[parisc-linux] Spinlock patch, take 2

[Randolph Chung]

> Looks beautiful. Is it measurably faster, or just semantically more
> correct than the existing implementation? :)

i have not done any measurements. the spinlock_t stuff is needed since
we've agreed that the attribute((aligned(16)) is not sufficient to
guarantee correct alignment. the atomic-lock optimization is based on
some discussions we had a few months ago. i'd like to actually try
these, but we still don't have a bootable SMP kernel yet :(

randolph
-- 
Randolph Chung
Debian GNU/Linux Developer, hppa/ia64 ports
http://www.tausq.org/