Re: VerilogAMS Committee Meeting Minutes - 15 December 2003


Subject: Re: VerilogAMS Committee Meeting Minutes - 15 December 2003
From: Geoffrey.Coram (Geoffrey.Coram@analog.com)
Date: Wed Dec 17 2003 - 11:11:06 PST


Sorry about the previous blank message; hit "send" instead of "quote" ...

I wanted to address two issues here.

> * Strings are currently supported in SystemVerilog and 1364-2001
> and should probably be resued directly. Also its probably a better
> idea to sync up with SystemVerilog including all extensions of SV
> in one shot rather than one feature at a time.

There were two other extensions, variable initialization (at time of
declaration) and variable declarations in unnamed blocks. I agree
that it is reasonable to wait for a full sync up with SV for those
two, since the issues are mainly convenience.

On the issue of strings, however, I think this is really something
we need in the short term, because people will expect to see models
for "NMOS" and "PMOS" -- not for type=+1 or type=-1 (especially
since compact models are traditionally written for NMOS, so model
writers know type=-1 is a PMOS whereas a designer might expect the
minus sign to indicate NMOS.

> Section 1.4 - Output, operating point parameters
> * The requirement was to hide certain variables being visible outside.
> The proposal was to declare all the variables at the top level, and
> making only those variables visible.
> * In the current language standards, all variables are available to
> be exposed. However vendors hide internal variables due to performance
> reasons but LRM does allows visibility of all the variables.

A new thought on this would be: what if all parameters and
module-level variables with (*description="string"*) attributes
were made visible? This would not affect existing modules, and
I expect that
a) compact model writers will attach a description to things
they want to be visible (gm, cds, etc.), so it would be
redundant to also specify a second attribute that specifically
says (*visible="yes"*)
b) no one will attach a description attribute for something
they don't want visible -- one would instead use a //comment

-Geoffrey

-- 
Geoffrey J. Coram, Ph.D.    Senior CAD Engineer     
Analog Devices, Inc.        Geoffrey.Coram@analog.com 
804 Woburn St., MS-422,     Tel (781) 937-1924
Wilmington, MA 01887        Fax (781) 937-1014



This archive was generated by hypermail 2b28 : Wed Dec 17 2003 - 11:11:27 PST