Re: pagefault scalability patches

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Thu, 2005-08-18 at 09:17 -0700, Christoph Lameter wrote:
> On Thu, 18 Aug 2005, Nick Piggin wrote:
> 
> > Nick Piggin wrote:
> > 
> > > If the big ticket item is taking the ptl out of the anonymous fault
> > > path, then we probably should forget my stuff
> > 
> > ( for now :) )
> 
> I think we can gradually work atomic operations into various code paths 
> where this will be advantageous and your work may be a very important base 
> to get there.

Don't forget however that when doing things like tearing down page
tables, it's a lot more efficient to take 1 lock, then do a bunch of
things non-atomically, then drop that lock.

At least on PPC, the cost of a lock is approx. equivalent to the cost of
an atomic, and is measurable on such things.

That said, I think your approach for the anonymous page case is a good
first step for now. I'll have to adapt ppc64 to it but it shouldn't be
too hard.

Ben.


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>