Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 4 Mar 2003 16:41:05 -0500 (EST)
From:      Andrew Gallatin <gallatin@cs.duke.edu>
To:        raj@cup.hp.com
Cc:        ia64@FreeBSD.ORG
Subject:   Re: initial netperf tests on rx2600
Message-ID:  <15973.7537.476710.330402@grasshopper.cs.duke.edu>
In-Reply-To: <3E65180A.8B39FA19@hp.com>
References:  <3E6401FD.E2B384E3@hp.com> <20030304201227.GA523@athlon.pn.xcllnt.net> <3E65180A.8B39FA19@hp.com>

next in thread | previous in thread | raw e-mail | index | archive | help

Rick Jones writes:
 > > I think getloadavg(3) might be the one. I don't know anything about
 > > netperf though...
 > 
 > The manpage for getloadavg seems to return just that - the load average.
 > While that is a measure of load, it isn't the same as the CPU util.  On
 > HP-UX I use pstat() calls, on Linux /proc/stat, on Solaris kstat(). 
 > 

I think you want the kern.cp_time sysctl.  

Look at sys/dkstat.h (or sys/resource.h if your source is new enough).
And see the sysctl(3) man page.

This provides a whole-system measurement, as taken from the kernel's
statclock routine.  I hacked an older version of netperf to use this
on an older version of freebsd (where I had to grovel in /dev/kmem).

Drew

PS: I work for Myricom now.  I have the (empty) bag of Famous Amos
cookies you gave Feldy for the first netperf submission hanging in an
honored spot in my office. 

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ia64" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?15973.7537.476710.330402>