[Asterisk-Dev] Re: ztdummy accuracy improvements on kernel 2.6

[Tony Mountifield]

In article <200505160906.57042.akohlsmith-asterisk at benshaw.com>,
Andrew Kohlsmith <akohlsmith-asterisk at benshaw.com> wrote:
> On May 16, 2005 08:59 am, Tony Mountifield wrote:
> > I think what is happening is that the zaptel processing invoked by ztdummy
> > is not happening quite often enough due to missed jiffies. Consequently
> > I suspect the incoming IAX channels are not being serviced often enough,
> > and are building up a backlog.
> 
> If ztdummy's hitting it 1024 times a second instead of 1000 wouldn't that 
> indicate that it was servicing it too often?

My paragraph above was referring to the current ztdummy method in 2.6
which is to hook into the kernel's tick processing or every jiffy. It is
that which I think is missing slots.

The 1024Hz is the completely separate RTC interrupt used by zaprtc and now
by my modified ztdummy. By skipping 3 every 128 (1 every 42 or 43), it
evens out at 1000/sec.

Cheers
Tony
-- 
Tony Mountifield
Work: tony at softins.co.uk - http://www.softins.co.uk
Play: tony at mountifield.org - http://tony.mountifield.org