Skip to Content.
Sympa Menu

sm-discuss - Re: [SM-Discuss] Benchmarking Spell?

sm-discuss AT lists.ibiblio.org

Subject: Public SourceMage Discussion List

List archive

Chronological Thread  
  • From: Andrew Stitt <afrayedknot AT thefrayedknot.armory.com>
  • To: "'sm-discuss'" <sm-discuss AT lists.ibiblio.org>
  • Subject: Re: [SM-Discuss] Benchmarking Spell?
  • Date: Sun, 4 Aug 2002 02:50:50 -0700

>
> Don't know how we could easily make it do the test after a reboot, at least
> not until we have standardised the startup/init stuff. Maybe the steps need
> to be more like:
>
> - if you find any of the files in the file systems delete them, and don't
> allow execution of test on this run.
> - if you don't find files, query user to see if they have rebooted
> machine to make sure disk caches will also be clean, if they answer yes then
> perform test.
>
the problem with doing just that is the disk cache. Basically once the
system accesses something from the (relatively) slow disk it gets thrown
into the (relatively) fast ram with the hope that we will access it again
soon. try running "time find / >dev/null" twice in short succession. The
first test will take a while for all the fs metadata to get read off
the disk, the second time it should all be cached at which point it will
run an order of magnitude faster. On my system the first test took about
36 seconds, then second 0.5 seconds. Thats a big difference especially
when compounded over 50 megs of files instead of at most a few megs of
fs metadata. So thats why reseting to the same state is preferable to
just making sure all the files are deleted from the system. I suppose
there might be a system call to completely flush the disk cache, but
there may be other variables involved that also can be just as easily
and more reliably reset by just a reboot.




Archive powered by MHonArc 2.6.24.

Top of Page