tasking priorities

[murphy@beatnix.UUCP (Michael Murphy)]

There recently has been much discussion about preemptive schedulers and
the interpretation of LRM 9.8(4).  I think I understand what it means for
the normal uniprocessor machine, but I'm not clear on its interpretation
for multiprocessor and distributed machines.  AI-00032 notes that:
   (Note:  the phrase "could sensibly be executed" refers to situations in which
   the  high  priority  task  can  actually  make use of the processor and other
   resources being used  by  the  lower  priority  task.   In  some  distributed
   processing  situations,  a  high  priority task may not be able to execute on
   some processors.  Preemption is only required for  processing  resources  the
   high priority task can use.) 

What if the processor the high priority task is currently located on cannot
run the task, but there is another processor "nearby" that could run it?
My question is, must the new high-priority task migrate to the 
processor (able to run the new task) where the lowest priority task 
is running (which would mean polling all processors on the system)?  
It would be much more efficient and would appear to be in the "spirit" 
of tasking priorities to only preempt the tasks running on your local machine.

Comments?  Anybody have a definitive ruling on this?
Note that this question can be applied separately (possibly with different
answers) to both distributed and multiprocessing machines.

-- michael murphy
-- UUCP: {uunet|sun}!elxsi!elk!murphy
-- AT&T: 408-942-0900