Working with the MIT Lincoln laboratory and the Information Sciences Institute, he developed software algorithms and fast, compact computers to demonstrate real-time encoding and decoding of speech. Culler was again on the leading edge, in this case with the digital transmission, processing, and analysis of speech, which is fundamental to the modern telecommunications industry, voice-mail, computer games, and more.

As part of his research on speech recognition, Culler used a physical model of the vocal tract to build an extremely compact encoding of the actual production of speech, rather than simply digitizing the electrical signal that comes out of a microphone. This work spawned several companies within the Santa Barbara area, including Digital Sound. Upon returning to the University, this modeling formed the basis of an invention called the Chromophone, intended to help hearing-impaired individuals improve the quality of their phonetics by being able to visualize on a TV screen the sounds produced. This too has been developed into products by IBM and others.

Culler-Harrison evolved into Culler Scientific systems, a company which initially specialized in the design of hardware for signal-processing applications. These designs pioneered what is now called the Very Long Instruction Word (VLIW) architecture. The key idea is that each machine instruction describes several independent operations. This allows many different hardware units to work simultaneously to obtain higher performance. Today VLIW is widely considered to be the breakthrough beyond RISC (Reduced Instruction Set Computers). It is scheduled to debut in Hewlett-Packard's PA8000 microprocessor next year, as well as in upcoming follow-ons to the Intel Pentium, IBM/Motorola Power PC, and others. Culler's work in this area started in the early 70's and included construction of a series of computers designed to support operations on large arrays of data. One of these designs was licensed to Floating-Point Systems, which produced the FPS AP120B in 1976. This was considered by many to be the "poor man's CRAY," referring to the Cray-1 introduced in the same year. The AP120B delivered over 3 MFLOPS for $50,000, compared to the Cray's 20 MFLOPS for nearly ten million. It took ten years and the RISC microprocessor revolution before the price/performance point of the AP120B was equaled by technical workstations. Glen continued to innovate with VLIW design at Culler Scientific, including development of an advanced microprocessor for digital signal processing with Motorola and general-purpose "personal supercomputers." In the 1980's The VLIW approach was adopted as well by Multiflow and Cydrome, producing fierce competition for the "minisuper" market, which eventually yielded to advancing workstation microprocessor technology. In the ten years since, the compiler technology developed around VLIW has matured and this architectural approach is about to break through to the mainstream of computing.

Looking back from the closing of the "pre-information" era, Glen Culler stands out as one of the under-recognized pioneers of modern computing. His work not only advanced the technical state-of-the-art by developing innovative computer system designs ten or twenty years ahead of the mainstream computing industry, but also advanced the vision of computers as tools for helping people to learn, to think about and explore important problems, and to see beyond the limits of their senses.

Moreover, the Santa Barbara community continues to benefit from Glen's work and the commitment of the University to fostering an entrepreneurial spirit in their students. At a time when high-technology business development and retention is so keenly on our minds, it is important to recognize that over 25 companies, many of which are still viable and growing organizations, spun out of Glen's work and the College of Engineering's Computer Research Laboratory.