🚀 go-pugleaf

In article <EACHUS.95Oct12163833@spectre.mitre.org>,
Robert I. Eachus <eachus@spectre.mitre.org> wrote:

>   I'd say overall not worth the effort, but in this case it was the
>difference between very marginal and acceptable performance.  (Looking
>for obvious coding boners like looping through a list instead of doing
>a binary search turned up another 6%, which left a reserve.)

Ummm, yes. We do manage to get into endless arguments over whether this
or that compiler, this or that language, is more "efficient." The big
performance payoffs are, in fact, often to be had from eliminating
"obvious coding boners" like using a linear algorithm where a log N
one was appropriate.

I am constantly amazed at the number of industry folks I've run into
over the years who put down stuff like "big O" algorithm performance
prediction as something horribly academic and useless.:-)

Also the number of students, grad and undergrad, who complain about
having to learn all that boring algorithm theory crap.:-)

I am certainly not flaming Bob here, just taking his point and
emphasizing that a little effort (or maybe a LOT of effort) spent
in learning and choosing appropriate algorithms is probably a better
investment of time than flipping compiler switches on and off.

Mike Feldman