On Tue, 1 Mar 2011, Arnd Bergmann wrote:

On Tuesday 01 March 2011, Sascha Hauer wrote:

quoted

quoted

Taking one step back from this, have you considered making this
a regular interrupt controller? That would make the client drivers
more standard -- you could define the interrupt numbers as resources
of a platform device or in the device tree, for instance.
The cost might be more complex code, e.g. when a device requires
many interrupts, but I think it will be at least as efficient
at run-time, and less surprising for readers and authors of
client drivers.

I thought about this, but hesitated to increase NR_IRQS by 463. Do you
think we should do this instead?

I think there is a plan to virtualize the interrupt numbers on ARM,
and in that case NR_IRQS becomes rather meaningless. I don't know
exactly how far that effort has come.

Also sparse irqs allows us now to allocate beyond NR_IRQS. With sparse
irqs NR_IRQS is pretty meaningless and just gives us an indicator how
large the irq space might become, but we allow up to 8k dynamically
allocated irqs beyond NR_IRQS, so this should be sufficient for your
problem.

Thanks,

	tglx