Date: Sun, 06 Sep 1998 14:37:47 -0700 From: Mike Smith <mike@smith.net.au> To: Thomas Dean <tomdean@ix.netcom.com> Cc: current@FreeBSD.ORG Subject: Re: Dmesg Timecounter Varies Message-ID: <199809062137.OAA10133@word.smith.net.au> In-Reply-To: Your message of "Sat, 05 Sep 1998 08:14:15 PDT." <199809051514.IAA00899@ix.netcom.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> I have noticed that the Timecounter value varies greatly in DMESG. > > I am running an SMP kernel. I extracted the values from > /var/log/messages and passed them thru a stat process: > > Timecounter Cost Statistics > Processing /var/log/messages and /var/log/messages.?.gz > > Frequency: 1193182 Hz - Always this value, unless stated. > 1/freq Time: 838 nsec > No. Cycles: 3 - floor( mean/(1/freqTime) ). > Cycle Time: 948 nsec - mean/(no. cycles). > Overhead: 332 nsec - Total overhead in Timecounter cost. > > Entries: 125 - Number of Timecounter entries in messages. > Minimum: 2517 nsec > Maximum: 41603 nsec - Wow! Flyer! > Mean: 2846 nsec > Std. Dev.: 3494.46 > > Notice the maximum flyer. What causes this variability? Unexpected execution discontinuities; most often SMI activity. We are going to have to discontinue our (ab)use of the TSC simply because it's not reliable enough. PC timekeeping *sucks*. -- \\ Sometimes you're ahead, \\ Mike Smith \\ sometimes you're behind. \\ mike@smith.net.au \\ The race is long, and in the \\ msmith@freebsd.org \\ end it's only with yourself. \\ msmith@cdrom.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199809062137.OAA10133>