[ibis-macro] Re: Format Corner - Interpretation?
- From: "Muranyi, Arpad" <Arpad_Muranyi@xxxxxxxxxx>
- To: "IBIS-ATM" <ibis-macro@xxxxxxxxxxxxx>
- Date: Wed, 30 Mar 2011 15:24:06 -0700
Bob,
I am not sure all previous models would stop working if
Default for Corner would be illegal starting in v5.1.
We do have a version number. Wouldn't that sort things
out?
Thanks,
Arpad
==========================================================
From: Bob Ross [mailto:bob@xxxxxxxxxxxxx]
Sent: Wednesday, March 30, 2011 5:12 PM
To: Muranyi, Arpad; 'IBIS-ATM'
Subject: RE: [ibis-macro] Re: Format Corner - Interpretation?
Arpad:
Your analysis is absolutely correct.
Also, we are dealing overall with a specification where not all people
were
on the same page - literally, nor did we have time to fully analyze the
interaction
impact of what was specified. That is evident in released models and
spec.
interpretations. Default was poorly thought out, and we are recovering
from
it with solutions compromising between perfection and what people have
implemented.
My overall recommendation is to keep default optional, as is currently
proposed, to avoid breaking models that work. In most cases, where
used, Default just repeats the typical value. For Corner, it makes no
sense, but tools could be advised to treat Default as an (unlikely)
but intended alternative to Typ. That also applies to List, Range,
Step,
and Increment - if Default is different than the Typ value.
To many changes from allowed to illegal will break many models
for unimportant reasons.
Bob
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Wednesday, March 30, 2011 11:28 AM
To: IBIS-ATM
Subject: [ibis-macro] Re: Format Corner - Interpretation?
Bob,
You are making good points regarding Default with Corner,
but the reason I was questioning this was from a different
perspective. Think about it this way:
Default is usually used for multi-valued parameters where
the user has to make a choice. Default serves the purpose
for having a value in absence of the user's selection. Put
this thinking together with Corner, but first a little
background.
Reading the existing specification I had no clue that
Corner was actually assigned by the tool (the note on
pg. 141 doesn't say anything about that). I only started
to get some clues about that when I began to study how
the Dependency Tables work. Some of the examples use
Corner as a control input for the table and I was
wondering why Corner was not defined in the same .ami
file (as required for all Dependency Table parameters).
Then I realized that this was one of those "predefined"
parameters listed in BIRD 124. But there is still no
description in BIRD 124 about how the predefined
parameters get their values, etc...
My "detective work" was misguided by the presence of the
Default value in some of those models which contain
Corner. The presence of Default gave me the impression
that Corner may be a user selection through the same GUI
which allows them to make choices for the other AMI
Algorithmic Model parameters. This "evidence" in my
detective work contradicts the logic that Corner is
assigned by the tool. There is no situation when a
tool wouldn't know what corner it simulates, therefore
there is no need for Default, right?
Now, think about an EDA vendor who needs to implement all
this. If the conclusion is that the tool makes this
assignment, the value of Corner should always be known.
But if the existence of Default is an indication that
this is a user selection, then we have to make a GUI
entry for it. Taking this one step further, we can also
ask the question: does the existence of Default imply
that regardless of what the tool uses for Corner in the
analog simulation engine to generate the channel
characterization impulse response, it should allow the
user to select a (different) Corner for the Algorithmic
Model simulation engine through the AMI parameter GUI?
To me doesn't make sense at all, but it is certainly a
possible conclusion based on what information is available
in the specification.
I think we have to address this issue and correct the
specification. This is why I am asking all these questions
about these rules. So while your reasons for keeping Default
may be doable from a spec perspective and for keeping the
existing models compatible with the new spec, I would
like to find out what was the intent of the authors when
the AMI specification was written regarding Corner. Given
our track record I don't know whether the appearance of
Default in certain models is only a "left hand doesn't
know what the right hand does", a "senior moment" or
intentional. Consequently I don't know which reasoning to
apply from my "detective work" to my interpretation of the
specification.
I would like to identify a consistent and reasonable solution
for this confusing situation for Corner.
Thanks,
Arpad
==============================================================
From: Bob Ross [mailto:bob@xxxxxxxxxxxxx]
Sent: Wednesday, March 30, 2011 11:00 AM
To: Muranyi, Arpad; 'IBIS-ATM'
Subject: RE: [ibis-macro] Re: Format Corner - Interpretation?
Arpad:
We would do exactly as you state. That is not my concern.
All entries in #6 start with a Typ value. The Optional Default
is consistent in all cases as an override for some parameters.
My concern is consistency of our rules and existing applications.
I believe many models already redundantly use Corner and Default.
There is no reason for change, and the Default overrule interpretation
can be valid for some analysis for all of these cases.
In the cases for #7, there is no definition of "Typ" entry. So the
Default exclusion rule is valid (as with the exclusive OR rule
with Value in #4).
I think Default was not fully thought out in its relationship to all the
other <data_format> parameters, but creating a spider-web of
inconsistent rules for our primitive elements does more harm than good.
Bob
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Wednesday, March 30, 2011 8:39 AM
To: IBIS-ATM
Subject: [ibis-macro] Re: Format Corner - Interpretation?
Bob,
I didn't remember that these "Notes" mentioned the same
as the text I quoted for change. I don't see why we
couldn't delete "Corner" from #6 and add to #7 in these
notes. Am I missing something?
Thanks,
Arpad
=========================================================
From: Bob Ross [mailto:bob@xxxxxxxxxxxxx]
Sent: Wednesday, March 30, 2011 10:34 AM
To: wkatz@xxxxxxxxxx; Muranyi, Arpad; 'IBIS-ATM'
Subject: RE: [ibis-macro] Re: Format Corner - Interpretation?
All:
My only concern is now we have an inconsistent rule. Here is where
this issue fits in:
|* All parameters must be in the following format:
|*
|* (parameter_name (Usage <usage>)
|* (Type <data_type>)
|* ({Format} <data_format> <data>)
|* (Default <value>)
|* (Description <string>))
|*
|* Notes:
|* 1) The order of the entries is not important.
|* 2) The word Format is optional as indicated by the curly
|* braces "{" and "}" and may be ignored by the EDA
tools.
|* (The examples do not show the word Format).
|* 3) Certain reserved parameter names allow only certain
|* <data_format> selections, as described below.
|* 4) The <data_format> selection of Value and Default are
|* always mutually exclusive. Certain parameters may
require
|* Value or Default, but Value and Default are not
allowed to
|* be present together for the same parameter.
|* 5) <data_format> is always required for selections other
|* than Value.
|* 6) Default is optional for <data_format> Range, List,
Corner,
|* Increment and Steps.
|* 7) Default is not allowed for <data_format> Table,
Gaussian,
|* Dual-Dirac and DjRj.
|*
List, Range, and Corner, Increment, and Steps all have the first entry
defined as a "Typ" value.
The same logic applies for Corner in rule 6.. You may choose the slow
or fast value for the Default just
as you might choose any of the non-Typ values for the other
<data_format> entries.
So I think the rule should remain as is.
Bob
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz
Sent: Wednesday, March 30, 2011 8:18 AM
To: Arpad_Muranyi@xxxxxxxxxx; 'IBIS-ATM'
Subject: [ibis-macro] Re: Format Corner - Interpretation?
Arpad,
I agree that (Default Value) does not make much sense for (Corner typ
slow fast).
Walter
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Wednesday, March 30, 2011 10:55 AM
To: IBIS-ATM
Subject: [ibis-macro] Re: Format Corner - Interpretation?
Walter,
I checked in BIRD 127 (Out Typos BIRD) and found this in it:
| Default <value>:
|* Default and Value are mutually exclusive, and must not be used
together
|* for the same parameter. Default is not allowed for Table,
Gaussian,
|* Dual-Dirac and DjRj. Default is optional for Range, List, Corner,
|* Increment and Steps. If a Default <value> is specified, its value
must
|* have the same Type as the parameter. For example, if Type is
Boolean,
|* <value> must be either True or False, if Type is Integer, <value>
must
|* be an integer. Also, if Default is specified, <value> must be a
member
|* of the set of allowed values of the parameter. If Default is not
|* specified, the default value of the parameters will be the <typ>
value.
I still wonder if Default makes sense for Corner. Wouldn't
a tool always know what corner they are working with, therefore
wouldn't Corner always be set to a known value?
What do you think?
Thanks,
Arpad
==================================================
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Tuesday, March 29, 2011 1:52 PM
To: IBIS-ATM
Subject: [ibis-macro] Re: Format Corner - Interpretation?
Walter,
Another question:
Does a parameter of Type Corner need Default? Or is it
not needed, or even prohibited? When would a tool make
use of a default value for them? The tool always knows
what corner it simulates, right?
Thanks,
Arpad
========================================================
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Tuesday, March 29, 2011 1:39 PM
To: IBIS-ATM
Subject: [ibis-macro] Re: Format Corner - Interpretation?
Walter,
Thanks for the answer. However, I think that's not the
end of the story yet...
If the value of corner (and the rest of the predefined
parameters) are set by the tool, the following statements
don't make sense in BIRD 124:
| "xyz" is a predefined AMI parameter of Type String and Usage Info
(Usage Info) is information that goes from the .ami file
to the tool. If these parameters come FROM the tool they
can't be Info.
Thanks,
Arpad
============================================================
From: Walter Katz [mailto:wkatz@xxxxxxxxxx]
Sent: Tuesday, March 29, 2011 1:22 PM
To: Muranyi, Arpad; 'IBIS-ATM'
Subject: RE: [ibis-macro] Format Corner - Interpretation?
Arpad,
Corner is determined by the EDA tool.
Walter
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Tuesday, March 29, 2011 2:15 PM
To: IBIS-ATM
Subject: [ibis-macro] Format Corner - Interpretation?
Hello AMI experts,
I stumbled on a question regarding how Format Corner
supposed to be interpreted by an EDA tool. Pg. 140
of the IBIS specification say this:
and pg. 141 has this:
Then on pg. 146 I see this:
and on pg. 147 I see this:
In BIRD 124 I read the following:
| "[Corner]" is a predefined AMI parameter of Type String and Usage Info
that can
| be used to create an input column in a dependency table. It has a List
format with
| the allowed values "Typ", "Slow" and "Fast".
but I don't see any text in the Specification or the BIRD that
explains how "Corner" is predefined and from where and how it
gets it value. Could someone explain this to me please? It
seems that this needs another clarification BIRD...
Thanks,
Arpad
===============================================================
Other related posts:
