pdahorek / seznam.cz wrote:
> There isn't noticable difference on Linux, it's even slightly faster.

The problem isn't the noticeability in Linux.  I suspect the
problem here is Linux hides performance problems with fast syscalls:

> Linux
> 
> ~~~
> 2.1.9 77991 i/s
> 2.4.1 78497 i/s
> ~~~
> 
> Windows
> 
> ~~~
> 2.1.9 1143000 i/s
> 2.4.1   39829 i/s
> ~~~

Are those numbers on the same hardware?  If so, it's because
our glob performance on Linux always sucked :)

So, I suspect the performance on 2.1.9 was good because Ruby
used Win32-specific APIs; but when the code path changed to
use work the same on both systems, it got silly slow.

I've been having a tough time figuring out what changes in the
2.1..2.2 era did what over time, especially on a platform
I don't run...

Can you run "git bisect" to narrow down the performance problem
to a particular commit?

Thanks.

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>