IBIS Macromodel Task Group Meeting date: 06 September 2016 Members (asterisk for those attending): ANSYS: * Dan Dvorscak Curtis Clark Broadcom (Avago): Xingdong Dai * Bob Miller Cadence Design Systems: Ambrish Varma Brad Brim Kumar Keshavan Ken Willis Cisco: Seungyong (Brian) Baek eASIC: David Banas Marc Kowalski Ericsson: Anders Ekholm GlobalFoundries: Steve Parker IBM * Luis Armenta Intel: * Michael Mirmak Keysight Technologies: Fangyi Rao Radek Biernacki * Ming Yan 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 Kevin Li Teraspeed Consulting Group: Scott McMorrow Teraspeed Labs: * Bob Ross TI: Alfred Chong The meeting was led by Arpad Muranyi. -------------------------------------------------------------------------------- Opens: - Walter said Michael M had emailed back-channel questions. - Arpad: Ambrish and Curtis emailed that they could not attend. ------------- Review of ARs: - Walter submit [Pin Mapping] BIRD to Open Forum. - Done. - Bob Miller submit Backchannel BIRD to Open Forum. - Done. ------------------------- No patents were declared. ------------------------- Review of Meeting Minutes: - No Aug 30 minutes yet. ------------- New Discussion: Michael M email about back-channel communication: - Michael M: Colleagues had concerns about using files for communication. - Are I/O restrictions needed? - Bob Miller pointed out that it is not a live simulation, that the models are called at separate times. - Walter: That is correct. - Synchronization might be every 1000 UI, for example. - Bob M: Block size could theoretically be 1 UI but that would be silly. - Michael M: With a large block size some events might be missed. - Arpad: EDA tools determine block size. - There is no restriction that it must remain the same. - What if the block size changes from call to call? - Walter: We will be requiring only a minimum. - The exact number should not be critical. - Arpad: Can this change after initialization? - Walter: Maybe the suggested block size should apply only during training. - Bob M: This is used for latency during message passing. - The EDA tool could choose a smaller block size for memory reasons, for example. - Walter: The model might not send messages every block. - Arpad: We should clarify that it is only for the training period. - Bob M: We could do that. - The overhead of a GetWave call is small. - Mike L: File caching can be a problem, particularly with Windows. - Has this approach been tried? - Bob M: Some testing has been done. - Walter: Caching should be a problem only for separate processes. - Arpad: We should not count on the TX and RX being in the same process though. - Mike L: The BIRD appears not to preclude using sockets. - Walter: This is much like the original BIRD 147. IBIS 6.2 BIRDs: - Walter showed an updated IBIS 6.2 BIRDs spreadsheet. - Bob: BIRD A is not BIRD 184, D is 184. - Walter: BIRDs F, G, and I all have to do with pin reference. - The question is what to do when the component has no reference pin. - Bob and Radek are working this out. - Bob: We have not connected to discuss that yet. - Walter: There are two worlds: - One in which everything is floating. - One which is ground based. - Arpad: A number of BIRDs were untabled in the previous Open Forum meeting. - It is not clear if we should discuss them here. - Walter: BIRD 128 was untabled because BIRD 147.1 doesn't need it. - BIRD 180 relates to BIRDs 125, 163, and 164. Maybe BIRD 165 too. - It says there are no joins or splits between pins and models. - You can't have the same pin twice. - The parser flags these as errors. - BIRD 180 makes that illegal. - It's not clear how an EDA tool should handle such a case. - Arpad would prefer to avoid voting on BIRD 180 until BIRD 125 is voted down, but that is waiting for the new Interconnect BIRD. - The Interconnect BIRD is inconsistent with BIRDs 125, 163, and 164. - Those should be voted down sooner rather than later. - Arpad: The question is if we discuss them here. - Mike L: It would be good to have them on this agenda. - Walter: It helps for IBIS to be consistent with the parser. - Arpad: We could put BIRD 180 on this agenda. - Bob R moved to put BIRD 180 on the agenda. - I would not change the parser to handle multiple pins. - Mike L seconded. - Without objection the motion passed. - Walter moved to remove BIRD 147.1 from the tabled BIRDs list. - This was to delete, not to untable. - Bob R seconded. - Without objection the motion passed. - Bob R: Is the New Redriver Flow BIRD an update to the Redriver flow BIRD? - Walter: There is no BIRD yet, just a PowerPoint. - It overlaps with BIRD 166. - It adds multiple Impulse Response inputs and outputs. - They could be combined into one BIRD. - Bob R: A draft BIRD was sent last October. - Walter: I have not seen it posted. - Michael M: There should be no interaction between BIRDs 147.1, 166, and this new BIRD. - Walter: There will be none. - Walter: Motion to adjourn. - Bob R: Second. - Arpad: Thank you all for joining. ------------- Next meeting: 13 September 2016 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives