SV-ACMinutes2010_06_22 < P1800

P1800 Web>SystemVerilogAssertionCommittee>SVACMeetingMinutes>SV-ACMinutes2010_06_22 (2010-06-25, ErikSeligman) (raw view)EditAttach
Minutes from SV-AC Committee Meeting

Date: 2010-06-22

Time: 16:00 UTC (9:00 PDT)

Duration: 1.5 hours

Dial-in information:

--------------------

Meeting ID:            38198

Phone Number(s):

1-888-813-5316         Toll Free within North America

Agenda:

-------

- Reminder of IEEE patent policy.

See: http://standards.ieee.org/board/pat/pat-slideset.ppt

- Minutes approval

- Email ballot results

- New issues:

3117: make it clear that rewriting algorithm (F.4.1) applies to checker

and let

3120: "expect" construct to refer to virtual interfaces

- Issue resolution/discussion

3113: Add port_identifier to constant_primary BNF for sequences,

properties and checkers

2732: Clarify timing diagram in Figure 16-4?Future value change

2362: 16.14 mention of assertion control system tasks is unconnected

2825: 16.16 Disable iff: checkers not included in list of default extensions

2754: P1800-2009 : Can clock change in conditional branch of 'if' operator

2927: Precedence between sequence/property operator and normal

expression operator

2452: No vacuity information about synchronous aborts

2557: Rules for passing automatic variables to sequence subroutines are

not clear

2556: Explicit package scope indication is not allowed for checkers

2476: Need clarification about system functions $onehot, etc

1763: The LRM does not define whether assertion control tasks affect

sequence methods and events

2485: terminology related to immediate and deferred assertions

1756: The LRM does not indicate how the control tasks $asserton/off/kill

 affect verification statements in initial blocks

2809: Checker instantiation in checkers' always procedure

2938: Surprising (to some users) interaction between deferred assertions

& short-circuiting

2353: 'classes' missing from description

2871: Clause 16 does not forbid assertion local variables within

clocking event expressions

2248: Champions feedback - items related to Mantis item 1683

1678: Clarify that rewriting algorithm doesn't replace name resolution

2934: Precedence and associtiativity of case operator is not shown in

the table

3008: In $past BNF, "expression" should be "expression1"

2479: Annex F.5.2.1 conflicts with changes from 2434

2494: 37.44 Assertion diagram missing restrict

2095: Clarify meaning of distribution as condition for "disable iff"

1627: 17.16: clarify that expect statement not allowed in functions

2904: Clarify when disable iff condition must occur relative to starting

and ending of an attempt

1853: BNF for calls to $rose and other sample value system functions

3015: Examples of $fatal have bad arguments

2839: Contradictory statement of increment/decrement operators usage.

2551: trivial example error

2558: Restriction inside checker construct

2340: clarifications needed on vpi_control for non-temporal and

immediate assertions

2271: sequence events require a clocked sequence

2255: clarifications on expect

2491: Conflicting rules in 16.17 (D7)

2552: Confusing comments regarding nexttime operator

2571: confusing assertion clock inference rule

2722: Errors in Figures 16-14, 16-15, and 16-16

2546: 'empty match' and 'vacuous success' are not clearly defined in LRM

2386: Rename 16.9 to "Local variables"?

- Opens

Attendance Record:

------------------

        Legend:

                x = attended

                - = missed

                r = represented

                . = not yet a member

                v = valid voter (2 out of last 3 or 3/4 overall)

                n = not a valid voter

                t = chair eligible to vote only to make or break a tie

Attendance re-initialized on 2010-04-13:

 v[xxxxxxxx] Laurence Bisht (Intel)

 v[xxxxxxxx] Eduard Cerny (Synopsys)

 v[xxx-xxxx] Ben Cohen

 v[-xx-xx-x] Surrendra Dudani (Synopsys)

 v[-x-xxxxx] Dana Fisman (Synopsys)

 v[x--xxxxx] John Havlicek (Freescale)

 v[xxxxxxxx] Tapan Kapoor (Cadence)

 t[xxxxxxxx] Dmitry Korchemny (Intel ¿ Chair)

 v[-xxxxxx.] Scott Little (Freescale)

 v[xxxxxxxx] Manisha Kulshrestha (Mentor Graphics)

 v[xxxxxx..] Anupam Prabhakar (Mentor Graphics)

 n[--xx-xxx] Erik Seligman (Intel)

 v[xx-xxxxx] Tom Thatcher (Sun Microsystems ¿ Co-Chair)

   |- attendance on 2010-06-22

 |--- voting eligibility on 2010-06-22

Minutes:

--------

1. Minutes from last meeting:

    Move to approve minutes: John

    Second: Ben

    Vote results: 8y, 0n, 0a

2. E-mail ballot results:

    2291, 2330, 2955 Passed

    3113 Failed

3113:

Manisha: It's not clear that re-writing algorithm applies to let.

     If language is clarified, this should be clear.

John: Are there technical issues w/ applying rewriting algorithm to

     checkers or let?

Manisha:   Difference between sequence and property arguments:

     Sequences cannot take a property argument

     Checker arguments should be similar to property arguments.

     Let arguments should be a subset of sequence arguments.

John We should look at and think about it.

Dmitry: Can we separate 3113 from new issue 3117?

John: Move to approve 3113

Second: Manisha

     Vote results; 8y, 0n, 0a

New Issues:

Dmitry: Note that PAR is approved.

John:  Any feedback from Working Group?

     Had heard that Karen Piepers had expressed concern about how AC

     had operated during the last PAR

Dmitry: We did slip schedule of last PAR because of significant

enhancements

     That turned out to be complicated and all interdependent.

     In current PAR, we can tread most issues separately.

     We need to work more closely with other committees.

John: We did have trouble last time getting other committees to think about

     our enhancements.

     We need to do better in this respect this time around.

3099: Action block triggering is not well defined.

3117: Make it clear that rewriting algorithm (F.4.1) applies to checker

     and let

     (Created in response to issues from 3113)

Manisha: Will write a proposal for 3117

John: We need to think about it carefully, as we are applying re-writing

     algorithm to two new structures: checkers & let

Manisha: John, can you review the proposal?

John: If you come up with a proposal, I will review it.

Dmitry     3120 "expect" construct to refer to virtual interfaces

Ben: In OVM/UVM/VMM you use classes. These classes use virtual interfaces

     to tie the test bench code to the RTL. An question was asked in

     the Verification Guild about whether an expect could be written which

     refers to signals in one of these virtual interfaces

     One vendor implements this: allows expect w/ virtual interfaces.

     Other vendors don't allow this

Ed:  Immediate asserts can be used on virtual interfaces

Ben: Expect is different from immediate assert, because expect will wait

     for a pattern to match

Ed:  Problem in the example is also that the variable was an automatic.

John: Some concerns: Entered a previous mantis item regarding expect

          Thinks that there are some things broken with expect.

Ben: If interface not virtual, there would be not problem.

     Why should it be an issue if it is virtual?

     OVM and UVM will all use virtual interfaces in classes

     The only thing that will work with classes is immediate

     Could we allow this use case, but just restrict expect to access

     only virtual interfaces?

John:  Can't remember what his problems were with expect.

     Will review

Manisha:   Expect is already supposed to handle automatic variables

     But handle for virtual interface may be null. This is not currently

     defined. In Verif Guild example was the expect on a clock interface

     or signal?

Ben  Can a virtual interface be NULL?

     Will show how the expect was used in the example.

Issue resolution:

2732: Clarify timing diagrams in Fig 16-4

John:  The current proposal just clarifies what we had intended for

     reporting the assertion failure time for assertions with future

     time functions.

Ed:  Don't like the requirement that tool must report earlier time

     Requires storage to remember the time that the assertion failed.

     It's doable.

John:  What is your proposal?

Ed:  Simpler: Report the current time when the message is being printed.

     We had to wait for the next cycle to see if the assertion actually

     failed, we should just report this time.

John: This requires user to think about when the assertion actually failed.

Dmitry: What about DPI behavior If you implement callbacks how can you get

     the correct time for the assertion failure? Not implementable in DPI.

John:  You just have to wait to see if the assertion failed.

Ed:  The time reported by DPI is the time when the DPI function is called.

     It is different from the time of failurefailure

John: Yes, but you'll have to figure that out somehow.

Ed:  If the callback asks for current time, it will get current time, not

     the time that the assetion actually failedfailed

John: We can't force omnicience.

Dmitry: I will call for e-mail vote:

2362 16.14 Mention of assertion control system tasks is unconnected

Ben: I agree with proposal

Tom: Move to accept proposal

John: Second

Vote results: 8y, 0n, 0a

2825:

Ben  Will write a proposal

2756:

John: Nothing wrong with LRM. However, the original LRM was stricter.

     2009 liberalized clock inheritance, but examples not changed.

     All examples demonstrate code complying with stricter 2005 examples.

Ben: If there was an example with looser clock restrictions, it would be OK.

Tom: Will write a proposal

Ed:  On p. 416, you can see that the rules are explained.

2927:

Ed:  Table 16.3 seems to have both precedence and associativity:

Anupam:    Question is about precedence when regular expression operators

     combined with property/sequence operators

Ben  Don't see the point

     Logical expression have precedence over sequence operators.

Tapan:     Precedence s only only needed when there are two ways of parsing an

     expression. BNF rules dictate the rules for processing expressions

     with both logical operators and sequence/property operators.

Dmitry:    Would it make things more clear to specify

Dmitry: Who want's to own it?

     Solution could be to just write a note on Mantis item explaining

     why no change is needed to LRM.

Ben: Will take this one.

2452: No vacuity information about synchronous aborts

Dmitry:    Assign to Dana for now. She can reassign it later

Dmitry: Other Mantis items:

Dmitry: Should I just assign people to mantis items so that we don't spend

     meeting time?

     I can allow two days for people to volunteer. After that, will assign

     people to mantis items.

Tom: Will write a proposal for 2722.

Ed:  Next meeting: Synopsys will be shut down. Won't be able to start

     phone conference.

Dmitry:    Will use the Intel phone conference, unless other people volunteer

     to arrange a phone conference.
Topic revision: r1 - 2010-06-25 - 17:58:54 - ErikSeligman
P1800