> While I haven't read the referenced paper, its title ("Fast Mutual Exclusion
> for Uniprocessors") does somewhat imply that it's only useful in the case
> where SWP is already good enough.

I have read the paper......
It is definitely only talking about uniprocessor systems.
In particular (effectively) stopping the kernel from process
switching within the critical region (by resuming at the start of it.)

This won't be any use at all for synchronisation on a MP system.

It should also be noted that 'Lamport's algorithm' will not work
for MP locking on many modern CPUs (eg sparc).

	David

-- 
David Laight: david@l8s.co.uk