P1800 Web>SystemVerilogBasicCommittee>SvbcMinutes20160412 (2016-04-26, MattMaidment)EditAttach

SV-BC Meeting Minutes for April 12, 2016

Attendee	Company	Mantis Perms	Mar 15	Mar 29	Apr 12
Shalom Bresticker	Accellera	D	X	X	X
Daniel Schostak	ARM	U
Steven Sharp	Cadence	U	X	X	X
Francoise Martinolle	Cadence	U	X	X	X
Mark Strickland	Cisco	U	X	X
Matt Maidment	Intel	M	X	X	X
Brandon Tipp	Intel	U	X	X	X
Boon Chong Ang	Intel		X	X
Jonathan Bromley	Oracle	U		X
Neil Korpusik	Oracle	M	X	X	X
Gauraw Singh Chandel	Mentor Graphics
Eric Coffin	Mentor Graphics	R	X	X	X
Alex Gran	Mentor Graphics	U
Sharad Jain	Mentor Graphics
C Venkat Ramana Rao	Mentor Graphics	R		X	X
Dave Rich	Mentor Graphics	U			X
Ray Ryan	Mentor Graphics			X	X
Gordon Vreugdenhil	Mentor Graphics	U
Sachin Pathak	Mentor Graphics			X
Dave Scott	Real Intent	U	X
Mark Hartoog	Synopsys	U	X	X	X
Brad Pierce	Synopsys	U	X	X
Arturo Salz	Synopsys	U	X
Justin Refice	nVidia	U	X		X
Karen Pieper	Accellera	M

Call Logistics

April 12, 2016 8am-10am PDT

Action Items

Open

Jan 19 Matt: Track version specific changes to LRM as AI and file if needed.

Jan 19 DaveR: Check with some of the original contributors about the intent of 315 for clarification of 3211.

Apr 12 Matt: Update proposal for 2500 to remove text regarding closure of 4691.

Apr 12 Matt: Add 484 and 305 to next email Vote

Complete

Feb 01 Dave R write a proposal for 2488

Agenda

Review IEEE Patent Policy

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

Drew everyone's attention to the policy.

Opens

Removed 5529. Wrong Mantis.

Approve Mar 29 Meeting Minutes

http://www.eda-twiki.org/cgi-bin/view.cgi/P1800/SvbcMinutes20160329

Mark H moves to approve.

Justin seconds.

Motion passes.

Action Item Review

Done.

Email Vote Discussion

For 2500 Brandon would like to remove the text in the proposal calling for closure of 4691 as a duplicate.

AI: Matt update proposal for 2500 to remove text regarding closure of 4691.

Brandon proposes to close 4691 as a duplicate of 2500.

Dave Rich seconds.

No opposed.

No abstain.

Motion passes.

3548

Brandon was wondering about this comment:

"Because variables declared in automatic tasks are deallocated at the end of the task invocation, they shall not be used in certain constructs that might refer to them after that point" is no longer correct because of fork/join blocks."

Not being in 1967. Matt added this as a note to 1967.

5217

Brandon raised that the proposal does not direct the editor to renumber subsequent sections. Neil believes that this is not necessary based on editing software capabilities and as a regular task for the editor. Brandon is satisfied with this response.

4939

If a non-random enum has a value outside of the named values of the enum will it result in a randomization failure. Does a rand enum have an implicit constraint that its value have to be one of the enumerated values? Should the solver be able to assign non-enumerated values to the enum variable?

New wording headed toward:

Constraints support only the named value set for enum types. In choosing a solution, the selected value for each random variable of an enum type shall be from the corresponding set of named enum values. Values which are outside of the named value set, even if they can be successfully cast to the enumerated type, are not valid solutions.

Note that each non-random variable of an enum type need not have a value in the corresponding set of named enum values.

Justin will continue to refine and upload a new revision.

Mantis items that will be dropped from e-mail vote eligibility:

5540, 5594, 4939 and 2500

AI: Matt send an email on this point to the reflector.

5594 - https://accellera.mantishub.com/view.php?id=5594

Explored issue of interconnect ports and whether or not they are permitted to be declared in the non-ANSI style. The grammar permits it and the feature has been around since 2012. Do not believe it is a good idea/possible to remove it now. Proposal will stay as-is and acknowledge possibility that interconnect ports can be declared using non-ANSI syntax.

Suggest changing:

"The data type may be overridden by a variable or net declaration for the same port."

to:

"This default data type may be overridden by a variable or net declaration for the same port."

5259 - https://accellera.mantishub.com/view.php?id=5259

Working on this via the reflector.

5540 - https://accellera.mantishub.com/view.php?id=5540

Should be taken off the vote. Feedback being collected via the reflector.

Original question: How to deal with rand members in a tagged union?

One approach is to require constraints on tagged unions require they include the tag. Steven raised that this would require a limit to the expression such that the implementation does not have to trace through function calls, etc. to resolve it.

Another approach is if an expression refers to a tag union member then the implicit guard is that the corresponding tag value is valid. Feedback is that this, too, is complicated to resolve for complex expressions.

Other approaches:

o Forbid randomization of tagged unions.

o Forbid randomization of unpacked tagged unions and permit randomization of a packed value.

Steven suggested that if enums had an implicit method to return the valid value set for use in constraints. This would need to wait for a PAR that permits enhancements.

Randomization of union members could be improved in the future through enhancements.

All suggestions seem to point to forbidding randomization of tagged unions.