Andy Lester wrote:
On Jul 4, 2005, at 3:03 PM, Nicholas Clark wrote:
On Mon, Jul 04, 2005 at 02:12:00PM -0500, Andy Lester wrote:
The attached patch cleans that up.
Thanks, applied (25079)
now, can someone with better measurement-fu tell us if it sped anything up?
xoxo,
Andy
No appreciable difference with my Real World[tm] tests, but as the
change only deals with destruction, it isn't particularly exercised in
this code. I guess the low-hanging fruit has already been picked.
I reran my hash performance code that I posted a while back. The short
results are as follows (load is lots of hash and array creations, reduce
is lots of hash lookups and minor modfifications, most extreme run-times
(fastest, slowest) of 22 runs discarded).
test perl____ -- load -- -- reduce --
file version_ min mean max min mean max
test.1 perl5.8.6 0.74 0.753 0.79 0.12 0.122 0.15
test.1 perl5.9.2 0.79 0.793 0.82 0.13 0.1325 0.14
test.1 perl5.9.3 0.88 0.8875 0.90 0.14 0.1515 0.18
test.2 perl5.8.6 14.69 14.9915 17.97 5.53 5.7305 8.21
test.2 perl5.9.2 15.98 16.146 17.10 11.54 11.6735 11.97
test.2 perl5.9.3 17.78 17.9195 18.09 6.81 6.8855 6.99
perl5.9.2 as released on CPAN and 5.9.3 is blead/DEVEL24148. Both
recompiled this morning with -Dusedevel -des.
previous run (unfortunately didn't test 5.9.2):
http://www.landgren.net/perl/blead/20050611.txt
this run:
http://www.landgren.net/perl/blead/20050705.txt
DAvid
