Ilya Lipovsky (lipovsky@cs.bu.edu) on 2/15/08 wrote:
---------------------------
>Interesting. Curiously, I work in the high performance numerical computing field,
>and all of my (former, potential, and current) bosses, I mean the ones who write
>optimized algorithms in assembly for a living - they all have *non-CS* education,
>at least originally - all of them majoring in either physics or math (but with Ph.D.'s),
>though I've seen one coming from chemistry.

Heh -- there are various reasons for that; for older guys, it's because there really weren't many CS courses about during those days. When I was a student all of the oldest generation had PhDs in either physics or maths.

For the younger generation, that might tell you more about the job prospects in their chosen area of study. :-)

Or, it could just be that numerical computing is an applied branch of computer science that is more useful in fields like physics than in computer science itself, and there's more motivation for a physicist to pick up the little bit of computer science they need to know to implement those algorithms effectively than there is for a computer scientist to do the reverse.

>But I don't know if this can be generalized to saying that a CS Ph.D. is incompetent
>when it comes to writing fast code. *If* this indeed is the case, then maybe it's
>just because a lot of CS programs in academia focus more on higher level languages
>and abstractions, omitting the engineering component. If I recall properly, one
>of my CS professors joked "Computer Science is the only "science" that's not."

It's a very broad field. Some parts are definitely science, others are definitely engineering. (One of my friends was doing a PhD so theoretical that some professors complained it should really have been a maths PhD. But that was only because they didn't understand his work. It was definitely appropriate for CS, in my view; they were just from the "applied" end of the spectrum.)

I would no more expect someone on the science end of the spectrum to write really tight code than I would expect a professor in mechanical engineering to do a really good job servicing my car. That doesn't mean they're useless, just that you need to hire the right person for the job, and a lack of technical skill does not reflect on their ability otherwise. I didn't like the code that my aforementioned friend wrote, but I respected the fact that he implemented a Turing machine in the K&R C preprocessor (the ANSI one isn't Turing complete).