[ibis-macro] Re: Suggested change to Usage Out and InOut BIRD
- From: "Muranyi, Arpad" <Arpad_Muranyi@xxxxxxxxxx>
- To: "IBIS-ATM" <ibis-macro@xxxxxxxxxxxxx>
- Date: Wed, 13 Apr 2011 12:47:28 -0700
Walter,
I have to disagree with your proposed text below.
First, I don't think we need to talk in the specification
about when a is model compliant with the specification.
The nature of a specification is to define the rules by
which the models and simulators operate. It is an obvious
"given" that if someone does things that is outside the
spec that they are not compliant with that spec. The spec
doesn't have to state that "if you don't do what I say, you
are not compliant", or "if you do everything the way I say
you are compliant".
Second, you seem to interpret "compliance" in a really
unusual and strange way. The way I understand your
suggestion is that you consider models and tools to be
compliant if they produce the same results, regardless
whether they do what the specification says or not. Why
bother having a specification then?
Third, the last sentence "There shall be no prohibition from a User
to direct a simulation to use a compliant model in a non-compliant way."
is
a clear and premeditated anti-trust idea. You want a model
to be able to tell the tool to perform non compliant things,
without letting the competing EDA vendors in this field know
what features they have to implement in their tools in order
to support such specific non-compliant actions the model's
user may want to perform. Without a specification describing
what actions the model's user can request from the tool, the
tool vendor will not know what features they need to implement
for a user who wants to perform non-compliant simulations.
This contradicts even your own views about models and tools
being compliant if they reproduce the same results. Without
a specification it is simply impossible to reproduce the same
results.
This would only work for companies whose product line involves
both models and tools, or if there are alliances between certain
model making companies and EDA tool vendors. But such alliances
might be secret, and they cannot be enforced to be public.
This scene is simply wrong and can set up a stage for unfair
competition.
I think it is totally inappropriate to put words like this
into a specification. The purpose of a specification is to
establish a fair playing field for all those who are interested
in implementing the rules spelled out in the specification so
that the models (which also obey the rules in the specification)
would work. We cannot put statements in the specification to
endorse (or encourage) non-compliant activities. This would
undermine the purpose of the specification.
Your suggestions even violate the first two goals you proclaim
for IBIS-AMI in various documents:
With the suggestions you made in your email these two goals
are NOT possible. And don't tell me that the models made
according to your suggestions are all compliant (because they
pass the IBIS parser), therefore they are interoperable and
portable. They may be for a limited set of their capabilities
or accuracy level, but the whole reason you want to go outside
the compliance is to be able to something bigger and better
that the spec doesn't support. This automatically sets a stage
for the above unfair competition because those who want to use
the bigger and better features of the model will have to go to,
guess who? The ONLY EDA vendor who knows how to simulate it,
because they make the model(s) and the tool.
Sincerely,
Arpad
==================================================================
From: Walter Katz [mailto:wkatz@xxxxxxxxxx]
Sent: Tuesday, April 12, 2011 5:28 PM
To: Muranyi, Arpad; IBIS-ATM
Subject: Suggested change to Usage Out and InOut BIRD
Arpad,
I suggest an alternative to the following paragraph in the Usage Out and
InOut BIRD
Model_Specific parameters must not be used for purposes which may alter
past, present or future simulation results in any way. In case there is
a need to use parameters of (Usage Out) or (Usage InOut) for such
purposes, those parameters shall be defined as Reserved_Parameters and
fully described in the specification.
Proposed replacement paragraph:
A Model is compliant if when presented with a set of inputs as described
by its .ami file and other inputs as described in the approved IBIS
specification gives the same outputs (within acceptable numerical
precision) on any hardware platform that the model supports. The various
IBIS AMI test benches can be used to validate this. A simulation is
compliant if it does not use the values of Out or InOut Model_Specific
parameters to modify the inputs to any future calls to any AMI model in
the simulation. A simulation is compliant if it does not use the values
of Out or InOut Reserved_Parameters to modify the inputs to any future
calls to any AMI model, unless such modification to inputs to any AMI
model in the simulation is fully described in the IBIS specification.
There shall be no prohibition from a User to direct a simulation to use
a compliant model in a non-compliant way.
Walter
Walter Katz
wkatz@xxxxxxxxxx
Phone 303.449-2308
Mobile 720.333-1107
Other related posts:
