This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: C optimization tests - almabench results
- From: Jim Wilson <wilson at specifixinc dot com>
- To: Uros Bizjak <uros at kss-loka dot si>
- Cc: gcc at gcc dot gnu dot org
- Date: Mon, 03 May 2004 16:45:42 -0700
- Subject: Re: C optimization tests - almabench results
- References: <4095FB93.20005@kss-loka.si>
Uros Bizjak wrote:
I don't think that 15% faster clock can produce such a difference...
Comparing benchmark results on two different machines is tricky. There
are lots of different variables that may be affecting the results.
1) different linux distros
2) different kernel versions
3) different glibc versions
4) different memory architectures
5) different main memory sizes and speeds
6) different disk speeds
7) different processors
The last one in particular may be an issue. You seem to be assuming
that all Pentium4s are the same, but they aren't. There are different
kinds of Pentium4s with different instruction sets and different
performance profiles. For instance, your Pentium4 probably has the
hyperthreading feature, whereas the other one probably doesn't. The
previous post mentioned it was using a "Northwood" Pentium4. You
probably have a "Prescott". There are probably other differences
between Northwood and Prescott. If they have different cache sizes, and
the program fits in the Prescott cache but not the Northwood cache then
that could give significant performance differences.
It is also possible that one of you made a mistake.
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com