Jack Schwartz

Jack Schwartz died last week. You can read John Markoff's NYT obituary here:

http://www.nytimes.com/2009/03/04/science/04schwartz.html

He was a remarkable person, bigger than life in a lot of ways. He was a central part of my friend and mentor Martin Davis' life, so I always heard a lot about him. His death made us really sad around here. Here are some thoughts on my own association with him, and my personal indebtedness to him:

I saw him more at parties than professionally, though he was on my thesis committee and my oral exam committee, and he had a huge cultural influence over the department I was educated in--he started and ran the department--and Cocke and Schwartz was my compiler bible.

But, oddly, while I had very little to do with him directly, he had a remarkable (positive) influence on my life and career. It wasn't until many years after the last of these incidents that I put it all together and realized how indebted I was:

(1) He turned me on to a lifetime of computer architecture/hardware design with one single substitute lecture in a course I was taking, pretty much at random. I've still never seen anyone describe how computers work quite as comprehensively and intuitively as that.

(2) When I went up to him after that lecture, excited, he told me I should see Ralph Grishman (eventually my advisor) and work with him on Puma (which led to Trace Scheduling, VLIWs, etc).

Maybe 3-4 weeks after I started on the PUMA project, Jack appeared on the 2nd floor, and asked Ralph how it was going. Ralph replied, and believe me I am telling you his response word-for-word because of how touched I was, "Send me another Josh Fisher and I'll be done in 2 weeks." Jack turned to me and asked if I had any support. My answer was no, and I had an assistantship in minutes.

(3) He got excited during my defense, and I could see right then and there how sound my judgment had been when I decided not to ask him to be my adviser. I might still be a graduate student today if I had; there was a lot of work in trying to prove out what I had put forward. Anyway, shortly afterward, I had lunch with him at some horrible Grand Concourse-style Cantonese restaurant on like 2nd avenue that he suggested, and he told me he had just been visiting Yale, that I'd be a good fit there, and I should call Alan Perlis.

(4) During his DARPA stint, I visited him for half a day. The next thing I knew, I had sold a 28 operation per cycle Multiflow Trace to Paul Schneck at the "Supercomputer Research Center" (the NSA's public thing). That sale cemented Multiflow's getting off the ground--it was the single most important sale the company made.