Brian Candler wrote:
> 
> On Mon, Apr 07, 2003 at 09:21:04PM +0900, ts wrote:
> >  See [ruby-talk:68578]
> >
> > ruby168 BIG    = 319
> > ruby168 SMALL  = 95
> >
> > ruby18p2 BIG   = 12
> > ruby18p2 SMALL = 12
> 
> Oh I'm not arguing that ruby 1.8p2 isn't better than ruby 1.6.8. I believe
> that those tests were done under Windows, and that just shows how dreadful
> Windows I/O libraries are.
> 
> Under FreeBSD I see almost zero difference between ruby 1.6.8 and ruby 1.8p2
> for the attached program, showing that a real operating system comes with
> decent I/O libraries.
> 
> Perl is still nearly twice as fast as ruby though:
> 
> [brian@vaio ruby]$ time ruby x.rb /usr/share/dict/words
> Elapsed = 2.146236062
> 
> real    0m2.492s
> user    0m2.362s
> sys     0m0.120s
> [brian@vaio ruby]$ time perl x.pl /usr/share/dict/words
> Elapsed = 1
> 
> real    0m1.347s
> user    0m1.251s
> sys     0m0.080s
> 
> However evil or optimised perl's I/O libraries are, I doubt that that's the
> reason for the difference.
> 
> Regards,
> 
> Brian.

Does running your code through the profiler provide any insight?  Any
chance you could post the Perl and Ruby profiler results side by side?

perl -MDevel::Profiler x.pl # assuming you have Devel::Profiler
ruby -rprofile x.rb

Regards,

Dan