From: Adam Krolnik (krolnik@lsil.com)
Date: Mon Feb 14 2000 - 07:36:05 PST
Good morning Steve, James;
<p>What a discussion to come to in the morning.
I reviewed the proposed and reproposed BNF for attributes. You found
most of the places I though could also be included, here are a
few more:
udp_port_list
function_item_declaration
list_of_parameter_assignments
combinational_entry
sequential_entry
case_item
function_case_item
To note, do you wish to attach an attribute instance to
module_or_generate_item, or only the place in generate_item?
This specific place would allow attributes on any generated element.
The translate_on/translate_off discussion was interesting. If a
compiler extended the attribute to following objects (until the end
of the module) it would be well within the functionality provided.
One would not have to then resort to begin/end markers, instead
only one marker for an object and a separate marker for
'ignore until enabled.' The tool that defines this functionality
will most like implement to allow it. Most tools won't need
this ability. Yes, it is disappointing that it's not a general
capability. Having attributes attached to objects is much more
useful that implying attachment due to position.
The attribute placement really needs to be specified and standardized.
If compilers forbid them where they aren't allowed, then multiple
definitions will render them useless. Consider if your synthesis
tool won't accept them. Will you use them in code that the synthesis
tool will read? Many folks will decide not to.
With regard to multiple attribute instances, I think this is
a reasonable request. We already allow multiple attributes
to be specified, it will only make tools (read user scripts)
easier to write to allow multiple instances.
Are attribute instances going to be inside a module instantiation?
>From my experience most people are unaware a list of items can
be instantiated, (declared, etc.) So the requirement to specify
each instantiation as a separate element seems to be the standard
now.
With regard to the list at the end of the text section on attributes,
we will have to remove system function calls. Any others?
This is a really good step toward making this generic across
the description.
Adam Krolnik
Verification Mgr.
LSI Logic Corp.
Plano TX. 75074
This archive was generated by hypermail 2.1.4
: Mon Jul 08 2002 - 12:54:12 PDT
and
sponsored by Boyd Technology, Inc.