[ibis-macro] Minutes from the 01 Jul 2014 ibis-atm meeting
- From: Mike LaBonte <mike@xxxxxxxxxxx>
- To: 'IBIS-ATM' <ibis-macro@xxxxxxxxxxxxx>
- Date: Tue, 01 Jul 2014 16:28:07 -0400
Minutes from the 01 Jul 2014 ibis-atm meeting are attached.
Mike
IBIS Macromodel Task Group
Meeting date: 01 July 2014
Members (asterisk for those attending):
Agilent: * Fangyi Rao
Radek Biernacki
Altera: David Banas
ANSYS: * Dan Dvorscak
* Curtis Clark
Avago (LSI) Xingdong Dai
Cadence Design Systems: * Ambrish Varma
Brad Brim
Kumar Keshavan
Ken Willis
Ericsson: Anders Ekholm
Intel: Michael Mirmak
Maxim Integrated Products: Hassan Rafat
Mentor Graphics: * John Angulo
* Arpad Muranyi
Micron Technology: * Randy Wolff
Justin Butterfield
QLogic Corp. James Zhou
Andy Joy
SiSoft: Walter Katz
* Todd Westerhoff
* Mike LaBonte
Synopsys Rita Horner
Teraspeed Consulting Group: Scott McMorrow
* Bob Ross
The meeting was led by Arpad Muranyi.
------------------------------------------------------------------------
Opens:
- None
--------------------------
Call for patent disclosure:
- None
-------------
Review of ARs:
- None
-------------
New Discussion:
BIRD 147:
- Arpad: Walter sent an email he wanted to share
- Arpad showed the email
(text of email here)
- Arpad: SiSoft's removal from BIRD 147 is not really new.
- Is Walter withdrawing his own proposal?
- Todd: Yes
- We believe co-optimization is about optimizing some system level metric
- BIRD 147 is about hardware level emulation
- Our customers optimize thousands of channels on a per channel basis.
- Arpad: It sounds like the proposal is withdrawn due to lack of interest
- We actually have about 50% interest in it
- Todd: SiSoft has been looking for a broader discussion of optimization, with
less acrimony.
- Modeling hardware is a valid way to achieve optimization.
- There are other ways, but we haven't been able to discuss them.
- BIRD 147 believes it can model the optimization each channel will find.
- We have other needs but have to be careful how our time is spent.
- Arpad showed a presentation
- slide 2:
- Arpad: Two IC vendors are important but not here today, Intel and Altera.
- slide 3:
- Arpad: Here is a high level comparison of the proposals.
- With BIRD 147 the EDA tool facilitates communication.
- The .bci file has the communication information.
- With Walter's proposal the .ami file has the communication information.
- Those are already passed by the tools.
- slide 4:
- Arpad: In both proposals the EDA tool passes parameter strings.
- Both proposals will need BIRD 128.
- Ambrish: I'm not sure the SiSoft proposal has string passing.
- Todd: The string to AMI_GetWave is passing information.
- Arpad is right that passing is needed.
- Arpad: In both cases the DLL function signature is not changed.
- With BIRD 147 the EDA tool is only a "mailman".
- slide 5:
- Arpad: Intel likes the BCI approach because it supports proprietary content.
- The BCI approach is good because the TX and RX use the same file.
- Maybe the SiSoft approach could do the same with Model_Specific
parameters.
- slide 6:
- Arpad: Altera likes not having to recompile the TX with the SiSoft approach.
- It might be possible if the EDA tool pretended to be the TX.
- The EDA tool would only know about standard protocols.
- The SiSoft special AMI parameters could help with this.
- slide 7:
- Arpad: We need a solution that serves both of these vendors.
- Todd: The two proposals are similar in many ways.
- We haven't defined the problem well enough.
- Intel values supporting proprietary protocols.
- Altera likes supporting legacy models, that could be a requirement or a
goal.
- The RX could ask for things the TX can't do.
- Enforcing consistency will be a challenge.
- We have to ask which co-optimization methods we want to enable.
- We need to take a step back and prioritize the various needs.
- Ambrish: The problem is well defined.
- We work with 3 or 4 IP vendors.
- Our proposal meets their requirements.
- BIRD 147 is at least a subset of the complete solution.
- David idea to not change the TX is nice to have, but not high priority.
- Arpad: This presentation is 2 weeks old, not a response to Walter.
- With Walter withdrawing we have only BIRD 147.
- Todd: Do we just vote on the BIRDs brought to us?
- Or as a standards body do we solve problems?
- BIRD 147 might not evolve well to a more general solution.
- SiSoft believes it is not enough.
- Arpad: should we send BIRD 147 to the Open Forum for a vote?
- Bob: We should hold off, the SiSoft proposal has good points.
- We should work on the points brought up by Intel and Altera.
- More discussion would be helpful, currently I'm neutral about this.
- Ambrish: What is needed to complete BIRD 147?
- Bob: I don't know if we can have BCI for co-optimization.
- Arpad: If BIRD 147 is a subset it is therefore incomplete.
- Ambrish: It solves the problems given to us.
- Todd: It does solve the time domain emulation of hardware back-channels.
- If we are trying to solve co-optimization for buffers that are not designed
to talk to each other,
do we need to modify the TX and RX have to talk to each other?
- Arpad: We have back-channel setups where the chips optimize each other.
- Or the EDA tool could be doing it.
- Todd: In either proposal the brains are in the RX.
- Arpad: I thought the SiSoft proposal allowed the tool to be more involved.
- Todd: The RX knows how to optimize itself, and it can view the TX as generic.
- We only need to abstract the TX operation so the RX understands it.
- The TX DLL does not need to change, we just describe it.
- The RX might not call the TX in the language it understands.
- The wrapper function Walter mentioned would do that.
- Waiting for steady state after each command could eat up a lot of time.
- System level statistical should be part of the proposal.
- Arpad: BIRD 147 supports that.
- Todd: It seems to have been tacked on.
- Ambrish: We made a serious effort on that and we know it works.
- Todd: The flow involves calling the model multiple times.
- That will have a significant impact.
- It needs to be thought out more and tested.
- Arpad: How do we go forward?
- Todd: We could have a vote.
- Bob: If there is other information we need it should be put on the table.
- SiSoft's withdrawal takes away one option.
- Arpad: The vote would be to promote BIRD 147 or continue working.
AR: Arpad formulate a vote for next week.
-------------
Next meeting: 08 Jul 2014 12:00pm PT
-------------
IBIS Interconnect SPICE Wish List:
1) Simulator directives
Other related posts:
- » [ibis-macro] Minutes from the 01 Jul 2014 ibis-atm meeting - Mike LaBonte
