Comments on: 64k version of python is 4234x faster* than c-python! http://www.philhassey.com/blog/2007/12/22/64k-version-of-python-is-4234x-faster-than-c-python/ game dev blog Sun, 29 Aug 2010 18:58:00 -0600 http://wordpress.org/?v=2.8.6 hourly 1 By: Ycros http://www.philhassey.com/blog/2007/12/22/64k-version-of-python-is-4234x-faster-than-c-python/comment-page-1/#comment-433 Ycros Sun, 23 Dec 2007 07:34:07 +0000 http://www.philhassey.com/blog/2007/12/22/64k-version-of-python-is-4234x-faster-than-c-python/#comment-433 ctypes is actually really easy, the bad thing about ctypes is that you're going to a C-libraries ABI, not to its API, and this can be brittle. Using the C-Python API is much better, although it takes longer to use. You should also take a look at Pyrex, which lets you write very python-like code which gets compiled down to C - which might be what you want. Also, take a look at: stackless python, psyco, libs like numpy/scipy. And make sure to profile your apps before trying to fix a performance problem! Sometimes performance problems can be easily fixed without having to rewrite things in C. I don't think py3k is going to have massive performance improvements - it's more of a backwards-incompatible cleaning-things-up release with a few extra funky features, nothing on the scale of the rewrites that ruby 2 and perl 6 are. You could also try looking at other languages which might be faster but aren't as painful as low-level C, such as: OCaml, haskell, one of the many common-lisp implementations, hell maybe even D/Java/C# (though I consider these lower-level than the other languages I've named). I guess Python has an awesome amount of libraries though. ctypes is actually really easy, the bad thing about ctypes is that you’re going to a C-libraries ABI, not to its API, and this can be brittle. Using the C-Python API is much better, although it takes longer to use. You should also take a look at Pyrex, which lets you write very python-like code which gets compiled down to C – which might be what you want.

Also, take a look at: stackless python, psyco, libs like numpy/scipy.

And make sure to profile your apps before trying to fix a performance problem! Sometimes performance problems can be easily fixed without having to rewrite things in C.

I don’t think py3k is going to have massive performance improvements – it’s more of a backwards-incompatible cleaning-things-up release with a few extra funky features, nothing on the scale of the rewrites that ruby 2 and perl 6 are.

You could also try looking at other languages which might be faster but aren’t as painful as low-level C, such as: OCaml, haskell, one of the many common-lisp implementations, hell maybe even D/Java/C# (though I consider these lower-level than the other languages I’ve named). I guess Python has an awesome amount of libraries though.

]]>