We're getting "the usual" timing results here: it's a cross-platform, cross-release crapshoot. Aiming at (just) a few percent really is worthless -- modern processors and OSes are too complex to out-think uniformly "in the small", and even if Python variability didn't contribute to the differences. [Tres] ....
But then again, the same is true for pystones:
$ python ~/projects/Zope-CVS/lib/python2.3/test/pystone.py Pystone(1.1) time for 50000 passes = 1.39 This machine benchmarks at 35971.2 pystones/second $ python ~/projects/Zope-CVS/lib/python2.3/test/pystone.py Pystone(1.1) time for 50000 passes = 1.15 This machine benchmarks at 43478.3 pystones/second ...
My favorite in this respect will always be Win98SE. Running pystone after a boot gave a figure almost exactly twice as large (i.e., faster) than running pystone a second, third, ... time. One way to get back the original speed was to reboot. The other way was to write a little Python program that systematically allocated all of RAM, until it died with MemoryErrror. It didn't matter how much RAM you had -- a dozen MB or hundreds, same thing. This was some kind of Satanic OS "reverse caching", I guess <0.5 wink>.