Re: Compilation question

From: Geoffrey.Coram <Geoffrey.Coram_at_.....>
Date: Wed Jan 04 2006 - 04:20:41 PST
Arpad -
At the risk of annoying you by telling you things that aren't
in the LRM ...

1364 (digital) Verilog doesn't have any idea of a dc sweep,
in particular parameter sweeps.

There is a proposal from Cadence somewhere in Mantis or the open
issues list for something called "dynamicparam" which would be
allowed to change during a dc sweep.  I'm against such a proposal,
because then by default "parameters" would not be allowed to
change, which would make life difficult for analog designers who
like to change everything as well as for characterization engineers
to vary all the parameters to get a transistor model to fit the data
(or else all the model-writers would have to go through their models
and change every "parameter" to "dynamicparam").

Digital Verilog understands parameters to be a sort of constant.
However, since you can have different values for different instances
(and even override those values in various ways), the simulators
must somehow take into account that parameters aren't constants
like "3.0" and thus optimization must be handled differently.

Verilog people often talk about "elaboration" which is a phase in
which all the parameter values are set (among other things).  As
noted in Kevin's and Jonathan's responses, it is left to the
simulator to decide if/when to compile, and thus what
optimizations happen when.

The analog simulators I've used with Verilog-A that compile
modules do allow the parameters to be changed and swept without
recompiling, just as they do for transistor models that are
built-in.

-Geoffrey






"Muranyi, Arpad" wrote:
> 
> Happy New Year to everyone!
> 
> I have a question about the compilation of modules.  Let's
> take an example which is a model for a resistor, in which
> the resistance value is parameterized.
> 
> 1)  Does this module have to be re-compiled for every value
> of the parameter value (in multiple instances in the same
> simulation, or in the same instance of a sweep simulation)?
...
> 
> 5)  Any related comments, cautions, suggestions?
Received on Wed Jan 4 04:20:47 2006

This archive was generated by hypermail 2.1.8 : Wed Jan 04 2006 - 04:20:55 PST