Rich Burridge wrote:
Hi,
Looking for a code review for the fixes for
10485 move pkg(5) to Python 2.6
http://defect.opensolaris.org/bz/show_bug.cgi?id=10485
10482 upgrade to cherrypy 3.1.2
http://defect.opensolaris.org/bz/show_bug.cgi?id=10482
11836 shebang line for python modules should be python version-agnostic
http://defect.opensolaris.org/bz/show_bug.cgi?id=11836
11950 ldtp used by pkg build process not setup to easily use Python 2.6
http://defect.opensolaris.org/bz/show_bug.cgi?id=11950
11989 pkg python dependency analysis tests fail
http://defect.opensolaris.org/bz/show_bug.cgi?id=11989
Webrev is at:
http://cr.opensolaris.org/~richb/pkg-10485-v1/
So, one unfortunate side-effect of this upgrade seems to be a slight
loss of performance. I've timed various operations using the gate tip,
and then using the gate tip + your python 2.6 patch. On the plus side,
I saw an average of 2% reduction in memory usage.
============================
time pkg list -a > /dev/null
============================
python2.4:
real 0m4.174s
user 0m4.061s
sys 0m0.098s
python2.6:
real 0m5.196s
user 0m5.069s
sys 0m0.112s
============================
time pkg list -af > /dev/null
============================
python2.4:
real 0m5.174s
user 0m5.058s
sys 0m0.099s
69,392K peak pmap total
python2.6:
real 0m6.262s
user 0m6.135s
sys 0m0.111s
68,128K peak pmap total
============================
time pkg list -s > /dev/null
============================
python2.4:
real 0m2.518s
user 0m2.411s
sys 0m0.078s
30,056K peak pmap total
python2.6:
real 0m2.932s
user 0m2.811s
sys 0m0.092s
28,788K peak pmap total
=========================
time pkg list > /dev/null
=========================
python2.4:
real 0m0.328s
user 0m0.259s
sys 0m0.063s
python2.6:
real 0m0.399s
user 0m0.317s
sys 0m0.077s
==================================
time pkg info SUNWipkg > /dev/null
==================================
python2.4:
real 0m0.600s
user 0m0.487s
sys 0m0.098s
python2.6:
real 0m0.679s
user 0m0.553s
sys 0m0.109s
=========================================
time pkg contents -m SUNWipkg > /dev/null
=========================================
python2.4:
real 0m0.276s
user 0m0.209s
sys 0m0.063s
29,048K peak pmap total
python2.6:
real 0m0.336s
user 0m0.256s
sys 0m0.074s
27,532K peak pmap total
=====================================
time pkg image-update -nv > /dev/null
=====================================
python2.4:
real 0m15.439s
user 0m14.803s
sys 0m0.246s
196,780K peak pmap total
python2.6:
real 0m17.673s
user 0m17.021s
sys 0m0.262s
193,956K peak pmap total
=======================
time pkg refresh --full
=======================
python2.4:
real 0m58.922s
user 0m55.556s
sys 0m0.329s
python2.6:
real 1:11.8
user 1:08.0
sys 0.3
Comparing the dtrace output timings for the 2.4 vs the 2.6 version, it
looks like most of this can be attributed to fmri.py:__init__,
version.py:__init__, and version.py:dotsequence_val. For some reason,
those appear to be slower now.
Cheers,
--
Shawn Walker
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
