Skip to Content.
Sympa Menu

sm-discuss - Re: [SM-Discuss] Smallest Source Mage

sm-discuss AT lists.ibiblio.org

Subject: Public SourceMage Discussion List

List archive

Chronological Thread  
  • From: "Chris Brien" <christopher_brien AT hotmail.com>
  • To: sandalle AT hellhound.homeip.net, sm-discuss AT lists.ibiblio.org
  • Subject: Re: [SM-Discuss] Smallest Source Mage
  • Date: Tue, 30 Jul 2002 01:16:21 +0100

From: "Eric Sandall" <sandalle AT hellhound.homeip.net>
Somehow, we always come back to tmpfs lately. How 'bout we just close
this door since the patch is there and have Anders or Nathan (or Chris,
someone) put this into sorcery so we can end this. Those of you whom do
not want tmpfs, fine, you can now disable tmpfs usage in sorcery, those
who do want tmpfs, also fine, just do nothing. :)

I committed the patch today, and so the option will be available for the next release of sorcery (whenever Anders gets back from whatever oblivion he's wandered into).

Since the option now exists, and this is a non-issue, I'd like to put forward the argument that it is neither for speed or for fragmentation reasons. But simply the fact that it is much easier to clean up afterwards.

To delete a tmpfs directory, all the function rm_source_dir has to do is umount the dir, and then rmdir. The umount is instantaneous (well, as fast as it is possible to mark up to a gigabyte of physical memory as free), and the rmdir is safe, since it will only delete empty directories. On the other hand, a real directory on a real filesystem has to be physically deleted from the disk. This means an rm -rf, which is usually not the most pleasant command to even think about doing.
Tmpfs is safe.

Next time someone attempts benchmarks, I'd like to see wall time. Not as given by time(1), but as given by a good old fashioned mechanical clock. One which is not subject to innacuracies as interrupts cause the timer to block, or where the timings are split between system and user and whatnot. What matters here is real world time.

Saying that, I can't understand why anyone would think that a physical disk can be faster than ram. Still, it has to be proven to make some people see reality. Or, if you prefer, you can just ask Linus which will be faster.

Chris
aka zero_the_hero

_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com





Archive powered by MHonArc 2.6.24.

Top of Page