RE: [sv-ac] subroutines attached to sequences

Hi Doron,

Actually in retrospect, when I raised lvars my intent was to analyze why
this issue has percolated up for subroutines and was not apparent before
for local vars. I think it's because the subroutine call -- assume for
now it is $display, argument applies to others-- intrudes on the
subsequence and forces visibility inside whereas before we were content
in treating the lvar assignments "local" to thread path and only
referencing after join -- effectively leaving the visibility function to
a) evaluation (decides to shortcut or not may be based on user input
...), b) debug keeps track of data (optionally glitching) within (if
recorded). It really did not matter to "outside" ... Data could only
propagate outward after the join.

That said in my opinion we should keep LRM flexible and adopt the loose
approach -- meaning treat this action loosely just like we effectively
did for lvars (also when LRM describes composition operators). The tight
interpretation can be mimic-ed by user putting the action after the join
not inside. Moreover, the loose approach of course minimizes context
dependence of sub-evaluation -- good for both evaluation and also user
analysis/debug -- and still maintains users' ability to model all
situations since it's more general.

** So in my mind, putting an action inside is requesting visibility
inside and making this sub-evaluation observable to the outside ...
rather *preventing* evaluation from shorting. Now, beyond a $display, a
truly actionable task is also better done for every sub-match and not
shorted ... If users wished to do a shortcut (aka tight) they can store
the data within sub-evaluation and add a "join action" after the join
(aka composition).

We can decide to add some pragma/keyword/operator (variants) later but
first we should agree on the default underlying interpretation in LRM --
I argue loose. BTW we should also recommend that users put a %m in
$display raising an issue to tackle -- scoping inside assertion ...

Thx.
-Bassam.


--
Dr. Bassam Tabbara
Synopsys, Inc.
(650) 584-1973

-----Original Message-----
From: owner-sv-ac@eda.org [mailto:owner-sv-ac@eda.org] On Behalf Of
dbustan
Sent: Saturday, February 11, 2006 1:11 PM
To: sv-ac@eda.org
Subject: [sv-ac] subroutines attached to sequences

Here are two examples with local variables, I don't see where the
problem is.