IBIS Macromodel Task Group Meeting date: 27 May 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 Scott Huss 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 Teraspeed Consulting Group: Scott McMorrow * Bob Ross Texas Instruments Alfred Chong The meeting was led by Arpad Muranyi. ------------------------------------------------------------------------ Opens: - None -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - None ------------- New Discussion: Backchannel: - Bob: The issue is whether or not to have a BCI file. - The Cadence proposal has 3 reserved parameters, 6 BCI, and a new Type Bits. - The SiSoft proposal has 16 Reserved_Parameters, but it has more generality. - Don't have a good feeling about the BCI file. - Arpad: There is a question about the gatekeeper of the BCI file. - Do we need to decide that first? - Fangyi: We can't decide on a proposal without deciding that. - Ambrish: That language is in the BIRD. - The IBIS committee or a subcommittee would review and post on the website. - Fangyi: Would voting be required? - Ambrish: Yes, we need a standard way to accept it. - Ken: This committee will be important. - If it goes into the AMI file there probably will be new reserved parameters. - I like the Protocol_Specific, which avoids new parser requirements. - Ambrish: It would not depend on interpretation from anyone else. - Radek: There could be an interpretation problem, and private protocols. - Ambrish: Private protocols would not use the standard BCI files. - Fangyi: How will the too understand private protocols? - Ambrish: It does not have to understand anything. - Radek: We might ratify something not in BCI and have an empty solution. - Walter: The BIRD itself can not have any BCI examples that would be ratified at the same time. - Increment and decrement are part of the private protocol. - Ambrish: It is part of the BIRD right now. - Todd: If it's a black box then it can't say increment and decrement? - Walter: Yes. - Bob: That is confusing. - Walter: 1.5.1 says the RX can send increment, decrement, or unchanged commands. - PCIe doesn't do that, it sends a number 0 to 63 for tap coefficients. - Ambrish: We have a solution for that which takes just a few lines. - Walter: Why can't that be private? - Ambrish: The RX and TX need to understand each other. - John: Does anything need to be added to help designers? - Ambrish: Like the final values that are settled on? - There can be an output file. - John: EDA tools should be able to present that to the user. - Ambrish: That is not decided yet. - Ken: That is not a functional issue. - John: There might be a model specific part of the BCI file. - Ambrish: That's a good idea but not part of this. - Ken: Outputs beyond what a protocol requires should be accommodated. - Radek: We have not discussed dependence between BIRDS 147 and 128. - Communication between RX and TX may depend on it. - Arpad: Does Walter's proposal depend of 128? - Walter: It is preferable to use existing GetWave, but we could have a new one. - That should not affect the decision. - A alternative TD approach could call Init and GetWave. - My proposal does rely on BIRD 128. - Truly private communication would need only a new text parameter. - It may not be needed, all current TXs can be controlled with what I've proposed. - Ambrish: Statistical models are not a reflection of actual silicon. - Walter: That is an imaginary problem, every TX has dynamic range limits. - Arpad: Someone asked about training with only one GetWave call, can that be done? - Walter: A real system requires the RX to communicate to the TX between GetWave calls. - Ambrish: It is possible the RX could say nothing needs to be done. - Kumar: 600,000 bits in one call should not be a problem. - Arpad: This is about requesting a long bit stream for optimization. - Kumar: There may be practical memory problems, we should support multiple calls. - Todd: The RX only needs to not send another instruction until it has enough bits. - Arpad: Can the models tell the EDA too how many bits to provide? - Walter: There can be performance issues, but a new Reserved Parameter would be needed. - Any reasonable adaptation would require at least 10 to 20 iterations. - You can only change taps once per GetWave. - Fangyi: Will be final result depend on GetWave sizes? - Arpad: Could we revisit the motion from the last meeting? - Walter: Ambrish made the motion and I seconded. - Ken: It should be whether to proceed with BIRD 147. - Arpad: That BIRD is currently open but tabled in the Open Forum. - The SiSoft proposal is not a BIRD. - We can make recommendations to the Open Forum on how to vote on BIRD 147. - Walter: The question is if this committee should table work on BIRD 147. - Option B would be to work on another BIRD. - Michael M: BIRD 147 was sent to us by the Open Forum for our recommendation. - Radek: Last week we had IC vendors who are not here today. - We should have their votes. - Arpad: Two responses received from one company conflict with each other. - That might be an abstention or we could take the second response. - Not sure what the final opinion would be from the others. - TI is not a voting member however. - Michael M: It depends on who made the motion. - Ambrish: It should be to work on BIRD 147 or go with Walter's proposal. - Michael M: Usually the vote is Yes/No/Abstain - We votes on whether to proceed with BIRD 147. - Roll call vote: Agilent: N ANSYS: A Cadence Design Systems: Y Intel: Y Mentor Graphics: A Micron Technology: A SiSoft: N Teraspeed A - Tally: 2 Yes, 2 No, 4 Abstain. - Ambrish: Can we hold the vote later with other members present? - Arpad: Marvell may want to vote. - Bob: Marvell is not a member. - Walter motioned to invalidate the vote and schedule a new vote. - Arpad: That may not be necessary, we had a tie. - Walter: DAC is next week, we might not meet. - Arpad: We can meet in two weeks and vote then. ------------- Next meeting: 10 June 2014 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives