Making faster scientific calculations...

dsign dignor.sign at gmail.com
Wed Mar 21 17:52:27 CET 2007


 Ok, this is my first post to this list and don't know if it's the right
one.
I'm currently making sort of a scientific application. Most data structures
and algorithmic knots are in C++, and from there I have a set of extension
modules to Python. So, users of the app don't have to compile or ever see
C++ code to personalize the app to their ends, except for the inner
mathematical part. Regarding the last statement, I saw a few messages in
this list about Psyco, a JIT for Python and things like that. Well, I need
all the speed the machine can deliver in the mathematica l spot (I know by
profiling that it's is criticall), so, nothing of that can do the work.
Recently I read something about generating mathematical kernels in Python
for this kind of problems (can't locate the paper right now), the idea is to
just generate machine code and have it to do the work, but it wasn't for x86
based architectures (yes, portability is always a nightmare regarding tools
which generate native code). Do somebody know about a parallel effort for
x86? If there isn't such a thing out there, could it be of some interest to
people which is working numbers in Python?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-list/attachments/20070321/cfa2fa1c/attachment.html>


More information about the Python-list mailing list