Subject: Re: [sv-bc] Proposal for extern modules
From: Karen Pieper (Karen.Pieper@synopsys.com)
Date: Tue Feb 18 2003 - 14:21:02 PST
Steven,
What if the "extern"ed module comes into a tool in another format, like a
black box,
.db, .lib, or through the netlister, or so on. If the user also creates a
dummy so that
the RTL engine sees it, which one gets linked in? The dummy or the black box?
If it turns out to be the dummy by accident, the user will be upset. But
then, in
another situation someone will want the dummy, and to have the black box
ignored,
so how can the tool decide?
What if a dummy is used, and the tool decides to do an auto-ungroup on the down
design, replacing it in the top design with an empty design (rather than
recognizing
that it can't because the down design is unlinked?) The down design
instantiation
will correctly disappear in the up design. Since the down design
description exists
for the tool, it should be safe to ungroup that logic into the upper module
if the
timing or area is better met that way.
If the user doesn't know what the downside is, but does know the ports, it
is better
if the user just says that.....
K
At 02:28 PM 2/18/03 -0500, you wrote:
>Karen,
>
>Like Adam, I don't understand why you can't use the actual module
>declarations.
>They are apparently available, since you mentioned using scripts to create
>external declarations from them. And you proposed a .* syntax for the module
>declaration to make it use the port list from the external declaration, again
>implying that both were available.
>
>If there is some situation where the real module declaration isn't available,
>you could always use a dummy module (AKA shell or stub). They look pretty
>much like your extern declarations; just take the word "extern" off the front
>and add the word "endmodule" to the end. And if the synthesis compiler needs
>to know that these are dummy declarations, you can put the "extern" back on
>as an attribute. Also, configs could be used to compile all of the stub
>modules into a separate library if desired.
>
>Steven Sharp
>sharp@cadence.com
This archive was generated by hypermail 2b28 : Tue Feb 18 2003 - 14:21:53 PST