I'm having some trouble getting it to run on OS X, so I may have to try
running it on my Linux box a bit later. Could you try your run with -X-C
to turn off compilation? If you find that it's faster...well we may have
a good clue on where something is less than efficient in compiled code.
Uwe Kubosch wrote:
Great work Charles!
For issue http://jira.codehaus.org/browse/JRUBY-2803 I can see no
difference, however running with "-r profile" actually speeds things up
considerably, both 1.1.2, 1.1.3 and trunk. Maybe a hint for
improvement?
Uwe
On Sun, 2008-07-20 at 01:01 -0500, Charles Oliver Nutter wrote:
I just committed a series of changes I've been working on to refactor
Java Integration for performance, memory, etc. On my tests, it seems to
be much faster, and should be using less memory and be easier on the GC.
But since I've only run my own microbenchmarks, I'd appreciate others
testing out their apps.
It's on trunk now, and because this work has gone so well I think we're
now planning to push out a 1.1.4 release in the next month (much shorter
cycle than 1.1.3) that includes as much cleanup of the existing JI stuff
as possible. Given that more and more folks are getting into JI, we
recognized that some help is warranted to get it cleaned up and
performing better. Also, we can get ourselves better acquainted with the
current JI code, which will mean a shorter path to the "new JI" we want
for the next major JRuby release toward the end of the year.
So give it a go folks. Report any bugs, and definitely let me know how
performance looks.
- Charlie
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
