brainwagon

This begins as a product announcement, but is really just a clumsy way of introducing a story.

Flipping through my usual news sources, I’ve seen a couple articles about PeakStream, a company which has a set of software products for creation of high performance code for GPUS. Normally, such a thing would be of mild interest to me, but as I read some more, I realized that one of the brains behind the company is Stanford professor Pat Hanrahan, a name which is familiar to most in the computer graphics industry, and a name which has played a role in my own career.

Back in 1989 I had just graduated with a Masters in Computer Science from the University of Oregon. I had gone on a number of interviews without much success, but then flew to New Jersey to interview for a position in scientific visualization at the Department of Applied Mathematics at Princeton. Over the next year and a half, I helped a number of their staff to create animated visualizations of fluid flow simulations (a topic which, admittedly, I know very little about even now), mostly by writing simple renderers to display particle and flow line data, and such mundane tasks as writing control software for Lyon Lamb controllers for video controllers and dubbing video tapes.

But somewhat coincidently, I had arrived at the time that Pat Hanrahan had signed on to create a computer graphics lab at Princeton. I remember our first meeting at P J’s Pancakes in Princeton, where we discovered that we had a number of interests in common. Over the period of a few months, I’d drop in on him and have a peek into what he was doing, and sat in on some of his undergraduate lectures, where I must admit that I learned quite a bit: after all, I was almost entirely self-taught in computer graphics, and let’s face it, it is hard to spot your weaknesses when you have only yourself to rely on.

The thing that amazed me most about Pat was his amazing breadth of knowledge, and the seemingly effortless way that he unified seemingly widely different disciplines and created new ideas from the result. The most memorable of these was on the day that I walked into his office and he demonstrated his first program that did rapid hierarchical radiosity. He had a cube environment with a triangle blocker on the screen, clicked a button, and within a second, a complete radiosity solution was calculated on what was probably an SGI 4D25 or so. I was amazed. What was even more amazing was his explanation. A half an hour later, as he explained how it was related to n-body simulation (a subject with which I was not completely unfamiliar) and how these systems resulted in block structured matrices that could be efficiently solved, I was blown away.

As I recall, that was one of probably five or six papers he was working on at the time.

I consider him to be certainly one of if not the smartest guys I’ve ever met.

By the time 1991 rolled around, I had tired of feeding VCRs and was seeking new employment. Due in some measure to the strength of Pat’s recommendation, I secured an interview and later a job at Pixar. And, the rest, they say, is history.