[ibis-macro] Re: Backchannel BIRD discussions
- From: Ambrish Varma <ambrishv@xxxxxxxxxxx>
- To: "twesterh@xxxxxxxxxx" <twesterh@xxxxxxxxxx>, "ibis-macro@xxxxxxxxxxxxx" <ibis-macro@xxxxxxxxxxxxx>
- Date: Mon, 19 May 2014 14:50:26 -0700
Todd,
I think I know where the confusion is coming from.
The example files in my BCI presentation are actual 10GBase-KR, PCIe3 and SAS3
BCI files that we are proposing to use for the Backchannel communication.
The means by which to identify the Training pattern and the protocol specific
section in the BCI file is to 'mine' the protocols and fill in the desired
section. For SAS3, I have a couple of tables on slide 8 that contain relevant
information:
[cid:image003.jpg@01CF738A.D1305340]
Every protocol is different - so there is no fixed method for doing this.
I want to point out that this is another advantage of the BCI file. The
protocols can be interpreted differently by different device model makers -
creating a source of confusion. With the BCI file, the model maker will get a
clear idea of what their model needs to support for a particular protocol.
There will be no need to read the protocol and map the various features that
may or may not be important to Backchannel.
Thanks,
Ambrish.
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx]
On Behalf Of Todd Westerhoff
Sent: Monday, May 19, 2014 5:09 PM
To: ibis-macro@xxxxxxxxxxxxx
Subject: [ibis-macro] Re: Backchannel BIRD discussions
Ambrish,
I guess I'm still confused.
Does BIRD 147, as proposed, include any .bci files that would be approved
as part of the standardization process? (my guess is no).
If that guess is correct, then we will need standardized .bci files _in
addition to_ BIRD 147, unless the only intent is to support private
protocols (which I assume it isn't).
Thus, I repeat the original question - did we identify the means by which
.bci files will be defined and standardized going forward?
This is a Standardization issue. Example .bci files are helpful for
purposes of understanding Cadence's proposal, but they mean nothing with
respect to ensuring model interoperability between vendors. If we don't
have a plan for standardizing .bci files, then we don't have a complete
proposal to consider.
Todd.
Todd Westerhoff
VP, Semiconductor Relations
Signal Integrity Software Inc. * www.sisoft.com
6 Clock Tower Place * Suite 250 * Maynard, MA 01754
(978) 461-0449 x124 * twesterh@xxxxxxxxxx
"I want to live like that"
-Sidewalk Prophets
-----Original Message-----
From: Ambrish Varma [mailto:ambrishv@xxxxxxxxxxx]
Sent: Monday, May 19, 2014 4:43 PM
To: Mike LaBonte; twesterh@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx
Subject: RE: [ibis-macro] Re: Backchannel BIRD discussions
Hi Mike,
I was referring to future .bci files and not the ones that would be
released as examples.
Also - BCI files will not be tied to IBIS release at all. There may be 3-4
BCI files getting approved by IBIS in a single year. This process will not
be connected to the IBIS specification in any way.
Thanks,
Ambrish.
-----Original Message-----
From: Mike LaBonte [mailto:mike@xxxxxxxxxxx]
Sent: Monday, May 19, 2014 2:55 PM
To: Ambrish Varma; twesterh@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx
Subject: Re: [ibis-macro] Re: Backchannel BIRD discussions
The IBIS parser would be updated to recognize .bci files only after an
IBIS specification covering the BCI format is approved. How would tested
.bci files be included in the same specification? Yes, BCI files are
similar to AMI files so maybe ibischk6 -ami could be used, but there are
syntax rule differences. And getting the syntax right is only part of it.
Overall I'm concerned that we would publish BCI contents in an IBIS
release only to soon find that changes are needed. Our experience with
IBIS 5.0 teaches us that getting things wrong in a formal specification
can be painful. I would recommend focusing on a verified publishing
procedure independent of the IBIS specification.
Mike
On 5/19/2014 11:24 AM, Ambrish Varma wrote:
> Hi Todd,
> We propose to publish 3 .bci files for recent standards as samples with
the BIRD. We can discuss the details of the contents when we are
finalizing the BIRD. Any future .bci files will be based on the protocols
that become available. We will set aside time in the IBIS Open Forum for
discussing the new .bci file at which point we may relegate the
discussion/approval to the IBIS ATM. Once approved, these files will be
publicly available at the IBIS website in a separate directory.
> These .bci files will be parsed/tested using the latest IBIS parser
before getting posted so there should be no surprises at run time.
>
> Thanks,
> Ambrish.
>
>
>
>
> -----Original Message-----
> From: ibis-macro-bounce@xxxxxxxxxxxxx
> [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Todd Westerhoff
> Sent: Monday, May 19, 2014 9:41 AM
> To: ibis-macro@xxxxxxxxxxxxx
> Subject: [ibis-macro] Re: Backchannel BIRD discussions
>
> All,
>
> Did we identify the means by which .bci files will be defined and
> standardized going forward?
>
> Todd.
>
>
> Todd Westerhoff
> VP, Semiconductor Relations
> Signal Integrity Software Inc. . www.sisoft.com
> 6 Clock Tower Place . Suite 250 . Maynard, MA 01754
> (978) 461-0449 x124 . twesterh@xxxxxxxxxx
>
> "I want to live like that"
> -Sidewalk Prophets
>
> -----Original Message-----
> From: ibis-macro-bounce@xxxxxxxxxxxxx
> [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Ambrish Varma
> Sent: Monday, May 19, 2014 12:41 AM
> To: ibis-macro@xxxxxxxxxxxxx
> Subject: [ibis-macro] Re: Backchannel BIRD discussions
>
> Thanks Arpad,
>
> I would like to take this opportunity to point out the main highlights
> of the Cadence proposals:
>
> - Our proposal implements a traditional backchannel system where the
> entire job of the EDA tool is to open a communication channel between
> the Tx and Rx and let them communicate without the need for the EDA
> tool to interpret anything from the Tx or the Rx. Once the training is
> completed, the EDA tool detects the 'Traning_Done' parameter and
> closes the communication channel.
>
> - Our proposal also allows for flexible interaction between the Tx and
> Rx based on a particular protocol. The details of the taps or any
> other parameters are encapsulated in a separate BCI file that is
> pointed to in the AMI file based on the protocols that are supported by
the AMI model.
> There is no hardcoded methods of communication that the other method
> prescribes.
>
> - The flexibility to implement and support a 'Private Protocol' is
> also possible without any further addition of Reserved Parameters.
>
> - The backchannel related parameters for training and any protocol
> specific parameters are confined to the BCI file - the contents of
> which (for a published protocol) are decided by IBIS and will be
> available publicly for anyone to see/implement in their AMI models.
> This leaves the .ami file relatively clean (with a couple of top level
> backchannel related parameters).
>
> These are a few salient features of our proposal. I would also like to
> point out that we have had the backbone of the proposal (passing a
> parameters string back and forth from the Tx to the Rx and vice versa)
> tested in many models. In other words, this is a tried and tested
> method and will add tremendous value to AMI and the IBIS specification
> and will fill a huge void in how real devices are modeled.
>
> We would be glad to answer any question that might still be out there
> about our proposal.
>
> Regards,
> Ambrish.
>
> -----Original Message-----
> From: ibis-macro-bounce@xxxxxxxxxxxxx
> [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
> Sent: Friday, May 16, 2014 6:26 PM
> To: ibis-macro@xxxxxxxxxxxxx
> Subject: [ibis-macro] Backchannel BIRD discussions
>
> Hello everyone,
>
> First, I would like to draw your attention to two presentations which
> have been posted recently on the ATM website from Cadence and SiSoft
> on the topic of Backchannel AMI modeling.
>
> http://www.vhdl.org/pub/ibis/macromodel_wip/archive/20140506/ambrishva
> rma/ The%20BCI%20File/BCI_File_and_Communications.pdf
>
> http://www.vhdl.org/pub/ibis/macromodel_wip/archive/20140513/walterkat
> z/Ba
> ckchannel%20Training%20Co-Optimization/Backchannel_Training_%20Co-Opti
> miza
> tion%20BIRD.pptx
>
> Once again, our discussions are at crossroads and some important
> decisions need to be made. I would like to make a request to everyone
> to become familiar with the two proposals and be ready to express your
> opinion and/or preference in the upcoming IBIS-ATM teleconference. I
> would like to encourage IC vendors and model makers to pay attention
> to this discussion and express your opinion in this discussion because
> the decision we are about to make will have a great effect on how
> models with Backchannel training and optimization features will have to
be written.
>
> Thanks in advance,
>
> Arpad
> ===================================================================
Other related posts:
