Friday, January 09, 2004

Comparing apples and apples when you should be looking at who's eating the apple..

This " Language Performance Round-up: Benchmarking Math & File I/O " article provides the wrong impression on languages use. For all its worth, Fortran would probably do better on all these if it were given a try. Similar types of findings were found in the great language shootout (even available for win32 machines). Yet, it's been my experience that this type of benchmark is only needed at the end of a development stage. It's fine and dandy to see Java doing better than say python, but if it takes your developers three to six times longer to develop it and longer to debug it, it won't matter that Java is quicker. Your project will run into the difficulty of not being responsive to either yourself (one can get bored easily) or the people giving you work, and you may be out of business quicker than java can run this benchmark. For that matter, it just doesn't make sense to look at a math benchmark without looking at matlab or octave.

No comments:

Printfriendly