Re: gettimeofday problem

[george anzinger]

"Richard B. Johnson" wrote:
> 
<snip> 
> I ran my program all night on another machine and it's still running.
> Neither of these machines are trying to sync with NIST. Machines that
> are running timing daemons that attempt to sync their clocks could, of
> course, have problems with time-jumps.
> 

Me thinks it is time to fix this NIST/NTP issue.  The
problem is that we are adjusting the wall clock every 1/HZ
tick instead of adjusting the 1/HZ tick AND the
interpolation constant.  What happens is (in the X86) is
that we assume that the conversion of TSC to usec is fixed
and exact as computed at boot time.  The time sync protocols
have a more "correct" story to tell.  We need to incorporate
this information into the TSC to usec conversion so that the
wall clock correction for times between 1/HZ ticks agrees
with what is done at the tick time.
> 

-- 
George Anzinger   [e-mail]
High-res-timers: 
http://sourceforge.net/projects/high-res-timers/
Real time sched:  http://sourceforge.net/projects/rtsched/
Preemption patch:
http://www.kernel.org/pub/linux/kernel/people/rml