Comments for Bits and pieces of thoughts and personal achievements http://hellohellonico.wordpress.com Just another WordPress.com weblog Fri, 21 Sep 2007 04:44:15 +0000 http://wordpress.com/ hourly 1 Comment on OpenWFEru on JRuby! Getting there … by on the JRuby front « processi http://hellohellonico.wordpress.com/2007/09/21/openwferu-on-jruby-getting-there/#comment-9 on the JRuby front « processi Fri, 21 Sep 2007 04:44:15 +0000 http://hellohellonico.wordpress.com/2007/09/21/openwferu-on-jruby-getting-there/#comment-9 [...] See Nicolas’ post for the full details. [...] [...] See Nicolas’ post for the full details. [...]

]]> Comment on Faster ruby(2) by hellonico http://hellohellonico.wordpress.com/2007/07/20/faster-ruby2/#comment-6 hellonico Fri, 20 Jul 2007 08:49:46 +0000 http://hellohellonico.wordpress.com/2007/07/20/faster-ruby2/#comment-6 Charles, I've updated with my results with your options. It does really speed things up. Since I am trying to get an application to work seamlessly whether on ruby or jruby, I am going to try those options on our own test suite as well. Will keep you posted ! Thanks a ton for that. Charles,

I’ve updated with my results with your options. It does really speed things up.
Since I am trying to get an application to work seamlessly whether on ruby or jruby, I am going to try those options on our own test suite as well.
Will keep you posted !

Thanks a ton for that.

]]> Comment on Faster ruby(2) by Charles Oliver Nutter http://hellohellonico.wordpress.com/2007/07/20/faster-ruby2/#comment-5 Charles Oliver Nutter Fri, 20 Jul 2007 07:13:47 +0000 http://hellohellonico.wordpress.com/2007/07/20/faster-ruby2/#comment-5 FYI, there's a few things you can do to speed this test up in JRuby: * Pass -J-server to JRuby to use the Java "server" VM * wrap the 0.upto pieces in methods and pass -J-Djruby.jit.threshold=0 to get them to compile (the whole script won't compile because the compiler doesn't compile class defs yet) * Pass -O to JRuby to turn off ObjectSpace With those changes made the two tests came in at around 9 seconds and 7 seconds, with MRI around 7.5 seconds on my sytsem. FYI, there’s a few things you can do to speed this test up in JRuby:

* Pass -J-server to JRuby to use the Java “server” VM

* wrap the 0.upto pieces in methods and pass -J-Djruby.jit.threshold=0 to get them to compile (the whole script won’t compile because the compiler doesn’t compile class defs yet)

* Pass -O to JRuby to turn off ObjectSpace

With those changes made the two tests came in at around 9 seconds and 7 seconds, with MRI around 7.5 seconds on my sytsem.

]]> Comment on Faster ruby by hellonico http://hellohellonico.wordpress.com/2007/07/20/faster-ruby/#comment-4 hellonico Fri, 20 Jul 2007 05:27:08 +0000 http://hellohellonico.wordpress.com/2007/07/20/faster-ruby/#comment-4 Charles, Thank you for your comments and insight. I went back and ran the experiment again: http://hellohellonico.wordpress.com/2007/07/20/faster-ruby2 I wonder what makes those two tests behaving differently from a jruby point of view. Charles,

Thank you for your comments and insight.

I went back and ran the experiment again:
http://hellohellonico.wordpress.com/2007/07/20/faster-ruby2

I wonder what makes those two tests behaving differently from a jruby point of view.

]]> Comment on Faster ruby by Charles Oliver Nutter http://hellohellonico.wordpress.com/2007/07/20/faster-ruby/#comment-3 Charles Oliver Nutter Fri, 20 Jul 2007 04:43:01 +0000 http://hellohellonico.wordpress.com/2007/07/20/faster-ruby/#comment-3 The second example is somewhat surprising to me. In the first case, if Niko is a constant in the current module/class namespace, it will do a single lookup to find it, followed by a single lookup in the Niko namespace to find MY_CONST. In the second example, it will first search the current module or class for MY_CONST and then search the included module for MY_CONST. Both cases work out to two constant hash lookups, so it's strange that one would be faster in Ruby than the other... The second example is somewhat surprising to me. In the first case, if Niko is a constant in the current module/class namespace, it will do a single lookup to find it, followed by a single lookup in the Niko namespace to find MY_CONST. In the second example, it will first search the current module or class for MY_CONST and then search the included module for MY_CONST. Both cases work out to two constant hash lookups, so it’s strange that one would be faster in Ruby than the other…

]]> Comment on Tracing calls in ruby by John Mettraux http://hellohellonico.wordpress.com/2007/07/19/tracing-calls-in-ruby/#comment-2 John Mettraux Thu, 19 Jul 2007 10:13:48 +0000 http://hellohellonico.wordpress.com/2007/07/19/tracing-calls-in-ruby/#comment-2 Nice ! Thanks ! Nice ! Thanks !

]]>