Re: [gmime-devel] parser optimization for mmap streams

[Jeffrey Stedfast <fejj novell com>]

These runs were done with several mbox files, adding up to a little over
1 GB of data and after warming up the cache on my office workstation
(previous results were on my home machine with about 1/4-1/3 of the
data):

w/o mmap patch (aka current svn):

[fejj localhost tests]$ time ./test-mbox

real	0m19.888s
user	0m18.369s
sys	0m1.464s
[fejj localhost tests]$ time ./test-mbox

real	0m19.517s
user	0m18.057s
sys	0m1.456s
[fejj localhost tests]$ time ./test-mbox

real	0m19.586s
user	0m18.185s
sys	0m1.388s
[fejj localhost tests]$ time ./test-mbox

real	0m19.725s
user	0m18.293s
sys	0m1.436s



w/ mmap patch:

[fejj localhost tests]$ time ./test-mbox

real	0m19.828s
user	0m18.505s
sys	0m1.240s
[fejj localhost tests]$ time ./test-mbox

real	0m20.028s
user	0m18.609s
sys	0m1.316s
[fejj localhost tests]$ time ./test-mbox

real	0m19.636s
user	0m18.401s
sys	0m1.232s
[fejj localhost tests]$ time ./test-mbox

real	0m19.714s
user	0m18.461s
sys	0m1.216s



There doesn't appear to be any noticeable difference on something a bit
less antiquated when testing on a larger data-set, so I'll probably just
pass on committing the patch.

Jeff