[Mono-dev] IronPython Performance
bfulg at pacbell.net
Fri Apr 21 19:01:54 EDT 2006
A while ago (nearly two years) an interesting paper was published by Jim Hugunin
(http://www.python.org/pycon/dc2004/papers/9/) crowing about the significant
speed advantage observed in this interpreter running on top of Microsoft's .NET
VM stack. I remember being surprised by these results, as Python has always
seemed fairly fast for an interpreted language.
I've been working on the Programming Language Shootout for a few years now, and
after growing tired of the repeated requests to include scores for IronPython I
finally added the implementation this week.
Comparison of IronPython (1.0 Beta 5) to Python 2.4.3  and IronPython (1.0 Beta 1)
to Python 2.4.2  do not match the results reported in the 2004 paper. In fact,
IronPython is consistenly 3 to 4 times slower (in some cases worse than that), and
seemed to suffer from recursion/stack limitations.
Since the shootout runs on a Debian Linux box, I was not able to use Microsoft's
CLR; instead, I was pleased to find that Mono successfully ran most of the
Shootout testing suite. The 1.0 Beta 1 version of IronPython was built using the
compiler chain that ships with Mono 126.96.36.199; the 1.0 Beta 5 version is an 'official' build
from the IronPython download site running on Mono 188.8.131.52.
I am aware of the following issues that might affect the results:
1) Mono vs. Microsoft's CLR.
2) Python 2.1 vs. Python 2.4
3) Changes in IronPython over the last two years.
In addition, there is a small startup cost paid by IronPython since the py_compile
libraries were not present in the IronPython beta, preventing me from precompiling
the *.py files to *.pyo. However, most of our benchmarks are now sufficiently long
that initial startup costs are less significant.
I thought these findings might be interesting to some of the Python, IronPython,
and Mono developers. Let the flames begin! ;-)
Current results for Python versus IronPython are posted here:
More information about the Mono-devel-list