from comp.lang.python.announce -- Helix encryption
This might be interesting to use as a pypy performance test, if we can get the python version acceptably fast. ------- Forwarded Message From: "James Makela" <jamesm249@comcast.net> Newsgroups: comp.lang.python.announce Subject: ANN: xhelix encryption/authentication module There was recently an article named "Helix: Fast Encryption and Authentication" by Niels Ferguson and Bruce Schneier, published in the Nov 2003 issue of Dr Dobbs Journal. In it was a module written in python to implement this encryption method. This module works very well, but is too slow to be used in any practical application. xhelix is a python C extension implementing Helix encryption and authentication. It is used in the same way as the Helix module descibed in the article, but runs many, many times faster. The project resides at: http://sourceforge.net/projects/xhelix/ The current version is 1.0 and is released under LGPL if you have any questions, mailto: jamesm249@users.sourceforge.net - -- http://mail.python.org/mailman/listinfo/python-announce-list Support the Python Software Foundation: http://www.python.org/psf/donations.html ------- End of Forwarded Message
Hello Laura, On Sun, Nov 16, 2003 at 06:40:51PM +0100, Laura Creighton wrote:
This might be interesting to use as a pypy performance test, if we can get the python version acceptably fast.
Could be. I tried Psyco on it, but it doesn't do better than 2x the plain Python speed. This is because this module does unsigned 32-bit arithmetic using longs. Well I could enhance Psyco to try and encode longs in a single unsigned 32-bit machine word if they fit... The result would certainly be quite fast. But anyway I guess that's not what you had in mind; we could try to genpyrex-ify the module. Would probably need some tweaking in the Python source, but could be interesting to try indeed. A bientot, Armin
Hello Laura, On Sun, Nov 16, 2003 at 06:40:51PM +0100, Laura Creighton wrote:
This might be interesting to use as a pypy performance test, if we can get the python version acceptably fast.
Could be. I tried Psyco on it, but it doesn't do better than 2x the plain Python speed. This is because this module does unsigned 32-bit arithmetic using longs. Well I could enhance Psyco to try and encode longs in a single unsigned 32-bit machine word if they fit... The result would certainly be quite fast. But anyway I guess that's not what you had in mind; we could try to genpyrex-ify the module. Would probably need some tweaking in the Python source, but could be interesting to try indeed. A bientot, Armin
participants (2)
-
Armin Rigo -
Laura Creighton