[riot-notifications] [RIOT-OS/RIOT] tests/xtimer_usleep: recognize negative offset (#11493)

Kaspar Schleiser notifications at github.com
Tue May 7 15:56:55 CEST 2019


> So in a timesloted networking protocol if I say:

I think we need to keep the timescales in context. I'd expect my timer subsystem to quantize any real time interval to the to the *closest* hardware timer tick instead of always rounding up. That would make it at most +/- half a tick wrong, being spot on on average. Always rounding up would mean it is between 0 and a whole tick *late* (half a tick on average).

Your argumentation for a timeslot would shift the problem to the end of the timeslot (always rounding up might make the protocol exceed its timeslot every time).

In practice, we're talking +- 0-15us vs +0-30us for a 32kHz timer quantized to 1000000us. The better option here is to either work with the low-level ticks or chose a faster timer. Or chose one that corrects in software via spinning, but that is inefficient.

I think we agree that the timer peripheral itself must not sleep too little. Is that the issue here?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/RIOT-OS/RIOT/pull/11493#issuecomment-490090788
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.riot-os.org/pipermail/notifications/attachments/20190507/56461b21/attachment.html>


More information about the notifications mailing list