Hello Everyone,
This is essentially a continuation of our discussion, I just gave it a new
subject line
to reflect the new direction in the discussion.
I would like to start with the question Walter raised, because I feel that the
answer
might have a significant effect on how we continue the work along the lines of
Fangyi’s proposal.
Walter’s question was:
“what do we do in the Redriver flow for legacy models that do not return the
new IR from the AMI_Init function”,
and I think this is an important question. I do agree that it would be
desirable to
find a way to support existing models also. The additional IR input/output
approach
in Fangyi’s proposal seems to be a roadblock in that regard.
This brings me back to the idea I mentioned several times already. Can we find
a
way to execute the Init function multiple times so that we could get all the
necessary
IRs to achieve the correct results? I know there may be challenges in this
approach,
but can we figure out a way to overcome them? Perhaps in a limited (but
acceptable)
way for legacy models, and a complete way for new models (with additional
parameters
if necessary)?
For legacy models I would imagine that running the Init functions multiple
times and
keeping its Initialization and optimization results only from the last run for
the
subsequent GetWave runs might be a workable option. For new models I would
think that we might want to introduce one or more new parameters if necessary to
suppress some of the unnecessary functionality to make the models more efficient
(and/or accurate).
Could we discuss these ideas a little further to see whether we can address
Walter’s
legacy model question before we go down the wrong path that might only support
new models? I tend to be optimistic, and I have a feeling that there is a
possibility
to achieve this goal.
Thanks,
Arpad
===================================================================
