Re: [pygame] IronPython for Game Dev?

[Anthony Baxter <anthony@xxxxxxxxxxxxxxxx>]

On Thursday 14 September 2006 21:47, Simon Wittber wrote:
> On 9/14/06, Richard Jones <richardjones@xxxxxxxxxxxxxxxx> wrote:
> > While this looks impressive, pystones is a little unrealistic (it's
> > basically the old dhrystone benchmark in Python code and tests pretty
> > much none of Python's language features). I'd be interested to see the
> > pybench* numbers for IronPython 1.0 vs CPython 2.5 if you could...
>
> It seems IP on mono won't run the pybench suite. Curses. I can't
> really have a go at fixing it... because there are no tracebacks! Egh.
> Looks like I've gotta play around a bit more...

pybench --debug is your friend (as the error message says). There's a couple 
of problems (I'll be fixing these in Python itself once 2.5 is done). You'll 
want to run 
ipy.exe ..../pybench.py --with-gc --with-syscheck

You'll also need to patch your platform.py file to make it understand 
IronPython's different sys.version string format - it's a fairly trivial
patch to that foul regexp at the top of the file.

Under Mono 1.1.17.1, I seem ironpython1.0 getting about a third of the 
performance of CPython 2.5. It's not consistently slower - some things are 
2-3 times faster, others are 20-30 times slower. If someone out there wants to 
point the Mono folks at pybench, they could probably use it as a nice 
indicator of things to fix. Many of the serious slowdowns are to do with 
string handling.

I'm currently playing with http://www.taoframework.com/Home under fepy 
(ironpython) - this provides .net/Mono bindings to a huge pile of graphics 
toolkits, including SDL and OpenGL. I'm hampered by my ignorance of raw SDL 
calls, but as far as I can see, it just works.