RE: [sv-ac] call to vote on 2110

Hi Manisha-- some responses to your points.


1. I am not convinced about the motivation behind this.

It's actually the same motivation behind 1995 (assertions in loops):  if
some type of model activity is taking place in a procedural loop, then
in order to have the checking close to the logic, we need to allow the
checker in a loop.  Otherwise the checker would need to be placed in a
remote generate block, the same problem with assertions before 1995. 


2. This proposal only talks about the loop aspect for assertions in
checker. But what about other things in checkers like initial_check,
always_check, final blocks etc. Do they get multiplied based on the loop
? If a checker is in a loop then everything inside should be part of
loop and not just the assertions.

Unfortunately, since we are not allowed to rely on synthesizability of
constructs, as you may recall from our long debates about the original
version of 1995, replicating a structural element like a checker in a
procedural context is very controversial.  Thus we have the limited
proposal you see in the current draft of 2110:  the assertions within
the checker may use loop-dependent elements, since we understand how to
handle that as described in 1995, but anything other than assertion
statements is *not* replicated, and must be identical for each iteration
of the loop.   (I would like to replicate everything, but a proposal
that does that in a way acceptable to all committees would not be
possible in our current time frame.)   This will form the starting point
of a more complete feature, probably in the next SVA standards
iteration, that will replicate all modeling code as you suggest.


 


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