Improving desktop application performance WAS Re: init observations
dmalcolm at redhat.com
Thu Nov 17 00:49:37 UTC 2005
On Wed, 2005-11-16 at 13:48 -0500, Dimi Paun wrote:
> From: "Stephen J. Smoogen" <smooge at gmail.com>
> > To elaborate.. there are factors beyond your CPU/memory. What kind of
> > diskdrive do you have? What is its transfer rates? What are it
> > hit/miss rates? How is the layout of the disk optimized for disk-seek?
> > A lot of commodity disk-drives are sub-par on these areas. This means
> > that programs will slow down because you are spending a lot of time
> > waiting for the disk-drive to give you bits because it keeps plugging
> > up its cache or not able to get various items off disk as fast as
> > advertised.
> I think this is missing the point -- we are not talking about some
> big database intensive application that needs special handling to
> work optimally. We are talking about the simplest application on
> the desktop, on a fairly beefy machine, running the most stock FC4
> out there.
> If I need to do _anything_ special to get gedit to start in a decent
> manner, something is clearly wrong. And BTW, try to start notepad
> in Windows. It's fast.
> I think that anything over 500ms for gedit is bad. Even on a 500MHz box.
> Granted, part of the problem may be in the application, but the user
> doesn't know (and frankly, doesn't care).
[changing Subject to better reflect this part of the thread]
There's lots of good work on improving exactly this kind of thing going
on at the moment in the upstream GNOME community (e.g. optimizing Pango,
the filechooser, session startup time etc).
If anyone with C coding skills is interested in helping out with this
effort, it'd be great if you could help out there, and we'll see the
improvements landing in FC5
The relevant mailing list is performance-list at gnome.org, where there's a
strong focus on good methodology i.e. measuring to determine exactly
what's really causing a problem, then trying a change, then measuring
the effects. Sometimes the proposed optimizations have turned out to
actually slow things down...
Some great resources are Lorenzo's Colitti's analysis page on speeding
See also Federico's blog:
Plus this page on the GNOME wiki:
Hope this helps
 with plenty of help from the #fedora-desktop gang :-)
More information about the devel