[Asterisk-Users] Generic X100P's

[Peter Svensson]

On Tue, 12 Oct 2004, Steven Critchfield wrote:

> On Tue, 2004-10-12 at 17:44 +0200, Peter Svensson wrote:
> > This is not correct. Each pci slot has four physical interrupt lines, A-D. 
> > The implementation is free to supply four separate interrupt lines to each 
> > card, i.e. the interupt lines are different for each slot on the bus.
> 
> Unless some PCI chipset is now dolling out 4 IRQ lines per slot, they
> have to merge somewhere and at that merge, they become shared. 
> 
> http://old.lwn.net/1998/0205/io-apic.html

Well, the poster a posts back said that some PowerPC machines had 
individual interrupts for the cards. I am not familiar with those, but 
some Alpha mchines used individual interrut lines. It was not uncommon to 
have a few hundred interrupt lines. Usually you only have a few pci slots 
per pci bus bridge. Behind the bridge some other technology (packets 
perhaps) can be used. 

Some chipsets use a new physical interrupt line per slot, so that no slots 
share the A interrupt. 

> > Does anyone have any hard numbers on the interrupt latencies tolerated by 
> > the digium cards? I would assume 100-200us to be acceptable. 
> 
> The facts about Zapata hardware is that there is no buffering. If you
> miss an interupt, you miss data. Interupts are fired 1000 times a second
> or as soon as 8 bits are collected on the 8000hz phone lines. 

That is pretty much what I said. The lack of buffering is a feature - 
latency is the enemy in telecomunications. That is why I said 100-200us 
should be acceptable. This should be redily aceivable, even with shared 
interrupts.

Another option would be to handle _all_ 1kHz telephony cards from one 
interrupt, i.e. all zaptel cards are handled in the interrupt hanlder of 
just one of the cards. There is no reason why having two X100P on the same 
irq is an insurmountable obstacle.

Peter