Latest update was trying to decide if keeping Russell's stem work was
worth the code complexity given the speedups I can get just moving to
pcre and using a precompiled data format. I'm thinking they aren't as
huge a win, but it's still clearly a win. My testing is:
start timer
for (i = 0; i < 100; i++)
initialize db;
look up one path;
clean up db
stop timer
So the stems code will help even more if you do more than 1 lookup per
db load. Results at the moment for that testing looks like:
$ ./glibc
13.390 seconds used by the processor.
$ ./glibc-stems
8.750 seconds used by the processor.
$ ./pcre
2.160 seconds used by the processor.
$ ./pcre-stems
2.040 seconds used by the processor.
$ ./pcre-mmap
0.230 seconds used by the processor.
$ ./pcre-mmap-stems
0.170 seconds used by the processor.
So we're talking about a x50 speedup here for my best case vs
something analogous to what we are doing today.
-Eric
--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to majordomo@xxxxxxxxxxxxx with
the words "unsubscribe selinux" without quotes as the message.
