How to make your python run faster

[izekia]

Here in my blog I've posted some test results that may be helpful.

[orochimaru_sof]

nice

Thanks

[izekia]

not at all
if you'll find any interesting boosts in perfomance please share this)

[hacktor]

gud work
keep it up

[zeppelin55]

Here in my blog I've posted some test results that may be helpful.

ok so I redid with the tester part also with range so this is the results the g functions use xrange and the f use range. The first set uses range in tester and the 2nd uses xrange lastly I transformed the first function into a lambda function g5 = lambda : [x+2 for x in xrange(n)] For the faster algorhythms it doesn't matter but for the slower ones it does. The hypothesis is that with psyco range should work better than x and not without.

xrange psy&gc gc psyco clean
g1 0.04600 0.03900 0.03900 0.04000 sec/repeat in 10 repeats
g2 0.03900 0.03900 0.04100 0.03900 sec/repeat in 10 repeats
g2_1 0.23700 0.23500 0.23500 0.23400 sec/repeat in 10 repeats
g3 0.24100 0.24100 0.24600 0.24000 sec/repeat in 10 repeats
g4 0.03800 0.03800 0.03700 0.03800 sec/repeat in 10 repeats
0.04000 0.04000 0.04000 0.03900 sec/repeat in 10 repeats
range psy&gc gc psyco clean
f1 0.04100 0.04000 0.04000 0.04100 sec/repeat in 10 repeats
f2 0.04000 0.04100 0.04100 0.04100 sec/repeat in 10 repeats
f2_1 0.23300 0.23900 0.23700 0.23800 sec/repeat in 10 repeats
f3 0.25400 0.24400 0.24900 0.24600 sec/repeat in 10 repeats
f4 0.03900 0.03700 0.03800 0.03800 sec/repeat in 10 repeats
0.04100 0.04000 0.04100 0.04100 sec/repeat in 10 repeats

import Ztester
print 'xrange',compareFuncs([g1,g2,g2_1,g3,g4,g5]),'range',compareFuncs([f1,f2,f2_1,f3,f4,f5])
xrange psy&gc gc psyco clean
g1 0.04500 0.03900 0.03900 0.04000 sec/repeat in 10 repeats
g2 0.03900 0.04000 0.04000 0.04000 sec/repeat in 10 repeats
g2_1 0.23600 0.23500 0.23600 0.23600 sec/repeat in 10 repeats
g3 0.24100 0.24400 0.24400 0.24100 sec/repeat in 10 repeats
g4 0.03800 0.03800 0.03800 0.03800 sec/repeat in 10 repeats
0.04000 0.03900 0.04000 0.03900 sec/repeat in 10 repeats
range psy&gc gc psyco clean
f1 0.04100 0.04100 0.04000 0.04100 sec/repeat in 10 repeats
f2 0.04100 0.04000 0.04100 0.04100 sec/repeat in 10 repeats
f2_1 0.23400 0.23900 0.23200 0.23100 sec/repeat in 10 repeats
f3 0.26100 0.24100 0.24100 0.24300 sec/repeat in 10 repeats
f4 0.03700 0.03900 0.03800 0.03900 sec/repeat in 10 repeats
0.04000 0.04000 0.04100 0.04100 sec/repeat in 10 repeats

[crimson]

:mrgreen: Ah! Nice work, Izekia! :mrgreen:

Thank you very much for sharing that! I knew some of that stuff, I learned some new things, and I got reminded about some ideas that I had sort of forgotten about that I can now use to speed up some of my submissions.

Readers may also be interested to review:

wiki.python.org/moin/PythonSpeed/PerformanceTips

I believe the results Izekia shows correspond quite nicely with the recommendations in that article, but that article is without reference to psyco. Izekia does a nice job of showing some cases where some of those guidelines really don't apply when using psyco.

Excellent.

[disillusioned_sof]

I fed this: a = [7,4,5,3,6] to the following python code, and "reduce"
doesn't seem to work as I expected:

import sys
if __name__ == '__main__':
    a = map(float,sys.stdin.read().split())
    print a
    n = len(a)
    m = sum(a)/n
    s = map(lambda x: x-m, a)
    s = reduce(lambda x,y: x+(y**2),s)
    print s

instead of calculating the sum of squares (10 in this case)
it outputs just 8.

[hendrik]

[quote="disillusioned"]I fed this: a = [7,4,5,3,6] to the following python code, and "reduce"
doesn't seem to work as I expected:

import sys
if __name__ == '__main__':
    a = map(float,sys.stdin.read().split())
    print a
    n = len(a)
    m = sum(a)/n
    s = map(lambda x: x-m, a)
    s = reduce(lambda x,y: x+(y**2),s)
    print s

instead of calculating the sum of squares (10 in this case)
it outputs just 8.[/quote]
I think your post does not have anything to do with making Python faster.
Anyway I analysed your code briefly. Using the initialiser for reduce (3rd optional parameter) should solve your problem. Set it to 0. Also using integer is actually sufficient.

[disillusioned_sof]

Hei Hendrik, thanks for your reply.
Actually I posted my stuff here since couldn't find other suitable place, and didn't want
to start a new thread. Typing "reduce" in the search wouldn't yield anything connected with Python.
As to the actual content of your reply, I'll try it. Thanks again.