Meeting Minutes: Analog System Verilog Assertions, Dec 16th, 2008

From: Anand Himyanshu-R61978 <Himyanshu.Anand_at_.....>
Date: Tue Dec 16 2008 - 11:53:55 PST
Cancelled: Meeting on Dec 30th, 2008 is cancelled due to holiday season.
Next Meeting on January 13th, 2009

Attendance: At the end, please correct if I missed someone.

Meeting Minutes from John Havlicek
Meeting 2008-12-16
------------------

- Committee to go forward, introduce requirements for analog assertions AMS.
- KB: Putting extension in Verilog AMS doesn't solve problem for 
customer who
  has mixed signal design.  Analog portions need to be integrated with 
digital
  (20% analog with 80% of problems, 80% digital with 20% of problems).  Pure
  analog designer operating only in AMS -- maybe they do not exist.  Big 
market
  is digital dominated mixed signal design.  Attach some analog to this 
in some
  verification runs.  This is an argument against putting the extension in
  Verilog AMS.
- ??? can you clarify interface?  KB: AMS has wreal wire.  SV has real 
port type
  or variable, but it is not net-like.  Connectivity information is not 
kept.
  The AMS resolution algorithms need this connectivity information.
- HA:  If we extend SVA and make it a part of AMS extended language, how 
would
  this prevent prototyping?  KB:  What has to come along with SVA?  Clocking
  blocks?
- JH:  Outlined what is typically in assertion module.  Assertions, modeling
  code, supporting tasks and functions, covergroups and coverage code.
- What of AMS needs to be imported into SV in order to define the extensions
  in SV?  KB:  There is an interface issue.

- KB:  Heart of matter is dense time, purely digital issue.  Make assertions
  on dense rather than clocked time.  Prerequisite to any consideration of
  analog crossings.  As a practical matter it has to be implemented by SV
  teams.

- Prabal:  There is a part of AMS extension needed.  Some SVA extension 
needed
  to get meaningful things done.  This is not something we can do 
without some
  SVA-AMS effort.  Think incrementally about what features need to be 
brought
  into SV.

- DM:  Verilog AMS committee hasn't started up really, and won't until 
next year.

- 1800 to support dense time.  Real expressions in booleans.

- DN:  SVA to dense time is a semantic issue.

- KB: Can we superimpose dense time semantics on top of clock semantics 
and have
  a single language for both?  DN: Map digital to dense time.  Other 
direction
  can lose information.  KB: A dense time language for SV is different 
from the
  clock event semantics.

- Prabal: Can users try to mix and match digital sensitivities with analog
  sensitivities?  KB: Mix dense time and clocked time.  Prabal: @(cross 
...) is
  not really boolean.  KB: cross has an event-like nature.  Prabal: Syntax
  issue.  Can I import mixed signal events into it?

- JH: I am hopeful that the current digital assertion semantics could be
  overlayed on top of continuous time without any severe impact to the 
digital
  assertion usage model.  This would open the way to import event 
controls from
  analog side.

- JH: Users are approximating analog assertions with digital.  They are 
using
  auxiliary variables to get the effect of comparison expressions involving
  reals.  They are using high frequency clocks to get sufficient 
sampling to be
  confident.  They are making engineering decisions about the tradeoffs
  involved.

Meeting Minutes from Himyanshu Anand
-----------------------------------------------------
The question of the committee's work going to System Verilog (P1800) or 
Verilog-AMS was raised. P1800 is winding down it's work related to 2009 
release. As such, it is unlikely it will open a new PAR anytime soon 
(2-3 years). On the other hand VAMS is starting up new work and may be 
able to incorporate analog assertions work. There did not seem to be a 
consensus on the where the analog assertions should reside. Dave Miller 
was of the opinion that some initial work needs to be done in System 
Verilog, that way SV-VAMS integration will inherit the basic support in 
VAMS for analog assertions. Ken believes that the SVA should be extended 
to dense time semantics. Himyanshu mentioned that the extension could be 
part of the VAMS language and left for implementation to decide on how 
to do the actual partition. Also, that later on when the P1800 opens up, 
some parts of the extension could be pushed back into System Verilog as 
appropriate. John mentioned that we should continue on extension work 
and decide later on the committee.

The following needs to be addressed - dense time semantics of assertions 
in SVA, real relational support and special functions defined for the 
analog properties like (frequency calculator, mean etc.). Dejan 
mentioned that moving dense time to digital time could possibly be done, 
but will incur loss of information.

Prabal mentioned that @cross is not boolean. Can it be viewed as a 
boolean event by the digital domain in a mixed signal simulator. Prabal 
also wanted to know about the interface resolution that Ken talked 
about. Dave Miller wanted to know whether SVA supported PLI calls. John 
mentioned that assertions (SVA) could call functions and tasks but was 
not sure whether that allowed PLI calls.

Ken asked who were the end users and what kind of assertions they liked 
to write. John, Scott and Himyanshu mentioned that the users were 
already using a mixture of System Verilog/VAMS using workarounds since 
there was no direct tool support for analog assertions. The examples 
that Freescale shared with the committee were motivated by what the 
users are/trying to do with assertions in their designs.

There was a discussion on unifying time semantics, using @cross 
naturally across the assertions and how @cross determines the time step 
and its comparison with a digital overhead of a high frequency clock 
running in the background.

Ken mentioned that designs are 80% digital with 20% problems and 20% 
analog with 80% problems. As an example of the interface issue Ken 
metioned the lack of wreal in digital (SV), and the fact that real 
variable is not really equivalent to wreal wire in VAMS. Such interface 
issues would need to be sorted out. Ken wants SV teams to implement the 
analog extensions.

Scott and Himyanshu raised the issue that some constructs are purely 
analog, like @cross, voltage and current probes. Thus, as it is, the 
question is not purely digital nor is it purely analog but a mixture of 
both. Hence, needs to be in both SVA and VAMS.

Prabal from Cadence believes that it will have to be addressed by both.

The conclusion: Continue with the work to define the extensions and the 
work required for analog assertions. Revisit the question of which 
committee this needs to go into later.

Attendees: Total 12
------------------------
Erik Seligman 	Intel
Mike Demler 	Individual Contributor
Prabal Bhattacharya 	Cadence
Kishore Karnane 	Cadence
Kenneth Bakalar 	Mentor Graphics
Patrick O'Halloran 	Tiburon Design Automation
David Sharrit 	Tiburon Design Automation
Dejan Nickovic 	EPF Lausanne
Dave Miller 	Freescale
John Havlicek 	Freescale
Scott Little 	Freescale
Himyanshu Anand 	Freescale



-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
Received on Tue Dec 16 11:55:01 2008

This archive was generated by hypermail 2.1.8 : Tue Dec 16 2008 - 11:55:19 PST