dean talbot <[email protected]> wrote:

> ...i think i found a couple of things that could shave a few
> cpu-cycles off of the BLFS 'compressdoc' script...

Update: I decided to test the modified script on a full system that i
could afford to trash...

# time compressdoc.orig -b -F /usr/share/man
  real  8m45.433s
  user  4m21.641s
  sys   4m19.278s

# time compressdoc.new -b -F /usr/share/man
  real  4m59.964s
  user  2m25.992s
  sys   2m29.770s

...This was on a system with 10,300 items under /usr/share/man, of
which 400 are hard-links (mostly from 'man-pages-posix'). So at face
value, it would appear that the original script ran "find <matching
inodes>" *9,900* times more often than it needed to!

I had to add the '-f' switch to "rm $SCRIPTSTART" to get it to run, but
the big difference comes entirely from using 'stat' to ID hard-links
(there was a negligible improvement with the timestamp included, too).

Hope this information is useful - and if it's flawed (besides being run
on a non-LFS system), i'd appreciate having my errors pointed out to me.

Dean
-- 
http://linuxfromscratch.org/mailman/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to