Subject: Accurate Mac clock == read from RTC?
To: None <port-mac68k@netbsd.org>
From: John <john@sixgirls.org>
List: port-mac68k
Date: 07/12/2001 00:33:49
> > With performance like that, you're no better off running "ntpd", than
> > running "ntpdate" periodically.
>
> Well, running ntp as a background process is (in my mind) better than
> starting a cron job to run ntpdate.  At least this way, you avoid all
> the process creation overhead.

ntpd runs niced -12 (otherwise known as nasty) and has accumulated 340
seconds of CPU time in 14 1/2 days of uptime on a 68060 with a reasonably
accurate clock.

If one is worried about performance, the process creation overhead isn't
likely to be as much of an issue as ntpd's fits over trying to get the
time right.

Another idea: how much overhead is there in reading the Mac's RTC? What if
the kernel were to use the RTC as it's timebase instead of the interrupt
counter? Or even if the RTC could be read periodically...

Time to look at the source...

John Klos