>>>>> "M" == Matt  <[EMAIL PROTECTED]> writes:

M> PCI has always had the ability to do shared IRQs, you are correct,
M> however, that is a bad idea for any real time application,
M> especially VoIP. If you have your NIC card and Digium TDM2400
M> sharing an IRQ, when you get a bunch of calls going over your NIC,
M> your Digium is going to start suffering. I've seen it happen,
M> Digium says that is what will happen, and it makes sense.

I know it happens, I know Digium says it happens, and I still insist
it doesn't make sense. All the Digium driver has to do when an
interrupt comes in and it turns out to not be from the card after all
is to say so. This should take perhaps 10 or 20 clock-cycles.

In the same way, if an interrupt comes in for the Digium card and the
Intel driver gets first dibs at it, that only delays the interrupt a
few cycles before the Digium driver gets it. It's not a problem.

Except it IS a problem because the Digium driver (or possibly the
hardware) is crap. Anyway, Sangoma's drivers/hardware don't have the
issue from what people have written. So there is no fundamental
problem with shared interrupts.

Anyway, it would be so much easier if one of the manufacturers got
their driver into the official kernel. That would probably make the
company I work for replace their cards. So far there seems to be
little chance of that, partly because of the lack of interest from
manufacturers, and partly because the code is very far from
kernel-quality.


/Benny


_______________________________________________
--Bandwidth and Colocation provided by Easynews.com --

asterisk-users mailing list
To UNSUBSCRIBE or update options visit:
   http://lists.digium.com/mailman/listinfo/asterisk-users

Reply via email to