On Fri, Feb 17, 2012 at 9:00 AM, Jon <jon.forums / gmail.com> wrote:
> Indeed!
>
> And I'm sure there's more to be had from https://github.com/jruby/jruby/wiki/PerformanceTuning (and master) and other goodies you guys have worked hard to implement.
>
> How about we easily knock off 1.2s/~37% for this short-running process?
>
> C:\projects\measurements-git>set JRUBY_OPTS="--server --fast"
>
> C:\projects\measurements-git>rci bench core_rd_filelines_crlf
> jruby 1.6.6 (ruby-1.8.7-p357) (2012-01-30 5673572) (Java HotSpot(TM) Server VM 1.7.0_02) [Windows 7-x86-java]
> Rehearsal ----------------------------------------------------------
> core_rd_filelines_crlf  2.793000  0.000000  2.793000 ( 2.746000)
> ------------------------------------------------- total: 2.793000sec
>
>               user   system   total    real
> core_rd_filelines_crlf  2.075000  0.000000  2.075000 ( 2.075000)
>
>
> I should have deleted the JRuby stuff as my point was the significant (relative) improvement of 1.9.3-p125 corresponding to a noticable usability improvement for MRI Windows users...MRI.prev vs. MRI.current vs. MRI.future.
>
> Sadly, it seems you must continually be diligent to ensure old perceptions don't re-sprout and take hold ;) Sorry for the distraction and muddying the waters.

It would probably get faster for a longer run...but yeah, I agree it's
great to see that MRI 1.9.3 and master have fixed the egregious perf
issues from 1.9.2 and 1.9.3p0.

- Charlie