ALF Working group Meeting - MGC San Jose - 2/12/98
---------------------------------------------------

Attendees:

Mike Andrews     (Mentor Graphics)
Kevin Grotjohn   (LSI Logic)
Mitch Heins      (Ambit)
Eric Howard      (Cadence Design Systems)
Timothy Jordan   (Motorola)
Tai Le           (Avant!)
Wolfgang Roethig (NEC)
Jin-sheng Shyr   (Toshiba TAEC)
Peter Suaris     (Mentor Graphics)
David Wallace    (Exemplar / Mentor Graphics)
Cary Wei         (Fujitsu)
Jeff Wilson      (Mentor Graphics)

Glossary:

ALF - Advanced Library Format
DCL - Delay Calculation Language
DPCM - Delay and Power Calculation Module

ALF-DCL framework strawman proposal
-----------------------------------

* Kevin Grotjohn gave an update to the group on the proposal
  for the ALF-DCL framework from the recent ASIC council
  technical committee meeting.

   Main points of interest to the ALF committee were:
    
 => The technical committee accepted the ALF-DCL strawman
    proposal. (The executive committee will make their
    decision this week).

 => In general, the ASIC council will only support an open
    standard in the public domain. This was in response to
    Synopsys' LIBERTY based proposal, which was also presented.

 => The recommendation is that EDA tools should only read
    the binary library information as compiled into a DPCM,
    rather than read ALF files directly. The main reason
    being that certain vendors will start from ALF source
    and others may write everything directly into DCL.

 => The intention is to have prototype applications ready in
    time for DAC 98. EDA vendors are invited to take part in
    this 'beta' program.

 => A combined ALF/DCL group meeting is proposed for the
    week of IVC/VIUF.

 => Accessing the function from the DPCM will be via a special
    API call, but all other extra ALF information will most
    likely be passed through as strings that retain the ALF
    syntax.
 
 => It is proposed that SI2 will program manage the generation
    of a flow, consisting of an ALF parser / compiler which
    will add information from ALF into the DPCM. (CAD service
    provider options are requested.)

The subject of whether a separate golden parser project should
also be funded by OVI was discussed, or whether this could be a
part of the SI2 program, in order to minimize duplication of
effort. It was generally agreed that there should be a free and
freely available utility for performing syntax and basic semantic
checks, regardless of who manages / funds the program.

Some ASIC vendors are now in a "wait and see" mode. It is crucial
that the golden parser and applications based on the ALF-DCL
framework become available soon in order to get development
of ALF libraries started. This is the reason, why this effort
is driven through the ASIC council, which has the most
leverage and impact on the industry.


Feedback on ALF v1.0
--------------------

Documentation Issues:

     In the ALF V1.0 document, the definition of the context
     sensitive keywords in BNF format is unclear. I.e. they read
     as hard keywords. (Page ref: 3-13)

     Definition of keyword-value pairs has lost the "=", in
     the syntax definition. (Page ref: 3-29, 3-36)

     In the section on polarity of pins it should be explicit
     that the attributes "inverted", "non-inverted", "both"
     are applicable only to outputs, whereas the attributes
     "high", "low", "rising_edge", "falling_edge", "double_edge"
     are applicable to inputs. (Page ref: 3-32)


Proposals for ALF v1.1
----------------------

General:

     A proposal for a bus resolution primitive was discussed
     but generally thought to be non critical at this moment.

ALF-DCL framework:

     Explicit classes of operation mode (e.g. scan and
     non-scan, read and write mode) would map to the current
     DCL mode-dependent timing methodology.

     To mirror the 'late mode-early mode' capability in DCL
     it should be stated that the use of enumerated, non-
     interpolatable tables can be used for this purpose.
     (See page 4-21, where this method is used to separate
     characterization corners.)

Signal Integrity:

     It was agreed that signal integrity modeling issues
     need to be well-defined and resolved first, before attempting
     to standardize models for signal integrity.
     Areas with open issues are crosstalk, voltage drop ...
     Constraints could be library constraints (candidate for
     future ALF version) or design constraints (candidate for
     other standards).

     One idea was to look at the set of well-defined signal
     integrity models in IBIS and investigate whether the portion
     relevant for cell characterization can be expressed in ALF
     (e.g. driver resistance as a function of slewrate and load,
     noise margin as a function of voltage swing ...).
     This would enforce ALF's role as "golden" characterization
     source from which IBIS models can be derived.

     The IBIS specification will be obtained for the next ALF
     meeting for the work group to review.

Available documentation
-----------------------

* from ASIC council work group

  http://www.si2.org/alf-dcl/final_presentation.ps
  http://www.si2.org/alf-dcl/final_presentation.pdf
  http://www.si2.org/alf-dcl/strawman.v1.1.pdf

* from OVI

  http://www.ovi.org/TS-PSC/

* latest press release

   http://techweb.cmp.com/eet/news/98/994news/advanced.html
   http://techweb.cmp.com/eet/news/98/994news/asic.html