[htdig3-dev] Re: More gprof results


Subject: [htdig3-dev] Re: More gprof results
From: Geoff Hutchison (ghutchis@wso.williams.edu)
Date: Mon Feb 28 2000 - 06:10:44 PST


At 1:28 PM +0100 2/28/00, loic@ceic.com wrote:
> I was first amazed to see this. It looks really bad. In a way it is.
>The NFields() method is used by the compression without restriction or
>cache, hence the huge number of calls. However, NFields() and Info() are
>inlined and only show when compiling with -g -p. Even inlined, given the
>number of calls, a cache of the result is more than necessary.

Yeah, I did see the inline statements, but I was worried mostly about
the sheer number of calls. For 5000-odd pages, it seems silly to be
calling something 2 billion times!

Good work though, that should help narrow the performance gap.

> Thanks a lot for doing this profiling and apologies for this performance
>bottleneck.

Well, that's the reason for profiling--it's somtimes hard to see
what's a bottleneck and what isn't! It's also hard to know how things
scale, so I'll probably run this test a few more times--it takes a
while because profiling is slow, but I'm willing to run overnight.

Next up, I'll try htsearch. That should also be interesting.

-Geoff

------------------------------------
To unsubscribe from the htdig3-dev mailing list, send a message to
htdig3-dev-unsubscribe@htdig.org
You will receive a message to confirm this.



This archive was generated by hypermail 2b28 : Mon Feb 28 2000 - 06:15:43 PST