Richardthughes
Posts: 11178 Joined: Jan. 2006
|
Mark CC destroys DaveTard, then takes a whiz on his still twitching V-shaped corpse:
http://scienceblogs.com/goodmat....hp#more
Quote | Davescott clearly isn't qualified to judge much of anything about compiler design. It is, in fact, an extremely mathematical field of computer science. You see, what you do working on an optimizing compiler is that to figure out what information is expressed in the static semantics of a program, and then use that to do performance-improving transformations transformations that provably don't alter program semantics. It's a field which is highly dependent on things like lattice theory, domain theory, graph theory, and denotational semantics. Parallel compilation also generally involves a fair bit of linear algebra and sometimes vector analysis.
What's more, if Davescott had bothered to find out a teeny tiny bit about my dissertation, he would found that the main contribution of it is something called the parallel continuation graph: a mathematical representation of parallel computation in terms of a š-calculus inspired variation of continuation-passing form compilation. The neat thing about the PCG is that many parallel performance optimizations could be expressed via simple graph restructurings. I'm quite proud of that dissertation; I still think the PCG is incredibly cool. And I challenge anyone to argue that proving that the transformation of source code into the PCG was semantically valid, and the proofs of the validities of graph-based program transformations was "very little in the way of math".
Quote | For someone who's been in commercial computer R&D for over 10 years Carroll's patent portfolio (2 patents, sole inventor on one of those) is abysmal especially for an uber patent-house like IBM where he spent most of his time so far. He's got a fair number of journal publications but that's a metric for academicians not industry. Both the patents were in client-server networking i.e. zero math content. I generated twice that many patents in half the time and I was just a non-degreed senior systems engineer. Even that was still short of my performance plan target which called for being a named inventor on two patent submissions per year. |
Bzzzt. Wrong. Industry actually quite likes publications. The mantra at IBM was that there are three things that IBM wants to see from researchers: papers, patents, and products. To be successful, you need to be generating at least two. I was mostly a papers and products guy. (Personally, I don't like software patents; 14 year monopolies on software concepts seems completely unreasonable to me, so I didn't file them unless I had to. You see, DaveScott, some of us have this quaint idea about this thing called "ethics". I realize that's probably a foreign idea to someone who works on a DI site.)
What's funny about this is that once again, Davescott blows it by not bothering to read. Because the second of those two patents - granted 1 week before I left IBM - is on search: source code search optimization based on a kind of multidimensional vector analysis. You assign program fragments to locations in a many-dimensional search space, and then find things that are close to a particular search vector. (And why the patent there? Because there are so many patents in IR that you have to file as a matter of self-protection, to establish when you did the work, so that you'll have a defense if someone else tries to file a patent that overlaps with it.) So in his attempt to smear me as unqualified of judging a mathematical argument based on search, he specifically mentions my work on search. Not super bright, Dave.
|
-------------- "Richardthughes, you magnificent bastard, I stand in awe of you..." : Arden Chatfield
"You magnificent bastard! " : Louis
"ATBC poster child", "I have to agree with Rich.." : DaveTard
"I bow to your superior skills" : deadman_932
"...it was Richardthughes making me lie in bed.." : Kristine
|