I was recalling my usage from memory; it was actually the threshold
parameters, not the offset parameters which were problematic.
I was doing this in the ami file:
(PAM4_UpperThreshold (Usage Out)(Range 0.33 0.0 1.0)(Type Float)
(Description "The decision level for the upper PAM4 eye")
)
(PAM4_CenterThreshold (Usage Out)(Range 0.0 -1.0 1.0)(Type Float)
(Description "The decision level for the center PAM4 eye")
)
(PAM4_LowerThreshold (Usage Out)(Range -0.33 -1.0 0.0)(Type Float)
(Description "The decision level for the lower PAM4 eye")
)
The EDA tool which did not send these parameters to the model, upon
receiving no values for the threshold parameters from the executable model,
quietly used {0.33,0,-0.33} for the trio. But the intention was for the
model to return these values to the EDA tool based on other hardware tap
settings and adaptation thereof.
Another popular EDA tool (the one I originally regressed the model on) sent
these Usage Out parameters to the model in the parameter string and the
model dutifully returned the real values to the EDA tool.
One could legitimately argue that I "got what I deserved." But perhaps a
little clarification in IBIS-AMI is in order.
Regards,
Bob
On Tue, Nov 1, 2016 at 3:27 PM, Curtis Clark <curtis.clark@xxxxxxxxx> wrote:
My guess is that Bob was referring to the default value that is given in
the spec (0.0) for use when that parameter is not declared in the .ami file
at all.
Is that right Bob?
Thanks,
Curtis
On Tue, Nov 1, 2016 at 5:23 PM, Mike LaBonte <mlabonte@xxxxxxxxxx> wrote:
The only allowable data Format for many Reserved_Paramneters including
the PAM4 Lower/Center/Upper ones is Value, and Default is illegal for
Format Value. So technically there should have been no Default to use, but
there should have been some kind of Value in the AMI file that a tool could
use when the model does not output it.. However, since IBIS says “For
Usage Out, the value in the AMI parameter leaves are determined by the
Algorithmic Model.” (oops, grammatical error), I don’t think tools
should be silent about substituting anything. Some people put unusable
Values like “NA”, attempting to cause a flag to be thrown on the play if it
isn’t replaced.
Mike
*From:* ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@free
lists.org] *On Behalf Of *Bob Miller (Redacted sender "bob.miller" for
DMARC)
*Sent:* Tuesday, November 01, 2016 5:05 PM
*To:* Curtis Clark
*Cc:* dmarc-noreply@xxxxxxxxxxxxx; Walter Katz; Mirmak, Michael; IBIS-ATM
*Subject:* [ibis-macro] Re: Draft Format and Usage Out Clarifications
BIRD
want, right?The model maker can specify Usage InOut if that's the behavior they
Yes. That's what I did to patch the model. In my particular case it
wasn't the *value* passed in that the model wanted or cared about; it
merely used the presence of the parameter in the parameters_in string to
decide whether to present the parameter at all in the parameters_out
string. So in strict terms the usage is not really InOut. This patch had
the side-effect of allowing the user to select a value for the parameter,
which was summarily ignored by the model.
Interesting enough (and possibly an issue), the EDA tool used the Default
value for this Reserved Parameter (PAM4_UpperEyeOffset) when the model
failed to return a value. No warning, no error, just mystically incorrect
post-simulation analysis.
I later modified the model to remove this "auto-hide" feature, rendering
the conundrum moot.
Question: What should the EDA tool do (if anything) when a Usage Out
parameter defined in the .ami file is NOT returned by the model?
Regards,
Bob
On Tue, Nov 1, 2016 at 2:26 PM, Curtis Clark <curtis.clark@xxxxxxxxx>
wrote:
Walter is correct, here's the appropriate section (page 202 of IBIS 6.1):
The EDA tool shall pass a string to the executable model through the
AMI_parameters_in argument. This string shall contain all of the
leaf-formatted Usage In and Usage InOut AMI parameters if there are any
defined in the .ami file. No other information may be included in this
string. The string shall always include the root name of the parameter
tree, even if there are no parameters to pass to the algorithmic model.
Even without that explicit statement, I would consider it an error for
the EDA tool to put Usage out parameters in the AMI_parameters_in string.
The model maker can specify Usage InOut if that's the behavior they want,
right?
Thanks,
Curtis
On Tue, Nov 1, 2016 at 2:16 PM, Bob Miller <dmarc-noreply@xxxxxxxxxxxxx>
wrote:
I agree with Walter. Been caught by this before (executable model subtly
expecting the EDA tool to pass Usage Out parameters into AMI_Init in the
parameter string). Some do, but more importantly, *some don't as there
is no requirement to do so. *I fixed my model rather than demand that
all EDA tools bow to me.
Regards,
Bob
On Tue, Nov 1, 2016 at 11:46 AM, Walter Katz <wkatz@xxxxxxxxxx> wrote:
MM,
The following implies that Usage Out arguments are passed by the EDA tool
to the executable model file in the **AMI_Parameters_In string to AMI_Init.
I do not think that this is correct.
Unless otherwise noted, specific Usage Out arguments or data provided as
Format are effectively ignored and will be simply passed by the EDA tool to
the executable model file.
Walter
*From:* ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@free
lists.org] *On Behalf Of *Mirmak, Michael
*Sent:* Tuesday, November 1, 2016 1:37 PM
*To:* IBIS-ATM (ibis-macro@xxxxxxxxxxxxx) <ibis-macro@xxxxxxxxxxxxx>
*Subject:* [ibis-macro] Draft Format and Usage Out Clarifications BIRD
As requested in a previous ATM meeting, please find enclosed a draft
BIRD, Format and Usage Out Clarifications.
Comments and questions are of course welcome and encouraged.
- MM
