Re: [squid-users] Benchmark of compiler options on Squid?

Hmm. OK.

I would have thought that the Select loop, at minimum, would benefit from
more efficient CPU utilization. After all, doesn't Squid have to examine
the cache bitmap to determine if the requested object is already in the
cache?

Thanks for the response.

On Saturday 29 September 2001 11:20 am, Joe Cooper wrote:
> Henrik is correct.
>
> Squid compiled -O and -O3 results in nearly immeasurably small
> differences (around 3%--I wouldn't even call it statistically
> significant), as does compiling for a specific compiler. The slowdowns
> in Squid, it seems, are not of the 'tight loop' sort that compiler
> optimizations can sometimes help.
>
> Henrik Nordstrom wrote:
> > I don't think the compiler optimizations will make much of a difference
> > from Squid. The bottlenecks in Squid is mostly on other levels.
> >
> > Regards
> > Henrik Nordström
> > Squid Hacker
> >
> > Steve Snyder wrote:
> >>Has anyone benchmarked various GCC compiler optimizations used when
> >>building Squid?
> >>
> >>I'm building for a Pentium3 box, using the compiler included in
> >> RedHat's v7.1 release. I understand that there's some
> >> environment-specific variables (CPU L2 cache size, etc.), but I'm
> >> hoping that some industrious person has tested differing compiler
> >> options when compiling for for the i686 processor.
> >>
> >>Any info on this? Thanks.
>
> --
> Joe Cooper <joe@swelltech.com>
> Affordable Web Caching Proxy Appliances
> http://www.swelltech.com
Received on Sat Sep 29 2001 - 10:55:33 MDT