Hi Tom,
I made the assumption that DSO support would affect all content engines
equally. Since it's only the relative numbers that matter, a uniform 5% hit
wouldn't affect the benchmark results. It's the same as if the benchmarks were
run on a 5% slower CPU.
5% is really in the noise. The real problem isn't the 5% that each engine
might be affected by DSO, it's the big question of whether the benchmarks are
representative. For example, mod_perl gets killed by adding function calls to
the benchmark while Java isn't affected. I suspect if I added field
references, Perl and PHP would start losing badly. In other words, as the
benchmark becomes more realistic, Java performance looks even better against
Perl and PHP. So, if anything, the benchmark is biased in favor of mod_perl
and php.
-- Scott
Renzo Tom` wrote:
> Hi there, on http://www.caucho.com/articles/benchmark.html you mention the
> following:
>
> <quote>
> Apache configurations used Apache 1.3.9. All modules were compiled with DSO
> support.
> </quote>
>
> Why has this been done?
>
> On http://www.apache.org/docs/dso.html you can find a few disadvantages of
> DSO:
>
> <quote>
> The server is approximately 5% slower at execution time under some platforms
> because position independent code (PIC) sometimes needs complicated
> assembler tricks for relative addressing which are not necessarily as fast
> as absolute addressing
> </quote>
>
> Please make sure DSO isn't a slowdown factor on your machines, benchmarks
> only count for something when done fairly.
>
> Cheers!
>
> -Renzo
