[LinuxPPS] Odd behavior of TimeStamps.

[clemens at dwf.com]

This is not a problem, just odd, different that it used to be, and
I thought Id let other readers of this list think about it.

I plot the difference between the timestamp, and the clock time as a 
function of time.  If there is drift on startup, you will see it. If there are
other problems you will see them too.

In the past I have seen a smooth 'sinusoid' as the clock drifted back and
forth, that has some larger (point) delays in the timestamp, when the machine
is bussy.

In Figure 1 following, with a scale of +-40us you will see how the curve stays
within +-10us except for these outriders.

In Figure 2 following, with a scale of +-1ms, you will see that these outriders
are all almost exactly 750us (sorry about the single digits on the scale).

So my question to readers of this list is why?
Im going to assume that the fact that the outriders are now all the same size
is due to the IRQ patch.  That's a plus, but why 750us?
-- 
                                        Reg.Clemens
                                        reg at dwf.com

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Frame2.jpg
Type: image/jpeg
Size: 31255 bytes
Desc: Frame2.jpg
Url : http://ml.enneenne.com/pipermail/linuxpps/attachments/20080624/ea2a5d27/attachment-0002.jpg 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Frame1.jpg
Type: image/jpeg
Size: 38944 bytes
Desc: Frame1.jpg
Url : http://ml.enneenne.com/pipermail/linuxpps/attachments/20080624/ea2a5d27/attachment-0003.jpg