RE: [sv-bc] RE: [sv-ec] That modport direction issue again

I agree the LRM is unclear about what input modport directions
means for net ports. 

> -----Original Message-----
> From: Gordon Vreugdenhil [mailto:gordonv@model.com] 
> Sent: Wednesday, October 17, 2007 9:55 AM
> To: Mark Hartoog
> Cc: Jonathan Bromley; sv-bc@eda.org; sv-ec@eda.org
> Subject: Re: [sv-bc] RE: [sv-ec] That modport direction issue again
> 
> 
> 
> Mark Hartoog wrote:
> >  
> > This is, however, something that is likely to cause problems for 
> > users. The simulator is using a last writer wins resolution 
> function, 
> > which may lead to unexpected behavior that will surprise users.
>  >
> > Since it is very expensive and a lot of overhead to do the single 
> > driver rule through virtual interfaces at run time, it 
> makes sense to 
> > exclude it for efficiency reasons, but when possible I 
> think we should 
> > not be adding blanket waivers of the single driver rule for 
> all test 
> > bench constructs.
> 
> Definitely -- this is something that tools likely want to try 
> to be helpful about but for which I don't think we can 
> realistically mandate an "accurate" answer in the LRM.
> 
> I agree that we need to be careful in exempting testbench 
> constructs from semantic rules.  There are very few 
> situations in which that would be desirable, but I think that 
> this situation is one of them.
> 
> 
> [...]
> > I think module are not allowed to write to variable input ports.
> > I'm not sure what other semantics you are thinking of.
> 
> Well, that is part of the question.  Are modport "inputs" 
> like port inputs in that they aren't really "ref"?  A module 
> isn't permitted to write to an input port variable because it 
> has a continuous driver so any sequential write conflicts.  
> But we've already said (I think) that the "output" is not 
> really a continuous assign.  Does that hold for "input" or 
> does "input" in fact behave continuously?
> 
> I wasn't very clear in my earlier comments -- I was thinking 
> about module *net* ports which collapse and are essentially 
> directionless in such cases.  Are modport "ref" semantics similar?
> 
> If modports for variables don't impose continuous assignment 
> semantics, then it should be immaterial what their direction 
> is stated to be, shouldn't it?
> 
> Gord.
> --
> --------------------------------------------------------------------
> Gordon Vreugdenhil                                503-685-0808
> Model Technology (Mentor Graphics)                gordonv@model.com
> 
> 

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.