RE: [sv-bc] DataTypes: The Enhancement

Creating a 'wone' data type is an orthogonal issue, and
`default_nettype' may be a tangential issue. I think we need to refocus
if we want to get the original enhancement - data types on wires'
through.

If the only change for this enhancement is to allow an optional data
type declaration after an explicit net type declaration, we should focus
on:

1. What other features or semantics are absolutely needed to make this
one enhancement work. For example, would the signal alias (section 5.7)
need to be modified?

2. How should we limit the allowed data types on wires? My suggestion is
to only allow fixed sized bit-stream data types (read Annex J Glossary)

3. Do we need the ability to get the actual value/strength of the
individual bits of the wire? I think the user would need to use either
an alias, continuous assignment, or go through a port and connect to a
wire of a different type. Which raises another question, how does this
affect the rules for connecting dissimilar net types. I think this
should not change and the data type only applies to the casting of the
local identifier when used in an expression.

So you can see there is a lot on our plates to think about with just
this 'one' enhancement.

Dave

-----Original Message-----
From: owner-sv-bc@eda.org [mailto:owner-sv-bc@eda.org] On Behalf Of
Krishna Garlapati
Sent: Friday, October 29, 2004 12:40 AM
To: Steven Sharp
Cc: btf-dtype@boyd.com; sv-bc@eda.org
Subject: Re: [sv-bc] DataTypes: The wone net type

Sure, but that's not exactly what I meant. My main point was that
why not extend 'default_nettype to user defined types (of nets) ?
If this can be done, an attribute route would still be acceptable.

That said, I dont have a problem with the idea of wone and I would
not get into the how many/few customers thing. If they want it, who
am I to say no.

Thanks,
- Krishna.

Steven Sharp wrote:
>>At the risk of getting into an argument for the sake of it, that
brings
>>up another question. When SV data types are extended to nets why not
>>allow `default_nettype to be legal with any user defined net data type
??
>
>
> Krishna, the "net types" in `default_nettype are not the data types,
but
> are the object "kinds". They are independent or orthogonal to the
> data types. They are not user defined, and the work being done in
this
> group doesn't change that.
>
>
>>Once that happens, a user can create custom data types with the wone
like
>>attribute and use it along with `default_nettype. While I have nothing

>>against wone like nets, I am only wondering why make the language any
bigger
>
>
> Since users cannot create new net types (i.e. "kinds" in the
nomenclature
> being used in this group), there is no facility for doing this
themselves.
>
>
>>1) if an attribute can capture the essence of a new construct
>>2) (and) it's something a select few would be interested in ??
>
>
> How few you think would be interested in it may depend on what
customers
> you have been talking to.
>
> Steven Sharp
> sharp@cadence.com
>
>
>
Received on Fri Oct 29 09:43:15 2004