On Tue, Aug 16, 2011 at 09:19:12AM +0200, Guennadi Liakhovetski wrote:

On Tue, 16 Aug 2011, Simon Horman wrote:

quoted

This avoids the need to look up the masks each time an interrupt
is handled.

Yes, almost... But I think, we can use the mask-caches even more 
extensively. In your patch you actually hardly gain anything, you continue 
reading the mask register instead of using the cache. Namely:

Sure.