A few disconnect follow-ups.  Somebody follow the dots!

* Pieter on #macdev said that:
  
    stat() can be a bit slow on MacOS X

    it has some caching for recent paths, but if you stat() too many
    paths, the cache is cleared before it can be reused

    the cache would be a few thousand items, but depends on the amount of
    memory a machine has

    you can try to increase sysctl kern.maxvnodes and see if that has any
    effect
  
  http://pastie.org/1063960 

* Florian tried darcs get, cp -R, cpio -p and cpio -p with Spotlight
  disabled:

   - darcs get: 16 min
   - cp -R:      5 min
   - cpio -p:    2 min
   - cpio -p     1 min (disabling spotlight)
 
  http://irclog.perlgeek.de/darcs/2010-07-27#i_2615872 

* Trent on #darcs mentioned a "hard link fiasco" on HFS+ and landof on
  #macdev said "hard links have a long history of poor implementation
  and performance issues on HFS+. It supposedly has improved, but I
  haven't been watching closely. "

  There's a tech note that seems to say that hardlinks are essentially
  implemented as symlinks would be
    http://developer.apple.com/mac/library/technotes/tn/tn1150.html#HardLinks

  Then again do the cpio results above indicate that hard links are not
  the whole story since it seems we can recursively link tolerably fast?

* Petr on #darcs overall observed that darcs on Windows is 20x slower
  than on Linux.  On MacOS X, it is 90x slower.

* Jason tried running darcs-benchmark on a ramdisk on his Mac.
  He found much better CPU utilisation, and also that doing a full
  darcs get of his darcs darcs repo went from 7 minutes to 15 seconds.


-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
For a faster response, please try +44 (0)1273 64 2905.

Attachment: signature.asc
Description: Digital signature

_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users

Reply via email to