Meeting minutes for the SV-CC Committee

October 29, 2002, 9:00-10:00am PST

Attendees:

[--xxxxxxx-] Yatin Trivedi (ASIC Group, Chair)
[-----xxx-x] Tarak Parikh (@HDL)
[--xx-xxxxx] Francoise Martinole (Cadence)
[xxx--xxxxx] Stuart Swan (Cadence)
[------xxxx] John Amouroux (Mentor)
[------x--x] Emerald Holzwarth (Mentor)
[-----xxxxx] John Stickley (Mentor)
[------xxx-] Doug Warmke (Mentor)
[--xx-xxxxx] Michael Rohleder (Motorola)
[xxx-xxxxx-] Kevin Cameron (National Semi)
[x---------] Tayung Liu (Novas)
[-xxxx-xxxx] Bassam Tabbara (Novas)
[-----xxxxx] Swapnajit Mittra (SGI)
[--x----xx-] Darryl Parham (Sun)
[xxx-x-----] Simon Davidmann (Synopsys)
[xx-xx-----] Peter Flake (Synopsys)
[xxxxxxxxxx] Joao Geada (Synopsys)
[xxx-xxxxxx] Ghassan Khoory (Synopsys, Co-Chair)
[xxxxx-xxxx] Andrzej Litwiniuk (Synopsys)
[---x-xxxx-] Alain Reynaud (Tensilica)
[x-----xxx-] Mike McNamara (Verisity)


1. Review and approval of last week's meeting minutes.

2. Vote on coverage API completed - was accepted 4 votes for, 2 abstain.

3. Face-to-face meeting confirmed for Nov. 7th. Tentatively planned to
    be held at Synopsys's Mtn. View offices.

    * ACTION: Swapnajit to confirm.

4. Discussions surrounding issue of whether function calls should support
    0-time only or allow time consumption.

    * This issue was discussed at length by e-mail earlier in the week.

      Some favor coming up with a way of modeling time consuming tasks
      on the C side and allowing time consuming HDL tasks to be called
      from the C side. Others prefer to restrict calls and tasks to 0-time
    * semantics. Joao suggested that at least for functions, since Verilog and SV require

      0-time, let's nail down the interface for 0-time function calls in both
    * directions and defer the issue of time consuming task support until later. Stuart felt that support for time consumption in calls needs to be worked

      out up front before we continue to far down the road with the specification
    * process for functions only. John S. favors requiring 0-time semantics for HDL-to-C or C-to-HDL

      function calls and for C-to-HDL task calls either restricting to 0-time semantics
      or allowing HDL tasks to consume time but not have the API itself
      worry about preventing or allowing time consumption. As far as allowing for
      time consumption on the C side, the API itself should not try to define a "task"
      structure per se, but rather let the C environment such as SystemC handle it and
      only consider function calls as 0-time synchronization points among time
    * consuming tasks or processes of the native language environment (C or SV). Joao stated that he would like to make it a goal of next week's phone conf.

    * and face-to-face meetings to get closure on this issue. Andrzej Litwiniuk also made some proposals regarding time consumption

in is proposal sent out shortly before the meeting.

5. Discussions about context sensitivity function call handling proposal from John S.

    * John S. reiterated that this proposal did not change the syntax or HDL-to-C

      call syntax and argument handling proposed in the DirectC donation other
      than to provide additional support for a context pointer that can bind calls
    * to specific C models. Joao commented that this might also be an elegant way of solving the "C task"

      dilemma by allowing model context pointers to refer to specific model objects
    * or time consuming threads on the C side. This proposal also tried to suggest a "mirror syntax" for the C-to-HDL

      direction that was modeled after the HDL-to-C direction already in the
      DirectC donation document. Context handling for the C-to-HDL direction
    * was also proposed. ACTION: Ghassan asked John S. to coordinate with Joao and Stuart to
      formalize the proposal for next week's face-to-face meeting.

6. Discussion about function name resolution and function override.

    * I missed some of the gist of this topic but I gathered that it was being

      suggested that we have some sort of function configuration capability
      whereby a function caller can bind through a configuration mechanism
    * to a function callee. Swapnajit also suggested that functions in local scope on the HDL side

      always override HDL functions of the same name in outer HDL scopes
    * or C functions of the same name being called on the C side. More discussions need to occur to formalize these requirements.

7. Touched on Andrzej Litwiniuk's proposal sent shortly before the meeting.

    * ACTION: Committee members should review this proposal and send feedback by next week's meeting.

Please forward any corrections or omissions to John_Stickley@mentor.com 

-- johnS
                                                           __