LKM Timemark Driver

Achim Gratz Stromeko at
Tue Aug 28 17:45:32 UTC 2018

Gary E. Miller via devel writes:
> Yes, thus the need for the PPS timestamp AND the TM2 timestamp.

All that NTP needs is a pair of measurements that show the difference of
the local and external view of time, at roughly 1 second intervals.  PPS
is not required to get that sort of result.

> The time from clock_gettime() is very coarse.  Not up to our needs.
> On a Raspberry Pi 3B+ running 64 bit kernel the granularity is only
> 52 ns.

That doesn't really matter, it is still about three orders of magnitude
improvement over what we get today for the error that matters.  In other
news, even if it had more resolution it wouldn't really make a
difference until you've also clocked the GPIO synchronizer with
something higher than 19.2MHz, something that would require a change to
the VC4 firmaware.

> Yes, plus the interrupt latency that we are trying to determine.  Which
> is why the Timemark method subtracts the two.

We don't need to determine any interrupt latency when there's no such
latency involved because we don't go through an interrupt for doing the

+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

Factory and User Sound Singles for Waldorf Blofeld:

More information about the devel mailing list