OK, very very quick benchmark on the realpath stuff
System used is an AMD Opteron 142 (1.6ghz) with a gig of pc2100 ram, and
Western Digital Raptor sata disk. It's running Debian/unstable, though
only in 32bit mode atm.
Both tests were run using the standard php5 release rather than current
CVS.
test.php just runs a passthru() 1000 times on another script which
includes 5 files.
-
before realpath modifications
root@mesa:/# time php /home/gaz/test.php
real 4m50.380s
user 1m23.635s
sys 1m22.761s -
after realpath modifications
root@mesa:/# time php /home/gaz/test.php
real 3m12.067s
user 1m22.602s
sys 1m24.608s
Now, imo, shaving over a minute and a half off that original time is
bloody impressive. Nice work, Andi.
OK, very very quick benchmark on the realpath stuff
Just as a bit of a follow-on......
- after realpath modifications
root@mesa:/# time php /home/gaz/test.php
real 3m12.067s
user 1m22.602s
sys 1m24.608s
The above was all done on absolutle paths. I've since tried a seperate
test with the same number of includes(same files infact) - but this time
with the files in php's search path and called them via include
"dir/file.php"; much like the typical PEAR usage. This one totally
surprised me.
root@mesa:/# time php /home/gaz/test.php
real 3m12.264s
user 1m39.049s
sys 1m23.718s
Virtually identical. Wow. Ok, I'm seriously impressed with this patch.