Kernel scheduling algorithm and X.Org performance

[Alan Coopersmith]

Jim Gettys wrote:
> Yes, certainly the Sun work is interesting/applicable (don't let the
> date on it scare you).  But the drivers are a fundamental problem that
> needs fixing: right now, X looks like (is) a CPU pig, and gets scheduled
> accordingly.

Oh, I'm sure there's plenty of blame to go around, and that we'll need
multiple fixes in multiple areas.

One other thing I meant to note in my previous mail was that while this
isn't as noticable on a single-user 2Ghz desktop, I've gotten reports
that users on our hundred-user Sun Ray servers did notice a subjective
difference when this was broken and then fixed about two years ago.
Unfortunately, no one's done any objective or formal testing in years.

One further optimization we've discussed but never implemented for
multi-user systems is to use the X servers built in idle timer/screen
saver/dpms routines to drop the priority boost for idle users to give
a bigger share of resources to people actually at their terminals than
to those that have wandered off for lunch.   (Especially DPMS - after
all, who cares how fast you're drawing to a powered off monitor?)

-- 
	-Alan Coopersmith-           alan.coopersmith at sun.com
	 Sun Microsystems, Inc. - X Window System Engineering