SV-BC Separate Compilation discussion 7/15/03

Attendees
00000
76654
12012
53928
00000
33333
aaaaa Karen Pieper
-aaaa Randy Misustin
-a--a Dennis Brophy
----a Danny Jacobi
aaaaa Brad Pierce
aaa-a Dave Rich
---aa Francoise Martinolle
a-aa- Matt Maidment
a-a-- Tom Kiley
-a--- Johny Srouji
aa--- Jay Lawrence
a---- Arturo Salz
a---- Peter Flake

We started with a discussion of Karen's proposal for separate compilation
support.  We noted that a many or most Verilog users are used to putting
all of their Verilog code on one line for the tool.  For this usage model,
$root is not a real problem.  

For tools that allow true separate compilation, this idea that $root keeps
growing is a real issue.  How do you know when to replace pieces?  How do
you know when to start over, and so on.  We also noted that procedural
statements are allowed in root.  We took a straw poll to see if there was
agreement to remove them.  Here are the results:

Peter, Matt want to leave procedural stmts in $root
Dave, Jay, Karen, Tom, Arturo are for removing
Brad abstains

Peter indicates that this is useful as a debugging feature.  Peter would
be happy if it was a tool choice to support this feature.

We also discussed the impact of $root on co-linking with VHDL.  We agreed
that $root did not pose a risk to the VHDL interpretation, but that the
issues surrounding how many $roots there were and how they merged together
still existed.

We tried to take a straw poll about $root.  It was felt too difficult to
vote given the large design space for solving the problem, so several 
people voiced their opinions:

Jay: nothing that has to be elaborated should live between modules
     including classes with static variables, non-reentrant tasks and 
     functions.
Dave: within a compilation unit, there needs to be "local" global data
Matt: $root for holding static variables is good for early design, but he can
	only justify it for testbenches.  Namespaces would be ok as a 
	replacement.

Karen will put together 3 versions of the proposal and we can do a straw
poll.

The next meeting to discuss separate compilation support is scheduled for 9am
Pacific on Tuesday 7/29 at 9am.