RE: [sv-bc] DataTypes: BNF changes

From: Mark Hartoog (Mark.Hartoog@synopsys.com)
Date: Wed Nov 10 2004 - 16:12:59 PST

  • Next message: Mark Hartoog: "uwire data type and System Verilog"

    I take it from reading the minutes that the restriction is that net type
    and 'reg' are not allowed next to each other, but if a drive strength or
    charge strength appears in between, then it is legal. For example these
    would all be legal:

    tri (large) reg w;
    typedef reg REG;
    tri REG w;

    but this would be illegal:

    tri reg w;

    Is this correct?

    > -----Original Message-----
    > From: Steven Sharp [mailto:sharp@cadence.com]
    > Sent: Wednesday, November 10, 2004 3:42 PM
    > To: Dave_Rich@mentorg.com; Mark.Hartoog@synopsys.COM; sharp@cadence.com;
    > btf-dtype@boyd.com; sv-bc@eda.org
    > Subject: RE: [sv-bc] DataTypes: BNF changes
    >
    >
    >
    > >> The datatypes groups voted to disallow <nettype> followed by 'reg' as a
    > >> lexical restriction.
    > >
    > >I cannot find this restriction anyplace in the proposed BNF, chapter 4 or 5
    > >changes to the LRM.
    > >
    > >Is it an oversight that this was left out?
    > >
    > >How exactly was this restriction formulated?
    >
    > I was asked to check with Brad Pierce about where it should appear, since
    > the BNF was one possibility. Brad expressed the opinion that it should
    > appear in the text instead. The best place would probably be the new
    > sub-section in section 5 on nets. It would be nice if the syntax boxes
    > were broken up so that the syntax for variable declarations appeared in
    > the sub-section on variables, and the syntax for net declarations appeared
    > in the sub-section on nets. Then the syntax would be right there and the
    > restriction could appear in the text after it.
    >
    > The restriction is supposed to be against having the net_type token
    > immediately followed by the "reg" token. An alternate restriction that
    > the group considered was just disallowing "tri" followed by "reg". Enough
    > people felt that constructs like "wire reg" were also confusing, that once
    > the restriction was made, it might as well be broadened.
    >
    > While this combination could have been prevented grammatically, it would
    > have required some hideously messy changes in the BNF. It was deemed
    > easier to express it as a separate lexical restriction against having
    > this sequence of tokens. It isn't pretty, but nobody came up with a
    > suggestion that people liked better.
    >
    > Steven Sharp
    > sharp@cadence.com

    This archive was generated by hypermail 2.1.4 : Wed Nov 10 2004 - 16:00:26 PST and
    sponsored by Boyd Technology, Inc.