Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 20 Jun 2001 21:40:48 +0200
From:      Poul-Henning Kamp <phk@critter.freebsd.dk>
To:        Brad Karp <bkarp@icsi.berkeley.edu>
Cc:        smp@FreeBSD.ORG
Subject:   Re: clock speedup on SMP boxes 
Message-ID:  <7314.993066048@critter>
In-Reply-To: Your message of "Wed, 20 Jun 2001 11:23:23 PDT." <200106201823.f5KINN321282@carp.aciri.org> 

next in thread | previous in thread | raw e-mail | index | archive | help
In message <200106201823.f5KINN321282@carp.aciri.org>, Brad Karp writes:
>There were several responses to my initial note that the time of day clock in 
>FreeBSD 4.3-RELEASE runs faster than real time (gaining up to an hour every 
>day, where the amount of clock speedup is in proportion to the load on the 
>CPUs) on my SMP boxes. In these responses, others mentioned they see similar 
>behavior on other hardware. I see this behavior on Intel STL2 and Tyan S2567 
>motherboards with dual 1 GHz P-IIIs installed.

I'm the author of our timekeeping ("timecounter") code,  I heard
about this problem first time today.

Unfortunately part of my lab is already packed down so I can't reproduce
this locally right now.

I don't have a good idea right now what causes this, but I have a number
of ideas I would like people to try out:

	0.  First off, kill ntpd/ntpdate-from-cron.

	1.  Try to quanitify the speedup at 100% CPU load with no
	    interrupts.  Basically run one copy of for each CPU
	    in the system:

			main()
			{
				for(;;);
			}

	2.  Try to quanitify the speedup at 1/NCPU load with no
	    interrupts.  Run just one copy of the above.

	3.  Increase kernel HZ to 500 and rerun #2.

	4.  Decrease kernel HZ to 50 and rerun #2.

Please include in your report the contents of /var/run/dmesg.boot

please email your data to timec@phk.freebsd.dk

I'll try to see if I can find out what's going on.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

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




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?7314.993066048>