Re: Hierarchical name ordering

On Jan 19, 2008 1:26 AM, al davis <[email protected]> wrote:
> As you might have observed, gnucap hierarchical names have the
> innermost name first, moving out.  It seems that most other
> simulators do the opposite.
[...]
> It has been suggested to change it, and I agree.
>
> Is it worth the effort and possible confusion to have a "option"
> so it can go either way?
>
> Is it worth the effort and possible confusion to make the search
> (used by probe and delete) search both ways?
>
> The coding effort is trivial.  The "effort" I refer to here is
> from a user perspective.  extra documentation, etc.
>
> It seems to me .. the only reason to keep the existing reverse
> ordering at all (even by option) is for backward compatibility.

Although consistency with other simulators is "nice to have" I don't
think it makes much difference for a gnucap user. It's just a
convention after all.

If it's simple to add and maintain, I would prefer an option (can be
hidden). But then, with so many other changes, I wouldn't bother with
backward compatibility much. In fact, the only compatibility I need is
compatibility of models (that often include hspice/spectre/veriloga
code). At testbench level anything is OK as long as it makes my job
done.

Personally, I would prefer some blocking issues (like probing voltages
of internal nets) to be resolved as well.

BTW, can you post a sample code that:
a) replaces a nested cell with a different one,
b) replaces an instance parameter of such a cell.

From the documentation, it seems like this is a supported feature of
Gnucap but I could have never figured out how to do it in practice.

-r.