Clay
10-12-2005 04:58:51
The SWIG for Python maintainer has sent out an email to maintainers of large Python/SWIG projects. He has made major changes to SWIG cvs and wanted us to test them.
The good news is, he has made major improvements to the generated code, which have some speed improvments to many aspects of SWIG. The bad news is, PyOgre runs much slower on the new version of swig. Upwards of a 40% decrease in speed.
I have been talking to him in e-mails trying to figure out what the problem could be, but in the end we really need to profile the code to find the problem. I do not have linux set up (nor a hard drive to install it on right now).
So, I need someone's help. Basically we need to build a copy of PyOgre under linux with Swig 1.3.27 and with SWIG from CVS and do the following:
The good news is, he has made major improvements to the generated code, which have some speed improvments to many aspects of SWIG. The bad news is, PyOgre runs much slower on the new version of swig. Upwards of a 40% decrease in speed.
I have been talking to him in e-mails trying to figure out what the problem could be, but in the end we really need to profile the code to find the problem. I do not have linux set up (nor a hard drive to install it on right now).
So, I need someone's help. Basically we need to build a copy of PyOgre under linux with Swig 1.3.27 and with SWIG from CVS and do the following:
- Check to see if the slowdown still occurs.[/*:m]
- Profile PyOgre using gprof or valgrind and find the difference between the two libraries.[/*:m][/list:u]
This may require building Ogre with -pg and obviously building PyOgre with -pg. Does anyone have experiance in profiling that can help out here?
This could also be done under windows if someone has a profiler. I'd be happy to wrap up two seperate versions of pyogre for someone to use. I just don't have the money to buy a profiler for windows myself.
In the mean time I'll try to build this under cygwin and use gprof, but I'm not sure how successfull I will be.
Thanks.