IEEE P1076.1 Subcommittee on VHPI-AMS - May 20, 2010 8am PDT

Attendees

• David Smith, champion
• Peter Ashenton
• Yaseen Zaidi, note taker

Agenda

• Call to order
• Approval of agenda
• Review and approve minutes from April 22 meeting
• Review action items and any content provided from them
• Discussion
• Next Meeting
• AOB
• Adjourn

1. Call to order

David opened the meeting and started presenting the slides

2. Approval of agenda

The agenda was discussed and there were no changes.

3. Review and approve minutes from April 22 meeting

Peter made motion, David seconded, no abstentions or objections. Minutes were approved.

4. Review action items and any content provided from them

• Ernst - get a copy of 1076-2008 for use by the working group
  • Posted on the server in working group document. Posted at WorkingGroupDocumentsPrivate.
• Ernst - contact Cadence (Francois Martinolle) to see if they have a VHPI definition for VHDL-AMS that they might consider donating
  • Francois responded but no clear position on donating. Ernst requested she work we David to pursue.
• David - send out information about where Martin's paper is posted (included in these minutes)
  • Done and on the site
• David - develop scope and review before next meeting
  • Presented at meeting (Meeting presentation)
• Peter - develop initial analysis of objects that need to be added to existing VHPI
  • No progress

5. Discussion

1. David presented slides (posted below) containing important aspects of Vlach’s paper and digital VHPI spec. This is first cut of VHPI-AMS requirements.
2. David made a point that 1076-2008 is essentially 1076-2002 as far object definitions are concerned, which we would use as baseline. 1076-2008 affects IP protection more than objects.
3. A number of requirements were brainstormed for better scoping VHPI-AMS: interaction with elaborated design, reelaboration, simulation control, simulation cycle and integration algorithm.

1. A point was made that digital VHPI does not address elaboration. For VHPI-AMS we need more clear view on reelaboration since topology change seems easier than writing to free quantities.
2. Modification of generics in the information model may change design structure, equations and Jacobian, therefore some restrictions would have to apply on generics’ change if reelaboration is supported with callbacks.
3. A point was made that digital VHPI provides RESET mechanism while STOP and FINISH return control to interactive user or command source. The control is minimal, therefore in VHPI-AMS we may need to provide more control for analyzing the obtained solution during transient or other simulation analyses. This capability coupled with reelaboration will allow replacing models during simulation with better/refined ones. Also to be noted is that digital VHPI does not support data processing.
4. Many people seem to seek VHPI for speed, simulation control and model library use during execution. Peter mentioned that VHPI could contain C routines to command simulator but this scope is simulator specific, out of VHPI domain and thus should rest with the outside manager/controller/ backplane/user.
5. Queries on analog objects seem straightforward, current routines have to be extended to act on analog data objects.
6. VHPI does not specify types of profilers, code coverage, and post analyzers, rather these are applications that are built using the routines provided. Peter mentioned there is always an inquiry how to use the routines to build API. A proper use model therefore would have to be developed.
7. Misc. discussion on interaction with integration algorithm and intermediate solution analysis: trajectories, linear first order approx., piecewise/constant approx., time step setting, global truncation error, smoothening. I don’t remember what came out of this discussion.
8. VHDL-AMS language provides attributes that assist in meta simulation procedures and also to generate events by detecting thresholds, so getting in (writing generics) and out (reading objects) of the information model in signal flow fashion is plausible, however writing quantities or topology changes need a deeper look.
9. VHPI-AMS should provide mechanism to call into integration algorithm even though there is no analog event, perhaps periodically or at specified time. The requirements for interaction with the simulation cycle and integration algorithm need to be understood in detail.
10. Misc discussion on writing to quantities. Attaching resistors to sources. Driving quantities means introducing topology changes by inserting sinusoid sources whose values originate from outside of the simulator.

5. Next Meeting

Thursday June 17, 2010, 8am PDT (Delayed until June 24 due to DAC).

Action Items:

• David - pursue contact with Francois to see if there is a donation
• David - update scope with feedback from meeting and post on site
• Peter - develop initial analysis of objects that need to be added to existing VHPI

6. AOB

7. Adjourn

Meeting was adjourned at 9am PDT.

-- DavidSmith - 2010-06-21