RE: [sv-bc] illegal priority if

>From: "Michael (Mac) McNamara" <mcnamara@cadence.com>

>Greg is right that it is only crazy cases where there is any order dependent 
effect in any case; but the corollary to that is the slow down would only be 
experienced in those crazy cases, and *gosh* those are the ones I want to spend 
simulation time on!

This assumes that the simulator can statically detect those crazy cases,
so that it doesn't slow down in other cases.  If it could reliably detect
those crazy cases, then we could make it illegal to have this kind of
order dependency, and the problem would go away.  But I don't think this
is practical to detect in the general case.  We could put restrictions in
that would prevent it from happening, but people might not be happy with
those restrictions.

And as far as being willing to spend extra simulation time on those cases:
we are talking about spending exponential amounts of time.  So I doubt you
really want to spend the simulation time on it in extreme cases.  If the
sun has burned out before you get your results back, they won't do you
much good.

BTW, defining a required evaluation order would avoid differences
between simulators and give predictable results.  However, it doesn't
really solve the issue between simulators and synthesizers.  Yes, the
synthesizer knows the order the simulator used.  But if the synthesizer
is required to follow that order, then I think it defeats the purpose
of "unique".  As I understand it, "unique" was supposed to mean pretty
much the same thing as "parallel_case".  It tells the synthesis tool
that it can evaluate these things in parallel.  If it has to evaluate
them sequentially to match side effects, then the purpose has been
defeated.

Note that any unique if/case that has an order dependency has inherently
violated this intent of unique, by not being computable in parallel.  So
it could be considered an illegal unique if/case.  The problem is how are
you going to detect this kind of order dependency in the general case?

Steven Sharp
sharp@cadence.com