mm: performance improvements

The following patchset is a set of performance optimisations
for the mm subsystem. They mainly focus on the page allocator
because that is a very hot path for kbuild, which is my target
workload.

Performance improvements are not finely documented yet, so they
are not indented for merging yet. Also some rmap optimisations
that Hugh probably won't have time to ACK for a while.

However, a slightly older patchset was able to decrease kernel
residency by about 5% for UP, and 7.5% for SMP on a dual Xeon
doing kbuild.

-- 
SUSE Labs, Novell Inc.

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