Re: Special 1364 Meeting to resolve Issue 680 (Configs)

From: Clifford E. Cummings (cliffc@sunburst-design.com)
Date: Tue May 10 2005 - 09:31:00 PDT

  • Next message: Adam Krolnik: "Re: Survey - use-model for configs?"

    Hi, All -

    I could support such an option under the following conditions:

    (1) The already defined config, endconfig and cell are the defaults.
    (2) 1364-2001+config (or whatever it is called) exchanges the keywords
    configuration, endconfiguration and cell name for config, endconfig and
    cell (I can't wait to see the BNF for this - it probably just has to be
    text and no BNF change)
    (3) Vendor can come up with their own command-line switch to change the
    default (but the switch is not required). As far as I am concerned, vendors
    could implement such a switch today without any modification to the standard.
    (4) We simultaneously pass the typo-changes proposal that I sent out on
    5/8/2005.

    Again, if we don' fix the config use-model, then 99%+ of all designs are
    going to ignore configs even with the above modifications, which means we
    are spending a lot of cycles on an enhancement that in its current
    implementation is 99% worthless.

    Regards - Cliff

    At 08:26 PM 5/9/2005, Steven Sharp wrote:
    >I still see a possible compromise that would allow configs in Verilog
    >source files while reducing the keyword conflicts caused by some of
    >the config keywords.
    >
    >We have added the `begin_keywords directive that "turns off" certain
    >keywords to match older versions of the language. We could add an
    >additional version string that selects the keyword set that is allowed
    >in Verilog-2001 source files but not in configs or library map files.
    >This set of keywords would allow compiling almost all Verilog-1995
    >source files, and all Verilog-2001 files that did not contain configs.
    >It would be the more backward compatible dialect of Verilog-2001.
    >There would be another version string that would reserve the same
    >keywords plus the config and library map file keywords. A possible
    >set of version strings would be "1364-2001" and "1364-2001+configs"
    >
    >The same could be done for Verilog-2005, since it also doesn't have
    >any problematic keywords except the ones in configs. However, since
    >the only difference between 2001 and 2005 is the uwire keyword, we
    >could probably do without this.
    >
    >The main goal of having these two keyword sets is not so much that
    >users will directly use the directives to get them. After all, if they
    >are willing to put directives in all their files, they can set the
    >keywords appropriately for each file. The goal is to define a dialect
    >that users can use to compile all their Verilog files (1995 or 2001)
    >without having to use any directives in them. This dialect could be
    >the default of a tool or could be chosen by command line option. The
    >`begin_keywords version string is just a way of defining this particular
    >keyword set in the standard, and ensuring that all implementations
    >support it.
    >
    >Steven Sharp
    >sharp@cadence.com

    ----------------------------------------------------
    Cliff Cummings - Sunburst Design, Inc.
    14314 SW Allen Blvd., PMB 501, Beaverton, OR 97005
    Phone: 503-641-8446 / FAX: 503-641-8486
    cliffc@sunburst-design.com / www.sunburst-design.com
    Expert Verilog, SystemVerilog, Synthesis and Verification Training

    This archive was generated by hypermail 2.1.4 : Tue May 10 2005 - 09:11:15 PDT and
    sponsored by Boyd Technology, Inc.