2009 Kaufman Award – Dr. Randal Bryant
Nonetheless, we do look to young grad students for breakthroughs. They come to their research at universities with different backgrounds [and fresh ways of looking at things]. And, it’s often people who come into a new area, who aren’t steeped in the lore of how things have always been done, that end up having a breakthrough.
The classic example is the World Wide Web. That was the brainchild of Tim Berners-Lee in Switzerland, who was supporting physics researchers there. He wasn’t a mainstream researcher in IT, but just someone who was trying to solve a problem
Q – Also speaking of grad students, why are there so many students at U.S. universities coming in from overseas? Do we have a problem with the quality of our undergraduate education here?
Dr. Bryant – Personally, I’ve enjoyed working with many students here at CMU who come from India, particularly in research areas related to IT. I attribute that to the selective filter in India, where only 2 percent of the students who apply to the IIT system are actually accepted. Naturally, filtered by that exam process, only the very brightest students are admitted.
However, I also think that at schools like IIT, there is more respect for a theoretical approach to problem solving. EDA combines practical instantiation [of a problem] with a mathematical perspective. It’s the putting together of the algorithms with thinking in a systematic way [that produces] the solutions. Integrating all of that together takes a special kind of talent.
We certainly get great American grad students at CMU, but they tend to be very focused. Either they are very theoretical, or they just want to write code.
Q – How would you restructure undergraduate education in the U.S. to meet both of those goals?
Dr. Bryant – Here at CMU, we keep a close eye on the twin footings for our undergrads, both the practical and the theoretical. It’s easy to swing too far in one direction, or the other.
In this country, our focus is often on the practical, our conversations are all about what succeeds in industry. But industry often concentrates on short-term results, which gets people thinking more about how to program using the latest tool, or how to use the latest development environment, while throwing away [considerations about] the more fundamental principles.
Q – I‘ve often had the impression that top-tier engineering schools in the U.S. have considered teaching VLSI design as something that belongs in a trade school, not a great university. Does CMU teach VLSI design?
Dr. Bryant – Well, we teach the algorithms, more than how to use this tool or that from Company X. There’s a similarity here to the balance between teaching programming versus teaching the theory of computer architecture.
You can teach one aspect of the latest tool from Microsoft, for instance, or you can teach the concept of a programming language. We are always trying to teach our students material that will be useful now, and will also help them throughout their entire career.
A lot of students like to see a physical manifestation of their work, of course, and CMU doesn’t have a chip fabrication facility. [For that reason], a lot of student design gets done in FPGAs.
FPGAs have made it possible to do really interesting logic design projects in the classroom that you would never be able to do if you had to wire together chips. Now, in a one-semester course, it’s possible to design a 32-bit microprocessor, map in onto an FPGA, and run it on a board.
Q – Does that mean the future of EDA is the FPGA?
Dr. Bryant – Various forms of programmable hardware will play bigger and bigger roles all of the time, given that the cost of manufacturing is so exorbitant.
Q – So just to clarify, could someone graduate from CMU with a degree in VLSI design?
Dr. Bryant – Yes, but it’s hard, given all of the deep-submicron issues today.
When I went through school, I learned the [Carver] Mead and [Lynn] Conway approach to VLSI design, which contained only a few basic ideas and was quite simplified. Nowadays, real circuit designers have got it a lot harder. Knowing how to do a circuit design [is very difficult], and it takes quite a while for an engineer to develop full competency as a designer.
That said, we do prepare our students for those types of jobs, and they do get job offers at semiconductor companies after they graduate. Some number of our Computer Science students go off to work at companies like Facebook, Google, or Microsoft, but our students in ECE often go to work at Intel.
Q – Do any of your graduates join startups?
Dr. Bryant – Yes, a fair number go to startups as well.
Q – What would you say to a young graduate student today who was choosing between EDA and other areas in technology?
Dr. Bryant – For me, EDA has always been a way to combine my interest in digital design with more abstract principles. It provides a very nice balance between hardware and software, between theory and practice.
Issues of career …
Q – Please tell me about your own career?
Dr. Bryant – I grew up in a suburb of Detroit, where my dad had started a company in the 1960’s. He also had his Ph.D. in electrical engineering. His trick had been to marry the daughter of the department head at the University of Illinois. My grandfather was Dean of Engineering at Illinois.
Q – Was it a burden or opportunity to be the son and grandson of Ph.D.s in engineering?
Dr. Bryant – I view it in positive terms. My mother also received her Ph.D. – in the same year I graduated from high school – so I definitely come from a strong academic heritage.
I went to the University of Michigan as an undergraduate, and earned a degree in applied math. I started on my Ph.D. at MIT in 1974, and finished in 1981. While I was at MIT, Lynn Conway came there and spent a semester teaching. I was one of the students in her VLSI design class and was really inspired by that. I am very much a protégéé of Lynn’s.
We were all given projects to do in the class, but no tools. I was in a group of 3 or 4 students, and we collaborated to come up with a circuit design. But we had no way of knowing if what we had designed would even remotely function.
After the course was over, I started to think about writing a simulator – a way to simulate a circuit that viewed each transistor as a simple switch, either open or closed. Most simulators at that time used fancy, numerical equations to solve the model of transistor behavior. I came up with something, instead, that actually worked. However, the class was already over, as I said.
That next semester, I was a teaching assistant for that same class, which was being taught at that point by Jonathan Allen. Part of my job as teaching assistant was to make my simulator available to the class. That was its first widespread use.
Over the years, switch-level simulation became the topic of my Ph.D. thesis and my work – both at Cal Tech, where I taught after finishing at MIT, and at CMU. I’ve now worked on several generations of switch-hold simulators, which have been widely used in industry at this point.
Q – When you finished your Ph.D., was it difficult to decide between industry and academia?
Dr. Bryant – No, not at all. I was inspired by my grandfather, a person I admired a great deal. When I started teaching at Cal Tech, it was a hotbed of VLSI design. I spent 3 years there, and [found it] extremely stimulating, but it was hard to be so far away from the East Coast, so I came to CMU.
Be the first to review this article