The attached patch-in-progress removes the per-cpu statistics from
struct kernel_stat and puts them in a cpu_stat structure, one per cpu,
cacheline padded.  The data is still coolated and presented through
/proc/stat, but another file /proc/cpustat is also added.  The locking
is as nonexistant as it was with kernel_stat, but who cares, they're
just fuzzy stats to be eyeballed by system tuners :).

A tool for printing the cpu stats specifically can be found near:

	http://www.osdlab.org/sw_resources/cpustat/index.shtml

Its output is almost identical to solaris' mpstat.  

I'm not sure I like the macro use, but it shields the callers from the
union garbage.  We can easily also make a THIS_CPU_STAT_ADD() interface,
as some have hinted would be nice :)

Currently its mostly ( :) ) only collecting the stats that were
collected in kernel_stat.  I'd like to add more stats -- page faults,
syscalls, cross-cpu calls, etc.  I understand people not wanting more
live cachelines in the fast paths.  I can make CPU_CRITICAL_STAT defines
that are config-ed out..

comments?  If its ok I can whip up a patch that updates all the ports
use of ->irqs[] as well.

- z
[ heading out for lunch :) ]