On Tue, Sep 30, 2008 at 04:54:16PM -0700, Dan Price wrote:
> I'm slightly shocked at the lack of speedup
/me is *very* sheepish.
Once again, I was pointing to the wrong PYTHONPATH when doing the
calculations.
The old code gives:
action creation
0.685171 29189 actions/sec
0.682373 29309 actions/sec
0.682449 29306 actions/sec
manifest contents loading
1.996338 500 manifest contents loads/sec (30055 actions/sec)
2.002437 499 manifest contents loads/sec (29963 actions/sec)
2.004785 498 manifest contents loads/sec (29928 actions/sec)
The new code gives:
action creation
0.323446 61834 actions/sec
0.322086 62095 actions/sec
0.322448 62025 actions/sec
manifest contents loading
1.191956 838 manifest contents loads/sec (50337 actions/sec)
1.191840 839 manifest contents loads/sec (50342 actions/sec)
1.192588 838 manifest contents loads/sec (50310 actions/sec)
so between a 67% increease and a 112% increase in speed. So a little less
shabby.
As for profiling, it looks like
libpython2.4.so.1.0`PyObject_Malloc+0xb3
libpython2.4.so.1.0`PyString_FromStringAndSize+0xf4
_actions.so`_fromstr+0x56f
is a pretty common stack, though coming from different points in
_fromstr(). Stacks containing add_to_attrs() (which does all the
dictionary insertion, list append, list test, etc) are pretty uncommon.
Baaaah.
Danek
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
