Here is a message from Morten whose large attachment is available for
download at http://homepages.ulb.ac.be/~arigo/psyco/mwpetersen.html.gz.
----- Original Message -----
From: "Morten W. Petersen" <morten@...>
Sent: Friday, March 01, 2002 8:19 PM
Subject: Psyco speed
> Hi all,
> I've played with psyco a bit, running a speed-test using the htmllib
> module and a large HTML-file (see attached files).
> htmllib.py runs the test function 100 times, htmllib2.py also runs the
> test function 100 times, but imports psyco and calls psyco.jit()
> The results are surprising (given that Psyco is expected to produce
> faster code):
> 'time python2.2 htmllib2.py' (with Psyco):
> real 4m57.797s
> user 4m53.960
> sys 0m2.040s
> 'time python2.2 htmllib.py' (without Psyco):
> real 4m36.409s
> user 4m35.420s
> sys 0m0.930s
> What is causing Psyco to be slower?
On Sat, Mar 02, 2002 at 05:59:59AM +0100, Armin Rigo wrote:
> Hello everybody,
> Here is a message from Morten whose large attachment is available for
> download at http://homepages.ulb.ac.be/~arigo/psyco/mwpetersen.html.gz.
And here's another go at it, without the huge HTML file attached this
Binding functions / classes is better than using the jit (as
it is noted in ISSUES.txt, albeit only for 'large applications');
binding two of the classes used in htmllib.py gave significantly
'time /usr/local/python2.1/bin/python2.1 htmllib2.py':
'time /usr/local/python2.1/bin/python2.1 htmllib.py':
So, psyco works, and it works quite well!
Get latest updates about Open Source Projects, Conferences and News.