On Sat, 2006-09-09 at 13:51 -0500, James Bottomley wrote:
> We have had a specific problem with the current generic linux IRQ code
> for ages, in that indirection via function pointers is horribly
> expensive for us (basically, the CPU pipes stall).  We'd like to use the
> new genirq infrastructure to mitigate some of the problem.
> 
> I didn't want simply to move the irq handlers back into our arch code
> again, since that would remove all the benefits of common handling code.
> What I did was template out the common code in a way that keeps it
> common but allows an architecture to modify it (for us to put our acks
> and eoi's in as functions instead of function pointers).
> 
> If everyone is OK with this, I'll introduce a new type of interrupt with
> a specific handler (so for our heavily called interrupts like timer and
> IPI we don't even need to indirect through action->handler()).

I don't have any objection to this patch as long as the current
behaviour with function pointers still works fine (which seems to be the
case from a quick look through the patch). On PowerPC, we have a huge
variety of interrupt controllers and cascaded controllers and thus we
can't really apply your shortcut.

Ben.


-
To unsubscribe from this list: send the line "unsubscribe linux-arch" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to