Comments on: Dismal performance with IronPython http://www.smallshire.org.uk/sufficientlysmall/2009/05/17/the-performance-of-python-jython-and-ironpython/ sin(x) = x Mon, 09 Jan 2012 20:46:03 +0000 hourly 1 http://wordpress.org/?v=3.1.3 By: Sufficiently Small » IronPython 2.0 and Jython 2.5 performance compared to Python 2.5 http://www.smallshire.org.uk/sufficientlysmall/2009/05/17/the-performance-of-python-jython-and-ironpython/comment-page-1/#comment-2063 Sufficiently Small » IronPython 2.0 and Jython 2.5 performance compared to Python 2.5 Fri, 22 May 2009 11:34:37 +0000 http://www.smallshire.org.uk/sufficientlysmall/?p=63#comment-2063 [...] previous post covering the performance problems I’ve been experiencing with IronPython raised some [...] [...] previous post covering the performance problems I’ve been experiencing with IronPython raised some [...]

]]> By: David Fugate http://www.smallshire.org.uk/sufficientlysmall/2009/05/17/the-performance-of-python-jython-and-ironpython/comment-page-1/#comment-1995 David Fugate Wed, 20 May 2009 04:04:26 +0000 http://www.smallshire.org.uk/sufficientlysmall/?p=63#comment-1995 I have a couple of suggestions/questions that might give you better performance with IronPython. First, are you NGen'ing IronPython/DLR assemblies by hand before running IronPython? The installer for IronPython 2.0.0 doesn't Ngen assemblies by default, but more importantly this feature is broken on x64 platforms - see http://ironpython.codeplex.com/WorkItem/View.aspx?WorkItemId=22001. Any ways, the difference between Ngen'ed and Non-Ngen'ed assemblies for a "hello world" program on 32-bit machines is about a factor of four in my experience. Next, are you using IronPython 2.0.0 or 2.0.1? The reason I ask is there were some improvements targeting performance that went into 2.0.1 and it's not clear from the report above which version you were using. Finally, I think you'll find IronPython performance to be a bit better on a 32-bit OS. It'd be very interesting to see the performance comparison between IronPython/Jython/CPython duplicated on a 32-bit OS... I have a couple of suggestions/questions that might give you better performance with IronPython.

First, are you NGen’ing IronPython/DLR assemblies by hand before running IronPython? The installer for IronPython 2.0.0 doesn’t Ngen assemblies by default, but more importantly this feature is broken on x64 platforms – see http://ironpython.codeplex.com/WorkItem/View.aspx?WorkItemId=22001. Any ways, the difference between Ngen’ed and Non-Ngen’ed assemblies for a “hello world” program on 32-bit machines is about a factor of four in my experience.

Next, are you using IronPython 2.0.0 or 2.0.1? The reason I ask is there were some improvements targeting performance that went into 2.0.1 and it’s not clear from the report above which version you were using.

Finally, I think you’ll find IronPython performance to be a bit better on a 32-bit OS. It’d be very interesting to see the performance comparison between IronPython/Jython/CPython duplicated on a 32-bit OS…

]]> By: Dino Viehland http://www.smallshire.org.uk/sufficientlysmall/2009/05/17/the-performance-of-python-jython-and-ironpython/comment-page-1/#comment-1991 Dino Viehland Wed, 20 May 2009 01:40:54 +0000 http://www.smallshire.org.uk/sufficientlysmall/?p=63#comment-1991 It would be interesting to at least see the run on 2.0.1 which is the latest stable version (and includes lots of performance fixes on top of 2.0). It might also be nice to see it run on x86 instead of x64 but I that certainly shouldn't be the cause of this massive of slowdowns. It could be that you've simply found some pattern that absolutely destroys IronPython perf. If you'd like to help us track down that issue so we can fix it feel free to e-mail me. It would be interesting to at least see the run on 2.0.1 which is the latest stable version (and includes lots of performance fixes on top of 2.0). It might also be nice to see it run on x86 instead of x64 but I that certainly shouldn’t be the cause of this massive of slowdowns.

It could be that you’ve simply found some pattern that absolutely destroys IronPython perf. If you’d like to help us track down that issue so we can fix it feel free to e-mail me.

]]>