mm: optimisations

The following patchset against 2.6.15-rc2 contains optimisations to the
mm subsystem, mainly the page allocator. Single threaded write-fault
based allocation performance is improved ~5% on G5 with SMP kernel, and
~7% on P4 Xeon with SMP kernel (this measurement includes the full fault
path, page copy, unmapping, and page freeing, so actual kernel allocator
improvement should be larger).

Thanks to feedback from Christoph, Andi, and Bob Picco.

This patchset is cut down to include just straight optimisations and no
behavioural changes.

Nick

-- 
SUSE Labs, Novell Inc.


Send instant messages to your online friends http://au.messenger.yahoo.com