Python 2.1 performance

[Michael Hudson]

Per Gummedal <p.g at figu.no> writes:

> Running pystone (average 3)
> AMD Duron 650 - win2000
> 
> 1.5.2 - 10605
> 2.0   - 10282
> 2.1   -  9482
> 
> Is pystone a special case or is there a general drop in performance ?

Both, probably.  pystone isn't particularly typical code - it accesses
globabl variables a lot, for one thing - but 2.1 probably is going to
be a little slower.  This is could be due to the rich comparisons, but
I doubt anyone really knows.

I had a whirl on this sort of thing a while back; you can see my
results here (on linux/x86):

http://mail.python.org/pipermail/python-dev/2001-January/012330.html

mal's pybench makes the CVS of about 3 months back about 3% slower
than Python 2.0; I *doubt* the released version was much different
(but as I said in a post yesterday - this is a hunch about
performance, and therefore probably wrong).

Can you try building using the optional memory allocator?  I don't
know how you do that on Windows, but it would be interesting.

Cheers,
M.

-- 
  I don't remember any dirty green trousers.
                                             -- Ian Jackson, ucam.chat