[Mono-dev] Benchmarking IronPython in Mono & .NET (and performance related stuff)

[Miguel Sousa Filipe]

Hi there,

I've just stumbled across this benchmarks that I believe are usefull
to pinpoint areas in the mono compiler/JIT that could improve.

The scenario is running PyBench in IronPython 1.1 vs CPython 2.5

Iron Python is ran on top of mono, or the MS .NET framework.

IronPython 1.1 ontop of Mono 1.2.3.1:
 http://sparcs.kaist.ac.kr/~tinuviel/pybench/
I believe this benchmarks were done on Linux.

IronPython 1.1 ontop of MS .NET (post by the IronPython main developer
- Jim Hugunin):
 http://lists.ironpython.com/pipermail/users-ironpython.com/2007-April/004773.html

A good analisys of the results is found in this email post:
 http://lists.ironpython.com/pipermail/users-ironpython.com/2007-April/004779.html

A better comparison would be to run all tests in a windows box, with
mono and .NET side by side..


Also related to the Mono compiler/JIT performance I have a couple questions:
- How is the new Compacting + Moving Garbage Collector going?

- Any short term remedies to those 9 methods that proved to be "at
least 3 times slower in mono than MS .NET" ?

- I've read in the mono "meeting" presentations of Q4 of 2006, some
remarks about work on several parts of the compiler to improve even
more its performance, how is that comming along?

- What is the status of the x86-64 compiler, how does it compare to the x86 JIT?
(i've been tracking commits in http://cia.vc/stats/project/Mono but I
haven seen much in the compiler/JIT+GC area...)

- I've read that a lot of effort went to make string operations quite
fast in Python, with a lot of algorithmic improvements, using state of
the art techniques and algorithms - how can mono take advantage &
integrate that kind of work that was done on CPython?

Keep up the good work!

Best regards, from a Mono fan!


--
Miguel Sousa Filipe