Subject: Re: [sv-bc] attributes on SV constructs and beyond
From: Steven Sharp (sharp@cadence.com)
Date: Mon Nov 03 2003 - 15:44:34 PST
>Due to my previous claims, I think that the use of attributes should not
>be limited to constructs that are viewable by the PLI routines, maybe
>some language should be added to the PLI section in the SV/V2K standard
>specifying which attributes are viewable by the PLI routines and which
>are not.
I have a different view of this. The major advantage of attributes over
the older comment-based pragmas is that their syntax and some portions
of their semantics that are non-tool-specific are defined. This gives
them some consistency across tools.
The PLI interface to attribute values provides a specification of the
non-tool-specific portions of the semantics. It acts as a sort of
theoretical model or reference implementation for what information can
be extracted from the attributes by a tool. It is difficult to see
how you could specify it better.
On a more practical level, PLI can be used as an interface to the design
for minor tools that can't justify having their own Verilog front end.
Such tools may need access to all the same attribute information that
other tools have. For example, a synthesis lint tool might want to see all
of the synthesis attributes to check them. Adding attributes that are
not visible to PLI could render such tools impractical. I believe that
PLI should be powerful enough that any tool could be implemented in it,
in order to ensure that new tools we haven't thought of can be implemented
in it.
This doesn't mean that attributes can't be added to new constructs. It
means that if some tool needs attribute information about a construct, then
PLI probably needs access to that construct too. Not necessarily just to
get the attributes either. If the construct is that important to a tool,
then PLI applications may need to know about it too.
Steven Sharp
sharp@cadence.com
This archive was generated by hypermail 2b28 : Mon Nov 03 2003 - 15:46:27 PST