Re: [sv-ac] checker: why are parameters not allowed? Need to file a mantis?

Dmitry, the topic is all about the use model,or application, of the
checkers in the verification environment. In some applications, like the
one you showed, one may want to build a library of checkers. In other
applications, I can see the declaration of checkers in a package, or just
within a verification unit. However, there are other applications where a
checker is defined a la module style, and that checker is bound to a DUT.
 There, parameters would be useful because they mirror a structure familiar
to SV engineers.
Why not using a module instead of a checker? One main reason I see is the
access to free variables, and that allows for the definition of assertions
in a style not allowed in modules. For example, local variables of a
property do not flow across the "implies" operator, but I believe that we
could use the rand const in formal to get that variable conceptually flow
across.

Bottom line, parameters are clearly constants, whereas inputs are not.
 Using inputs for ranges is not explicitly clear that the actual is a
constant. Adding parameters in the next PAR would be optional (just like
parameters in modules) and we could still use inputs to pass constants. It
would not hurt.
Ben Cohen

On Wed, May 23, 2012 at 5:04 AM, Korchemny, Dmitry <
dmitry.korchemny@intel.com> wrote:

> Hi Ben,****
>
> ** **
>
> This mantis already exists, see http://www.eda-twiki.org/svdb/view.php?id=2111.
> ****
>
> ** **
>
> However, IMO there is little added value in introducing parameters to
> checkers. Checker invocation should be similar to a property’s syntax, and
> not to a module’s.****
>
> ** **
>
> Say, you want to check that a request will be granted in n cycles. Then it
> is very natural to write it like this:****
>
> ** **
>
> // invocation****
>
> request_granted rg1(req, gnt, 4, clk);****
>
> ** **
>
> checker request_granted(request, grant, n, event clk);****
>
> a: assert property (@clk request |-> ##n grant);****
>
> endchecker****
>
> ** **
>
> Rather than to use this ugly syntax:****
>
> request_granted #(4) rg1 (req, gnt, clk);****
>
> ** **
>
> checker request_granted #(n) (request, grant, event clk);****
>
> a: assert property (@clk request |-> ##n grant);****
>
> endchecker****
>
> ** **
>
> What is a benefit of introducing parameters for checkers in your opinion?
> Do you have a specific example?****
>
> ** **
>
> Thanks,****
>
> Dmitry****
>
> ** **
>
> ****
>
> *From:* Ben Cohen [mailto:hdlcohen@gmail.com]
> *Sent:* Tuesday, May 22, 2012 22:12
> *To:* sv-ac@eda-stds.org; Korchemny, Dmitry; Eduard Cerny
> *Subject:* [sv-ac] checker: why are parameters not allowed? Need tofile a
> mantis?****
>
> ** **
>
> Dmitry, ****
>
> Why are parameters not allowed in checkers, in the same style as module? *
> ***
>
> If we do allow them, would that bring more consistency in design style,
> and possibly simplify the design of tools because of this parallelism to
> modules? ****
>
> Before I file a mantis for future exploration of this, I would like to
> hear your comments. ****
>
> Thanks, ****
>
> Ben ****
> ---------------------------------------------------------------------
> Intel Israel (74) Limited
>
> This e-mail and any attachments may contain confidential material for
> the sole use of the intended recipient(s). Any review or distribution
> by others is strictly prohibited. If you are not the intended
> recipient, please contact the sender and delete all copies.
>

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.