Is there any particular reason why the sparc64 locore sets `runtime'?
mi_switch() takes care of it right as the process resumes (i.e. upon
return from cpu_switch()), so setting it in locore is totally unnecessary,
and only complicates the modifications necessary to make the scheduler
aware of multiple CPUs.

Barring objections, I'm going to remove it (sparc64 is the only port that
does it).

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