Forcing context switches (was: Ada Tasking problem)

[cis.ohio-state.edu!zaphod.mps.ohio-state.edu!think.com!hsdndev!encore!max zilla!jcallen@ucbvax.Berkeley.EDU (Jerry Callen)]

> [various discussion about using delays, especially delay 0.0, to force
>  context switches deleted]

The Encore parallel runtime ignores delays of 0.0. Of course, since this
runtime is for a multiprocessor, the issue of context switches is a bit more,
um, "interesting." The vanilla Unix implementation maps each Ada task to
(essentially) a kernel thread, so Unix handles the timeslicing. The realtime
implementation uses a combination of priority and runtime-established
timeslice parameters to do timeslicing.

It would be nice to have some guidance from AJPO or someplace
regarding what runtime tuning knobs are expected to be available.
But then, maybe that's what "market differentiation" is all about. :-)

-- Jerry Callen
   jcallen@world.std.com