Re: [RFC PATCH 1/3] Unified trace buffer

[Ingo Molnar]

* Jeremy Fitzhardinge <jeremy@xxxxxxxx> wrote:

> Steven Rostedt wrote:
> > On Fri, 26 Sep 2008, Ingo Molnar wrote:
> >   
> >> Firstly they need a low-frequency (10khz-100khz) shared clock line 
> >> across all CPUs. A single line - and since it's low frequency it could 
> >> be overlaid on some existing data line and filtered out. That works 
> >> across NUMA nodes as well and physics allows it to be nanosec accurate 
> >> up to dozens of meters or so.
> >>     
> >
> > Can this possibly be true? I mean, light travels only one foot every
> > nanosecond. Can it really keep nanosecond accuracy up to dozens of  meters 
> > away? 
> 
> Sure.  NTP keeps machines within 1ms (or better) of each other even 
> though the network latency is much higher and jittery.

yes. And there are radio telescope arrays that are synced up to do delta 
interferometry, over thousands of kilometers. Syncing up time over a few 
dozen meters is no challenge - and the reason for that ease is that 
physical time is neatly and uniformly broadcasted by nature in a pretty 
dependable way, at around 300 thousand kilometers per second.

the challenge is to make it cheap enough for commodity hw. I.e. no extra 
CPU pins or lines in critical parts of the board, no extra power, low 
transistor count, no impact on any critical path, short and reliable 
clock readout after powerup, etc. But that is quite possible too IMO, 
and the payback is very real.

[ OTOH, this is a world that still ships FreeDOS on many whitebox PC
  instead of putting Linux on it, so dont expect logic to prevail
  in all cases ;-) ]

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/