Folks...

You may have noticed I've been cleaning up the pmap module over the
last couple of days.  The net result of this should be a slight to
moderate performance improvement, since a fairly expensive routine
was deleted.  Chris Gilbert has promised to run his standard speed
tests for me.

I've also taken a fairly expensive debug check out of the prefetch
abort path of non-SA110 CPUs, and this may improve things, as well.

I've tested all of this on a Shark with an SA110.  Please file a
bug report with send-pr(1) if you have any trouble.

"More to come."

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>