I attempted to measure the difference way back in the stone ages (about
a year ago), and couldn't find enough difference to make any call.
Literally, the difference was around 3%...In the case of Polygraph
benchmarks and Squid 3% is not far from the margin of error. ;-)
Florin Andrei wrote:
> Is there any significant advantage to compile Squid with -O3 instead of
> -O2? (anyone saw improvements in the benchmarks?)
> Or perhaps add other compile optimisations? (without hurting stability,
> which is very important in this case)
>
> My other compile parameters are: "-march=i686 -D_REENTRANT".
> The target system will be a dual PIII Coppermine (256kB cache).
> I'm compiling with gcc-2.96-112 on Linux Red Hat 7.3. I'm using the same
> distribution on the target system.
>
-- Joe Cooper <joe@swelltech.com> Web caching appliances and support. http://www.swelltech.comReceived on Sat Sep 14 2002 - 02:44:48 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:10:19 MST