At 12:34 PM 1/20/00 +0100, Peter Eisentraut wrote:
>No, compiler don't do this either. This is specifically for keyword
>lookup. The whole idea is that you have one set of keywords that hardly
>ever changes (such as for programming languages), then you take one
>afternoon off, play around with the different options until you have a
>lookup function which processes your particular keyword set fastest.
>Again, this is not a big deal to me, I just did it to play around. In any
>case it seems to run faster, but I wasn't sure if people wanted to bother.
The binary search could very easily be sped up, actually (I just peeked).
If you'd care to e-mail me your two test cases, the one using the
output of gperf and the one using the current binary search, I'd
be more than willing to demonstrate.
I have nothing against speeding things up, though again identifying
keywords takes a vanishingly small part of the time required to
execute a query (or compile a program, for that matter). I more
or less dislike adding dependencies to external tools when it
can be avoided, though.
I just built a new PC to do linux-based development on and it
would be fun to have your benchmark program to compare my
new and old linux boxes anyway :) (P 200 classic vs. P500E, guess
which will win?)
- Don Baccus, Portland OR <dhogaza@pacifier.com> Nature photos, on-line guides, Pacific Northwest Rare Bird Alert
Serviceand other goodies at http://donb.photo.net.