[ibis-macro] Re: Questions on API proposal
- From: Mike Steinberger <msteinb@xxxxxxxxxx>
- To: michael.mirmak@xxxxxxxxx
- Date: Tue, 12 Jun 2007 08:26:46 -0500
Michael-
I don't know enough about the IBIS language to address your main
question, but I can at least address the mechanics of the API.
Each instance of a given algorithmic model type has its own identity,
and that identity is expressed by its memory handle. e.g., two separate
instances of a SerDes design attached to two different pairs of pins
will have two different memory handles. There is no argument to the
functions which is a name string for a given instance; however the model
developer is free to include a name string in the model's parameter
tree. If the model developer chose to do that and the user wanted to
name one SerDes instance "George" and the other one "Fred", then the
user would be able to include in the parameter string something like
"...(name George)..." and "...(name Fred)...". When these parameter
strings get parsed, the model could then include in the memory space it
allocates a char* for the model instance name. In one case one might
have something like
char* instance_name = "George";
and in the other case
char* instance_name = "Fred";
The memory handles would then provide a means for the model to get to
these names.
On the larger question of the extent to which the IBIS language
does/should accept contributions from the department of redundancy
department, I can only observe that if an [Algorithmic Model] statement
could exist outside the scope of a [Model] statement, then there would
need to exist a mechanism for identifying/pointing to a specific
[Algorithmic Model] statement from inside the [Model]. I don't think
there's a clear answer as to whether or not the additional complexity of
such a mechanism would be justified.
"Whether 'tis nobler to suffer the slings and arrows of redundant code
or to take arms against a sea of copy-and-modify..."
My 2c.
Mike Steinberger
Mirmak, Michael wrote:
Looking at the API proposal text, I have two implementation questions:
1) Is [Algorithmic Model] below the level of [Model] in the keyword
hierarchy? In other words, can an [Algorithmic Model] only be used
with the particular [Model] that contains its definition?
I can imagine circumstances where I might want a single [Algorithmic
Model] definition to be used independently with several different
analog [Model]s (e.g., where the same engine design is used separately
to process data but for two different interfaces). Under the current
definition, I would have to re-define the [Algorithmic Model] several
times, even if identical, for individual [Model]s, even if they are
associated under [Model Selector] with a single pin.
2) On a related note, does [Algorithmic Model] not accept a text
string argument as an identifier? Again this implies that defining an
[Algorithmic Model] is explicit under [Model] but instantiating it is
implicit through the [Pin] list.
This effectively means that each [Algorithmic Model] is only
associated with a single [Model] and therefore a single [Pin]. I can
imagine -- and might be interested in creating -- cases where a block
of algorithmic code processes the results of data involving multiple
analog buffers associated with multiple pins. In this syntax, I have
no way of defining this association, at least not through [Pin],
[Model] or any argument to [Algorithmic Model]. The [Circuit Call]
keyword does allow me, by contrast, to associate a single block of
code with multiple pins.
Am I missing something?
- MM
Other related posts:
