RE: [sv-ac] mantis 1550

Ed,
Arturo and Stu raise a good point: does it even make sense to allow
sampled functions to generate an event? $time does not generate event
when the time changes. The reason for $time's behavior has more to do
with historical Verilog-XL and the fact that it only created events when
the inputs to a function changed, not based on when the result of a
function changed.

Using that definition, $sample(x) would only be evaluated when the raw
signal 'x' changes, and an update event would only be generated if the
return value was different from the last evaluation.


Dave


> -----Original Message-----
> From: owner-sv-ac@server.eda.org [mailto:owner-sv-ac@server.eda.org]
On
> Behalf Of Eduard Cerny
> Sent: Thursday, November 02, 2006 12:12 PM
> To: stuart@sutherland-hdl.com; sv-ac@server.eda-stds.org
> Subject: RE: [sv-ac] mantis 1550
> 
> Hi Stu,
> 
> I guess the question was related to the following: what happens if you
> write @($smapled(x)) where will it be scheduled. In my view and if I
> understood what Arturo wrote, if $sampled changed for the current time
> step and the event control is is in a module, it will unblock in
active
> region, if it is in program it will unblock in reactive region.
> 
> ed
> 
> 
> > -----Original Message-----
> > From: owner-sv-ac@eda.org [mailto:owner-sv-ac@eda.org] On
> > Behalf Of Stuart Sutherland
> > Sent: Thursday, November 02, 2006 2:55 PM
> > To: sv-ac@eda-stds.org
> > Subject: RE: [sv-ac] mantis 1550
> >
> >
> > This message thread has me confused.  I do not understand how
> > a "sampled"
> > value can have an "update" event.  From my user point of
> > view, I expect
> > $sampled and $past simply return to me the STABLE value of a
> > signal as it
> > exists in the Preponed region of the specified time step.  It
> > makes no sense
> > for a STABLE value to need an update event!
> >
> > I use currently use $sampled in pass/fail action blocks as a
> > way to print
> > the Preponed sampled values used by an assertion to determine
> > if a sequence
> > passed or failed.  If you now tell me that the sampled value can be
a
> > changing value, then $sampled and $past become completely worthless.
> >
> > Furthermore, if I am reading the message thread correctly,
> > there seems to be
> > discussion about acting on changes in the Postponed region.
> > This region is
> > very clearly defined as the point in a time step where all
> > activity for that
> > time step has been processed, and NO MORE EVENT ACTIVITY CAN
> > OCCUR.  Once
> > simulation reaches the Postponed region, there are no more
> > iterations within
> > the time step to process the affects of changes, and changes
> > are strictly
> > prohibited.  It is a read-only region.
> >
> > Can someone please explain to me what the AC committee is trying to
> > accomplish by looking for changes on a $sampled or $past value?
> >
> > Stu
> > ~~~~~~~~~~~~~~~~~~~~~~~~~
> > Stuart Sutherland
> > stuart@sutherland-hdl.com
> > +1-503-692-0898
> >
> >
> > > -----Original Message-----
> > > From: owner-sv-ac@server.eda.org
> > > [mailto:owner-sv-ac@server.eda.org] On Behalf Of Arturo Salz
> > > Sent: Thursday, November 02, 2006 11:18 AM
> > > To: Eduard Cerny; john.havlicek@freescale.com;
> > > sv-ac@server.eda-stds.org
> > > Subject: RE: [sv-ac] mantis 1550
> > >
> > > I believe a more accurate statement would be:
> > >
> > > "Sampled values such as those returned by the $sampled and $past
> > > functions are updated in the Postponed region"
> > >
> > > However, currently writing something like
> > > 	always @($time) ...
> > > will not work as expected. If we expect $sampled and $past to
> > > behave in
> > > this manner then they must be defined that way explicitly.
> > >
> > > 	Arturo
> > >
> > > -----Original Message-----
> > > From: owner-sv-ac@eda.org [mailto:owner-sv-ac@eda.org] On Behalf
Of
> > > Eduard Cerny
> > > Sent: Thursday, November 02, 2006 9:51 AM
> > > To: john.havlicek@freescale.com; sv-ac@eda-stds.org
> > > Subject: RE: [sv-ac] mantis 1550
> > >
> > > Hi John,
> > >
> > > I will try to put something like that in the text...
> > >
> > > ed
> > >
> > > > -----Original Message-----
> > > > From: owner-sv-ac@eda.org [mailto:owner-sv-ac@eda.org] On
> > > > Behalf Of John Havlicek
> > > > Sent: Thursday, November 02, 2006 12:47 PM
> > > > To: sv-ac@eda-stds.org
> > > > Subject: Re: [sv-ac] mantis 1550
> > > >
> > > > Hi Ed:
> > > >
> > > > Can we have a clear statement somewhere in 17.7.3 that
> > says that the
> > > > return values of all of the sampled value functions change
> > > only in the
> > > > postponed region?
> > > >
> > > > Also, can we have a statement that update events due to changes
in
> > > > sampled value functions are scheduled in the appropriate
> > > region of the
> > > > next time slot?  [It is still not clear to me that all
> > these will go
> > > > to the active region -- maybe some will go, e.g., to the
reactive
> > > > region.]
> > > >
> > > > If the redundancy is the bottleneck, one could say
> > > something like, "It
> > > > follows from the scheduling semantics (Section 9) that <blah
> > > > blah blah>."
> > > >
> > > > Best regards,
> > > >
> > > > John H.
> > > >
> > > > > X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0
> > > > > Content-class: urn:content-classes:message
> > > > > Date: Thu, 2 Nov 2006 07:58:41 -0800
> > > > > Thread-Topic: [sv-ac] mantis 1550
> > > > > Thread-Index: Acb+j4hFCYP8sfsBR0avmxNqtXjgewABQSDgAADLaVA=
> > > > > From: "Eduard Cerny" <Eduard.Cerny@synopsys.com>
> > > > > X-OriginalArrivalTime: 02 Nov 2006 15:58:42.0631 (UTC)
> > > > FILETIME=[C50FED70:01C6FE97]
> > > > >
> > > > > That's what I had in mind and tried to explain.
> > > > > ed=20
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: owner-sv-ac@eda.org [mailto:owner-sv-ac@eda.org] On=20
> > > > > > Behalf Of Rich, Dave
> > > > > > Sent: Thursday, November 02, 2006 10:53 AM
> > > > > > To: john.havlicek@freescale.com; sv-ac@eda-stds.org
> > > > > > Subject: RE: [sv-ac] mantis 1550
> > > > > >=20
> > > > > > I think the semantics of the return value of these
> > > function is no
> > > > > > different then the simple Verilog system function $time.=20
> > > > > >=20
> > > > > > You have to distinguish between the values returned by=20
> > > > > > references to the
> > > > > > function versus evaluation events scheduled by a
> > > > processes waiting on
> > > > > > the event expression. This is somewhat harder to put into=20
> > > > > > words than to
> > > > > > actually implement it.
> > > > > >=20
> > > > > > I think it is OK to say that the value that will be
> > > > returned by the
> > > > > > function is updated in the postponed region because no
one=20
> > > > > > can schedule
> > > > > > a call in that region. You can also say that an
> > update event is
> > > > > > scheduled for the active region of the next time slot.
> > > > > >=20
> > > > > > Dave
> > > > > >=20
> > > > > >=20
> > > > > > > -----Original Message-----
> > > > > > > From: owner-sv-ac@server.eda.org
> > > > [mailto:owner-sv-ac@server.eda.org]
> > > > > > On
> > > > > > > Behalf Of John Havlicek
> > > > > > > Sent: Thursday, November 02, 2006 6:59 AM
> > > > > > > To: sv-ac@server.eda-stds.org
> > > > > > > Subject: [sv-ac] mantis 1550
> > > > > > >=20
> > > > > > > Hi Ed:
> > > > > > >=20
> > > > > > > In general, I like the semantics for $sampled and $past
> > > > in your 1550
> > > > > > > proposal, but I have some concerns that make me vote
> > > > "no" at this
> > > > > > > time.
> > > > > > >=20
> > > > > > > 1. I don't think we have yet clarified when the return=20
> > > > > > values of these
> > > > > > >    functions change.  You say that $sampled is stable
> > > > throughout the
> > > > > > >    simulator timestep and that $past changes in the=20
> > > > > > postponed region.
> > > > > > >=20
> > > > > > >    Can the return value of $past really change in the
> > > postponed
> > > > > > region?
> > > > > > >=20
> > > > > > >    I think it is bad if there can be calls/references
> > > > to any of the
> > > > > > >    sampled value functions between the point that the
> > > > return value
> > > > > > >    of one changes and the point that the return value
> > > of another
> > > > > > >    changes in the same timestep.
> > > > > > >=20
> > > > > > > 2. A related question is that of the semantics of
> > > > events that refer
> > > > > > >    to sampled value functions.  The intuition seems to
> > > > be that the
> > > > > > >    return values of sampled value functions change "in
> > > > between" the
> > > > > > >    simulation timesteps, so when do we schedule
> > something like
> > > > > > >=20
> > > > > > >       @($sampled(p)) S1
> > > > > > >=20
> > > > > > >    when written in various contexts (e.g., in a
> > module, in=20
> > > > > > a program)?
> > > > > > >=20
> > > > > > >=20
> > > > > > > 3. I would like to see $rose, $fell, and $stable
> > > > defined in terms of
> > > > > > >    $sampled and $past.  I think this should be easy.
> > > > > > >=20
> > > > > > > We may need to get some SV-BC or other help with
> > > items 1 and 2.
> > > > > > >=20
> > > > > > > Best regards,
> > > > > > >=20
> > > > > > > John H.
> > > > > >=20
> > > > > >=20
> > > >
> > >
> > >
> > >
> >
> >