[EMAIL PROTECTED] (Hal Murray) writes:
> >Last time when I looked at it, it had no support for kernel PPS
> >("hardpps()"). That's what I also had a long time before. I haven't checked
> >recently.
>
> I've never understood this area. Why do I want a hardpps in the kernel?
Do you want a user-space algrithm play with kernel time variables, and if so,
why?
>
> What I want in the kernel is support to capture the information
> accurately. That means something like reading the time when the
> PPS signal into the serial driver generates an interrupt.
Thats PPS API without kernel bind support.
>
> I don't see any rush to process that data. The ATOM/PPS driver,
> for example, processes a batch of 64 samples at a time. So as long
> as the driver can read the data out of the kernel before it gets
> overwritten, that's good enough.
You know that hardpps can work completely without a running NTP daemon? (Well,
mostly ;-)
>
> Maybe in the old days ntpd couldn't get enough CPU cycles on busy
> machines. Is that a serious problem today?
If nobody needs hardpps in the kernel, why do all the people complain about
the "alpha" PPSkit? You have PPS API there, but nothing else.
Regards,
Ulrich
_______________________________________________
questions mailing list
[email protected]
https://lists.ntp.isc.org/mailman/listinfo/questions
