[comp.windows.x] is xload insane?

mansfiel@sapphire.sdr.slb.COM (Niall Mansfield) (12/29/88)

What does xload's output mean?  Is the histogram printed showing load as a
percentage, or what?

If as a percentage, then my understanding from the xload manpage is that
xload will shown the appropriate no. of horizontal lines in the window to
accomodate the current value.

But no matter how heavily loaded my machine is, the load average never
exceeds three percent (yes 3%!).  I hacked out the code from the widget,
and indeed printing the value ( *loadavg = (double)temp/FSCALE; ) I see it
varies between about 2.98 and 2.3.

(X11R3, Sun 3/50, SunOS 3.4 and 4.0)

swick@ATHENA.MIT.EDU (Ralph R. Swick) (12/30/88)

> What does xload's output mean?

The Unix implementation is supposed to show the average number
of runnable processes, as reported by the operating system.
Other meters are possible (and probably preferable).

jef@HELIOS.EE.LBL.GOV (Jef Poskanzer) (12/30/88)

>Other meters are possible (and probably preferable).

Yes, I've always preferred xcpu, which shows the percentage of cpu time
being used.  No averaging is done.  It seems extremely silly to me that
xload graphs a one-minute average once per second...

Also, on some systems xload can actually cause the load average to oscillate
between N and N+1 with a period of a few minutes, as the once-a-second xload
process goes into and out of sync with the once-a-second kernel load average
computation.
---
Jef

bin@primate.wisc.edu (Brain in Neutral) (12/30/88)

From article <8812291854.AA24373@LYRE.MIT.EDU>, by swick@ATHENA.MIT.EDU (Ralph R. Swick):
>> What does xload's output mean?
> 
> The Unix implementation is supposed to show the average number
> of runnable processes, as reported by the operating system.
> Other meters are possible (and probably preferable).

I prefer xcpu (ratio of user+system+nice to user+nice+system+idle) as
a more useful index of machine activity.  It's pretty easy to hack xload
to show this instead.

Paul DuBois
dubois@primate.wisc.edu	rhesus!dubois
bin@primate.wisc.edu	rhesus!bin