VHDL-200x Meeting
 Held 08-16-2004 14:00-17:00 PDT

Attendance

Charlie Guy
Eric Marshner
John Ries 
Steve Bailey
Peter Ashenden
Deepak Pant
Ajay Varikat


FT10A

   The grammar is ambiguous with itself.  The grammar is linear, i.e expression-N
   and condition.  The grammar needs to be changed. One way is the condition
   needs to be replaced with BOOLEAN_logic_expression. Another proposal is to
   change condition to a boolean logical instead of boolean expression. A plus
   for this proposal that conditions well defined. 

   Eric asks why not ?:.  Some people don't like it for aesthetic reasons. Another
   is reason is ?: is just a two way choice and get cumbersome for multi-way
   choosing.

   Result: condition must be change to a logic expression no just an expression.

   Example on line 37 is wrong, missing "else" value

   Eric: change production

   nnary_expression ::= 
     { logical if condition ,} logical


   Was some question of waveforms and n-ary expression but doesn't appear to
   have any problems.

FT10B

  Should just be removal of n-ary usage.

  Peter: Why was simple assignment factor out. John: because we need to define
  the basic sequential signal assignment that everything is transformed into.

  Line 110 is incorrect, Guarded is an option, not part of the
  waveform. Statement needs be changed to "It is an error if the reserved work
  guarded appears as an option in the sequential form of the signal assignment"

  Issue on how guarded is specified and handled in the transform.  Eric brought
  up issue that the grammar for sequential conditional assignment should be left
  recursive.  Discussion of why the grammar is as it is.  Do we want the ability
  not to have to have an assignment value?  One issue is how much to change in LRM
  vs regularize the LRM because we are making similar structure.  Eric argued
  that we should keep concurrent and sequential signal assignments apart and
  make the common syntax through duplication of text.

  We should have NOTE specified in line 185.

  Proposal is close to done.

IP proposal

  Outstanding issue

  1) Lexical vs Language structural.

  2) Specify encoding, six-el vs base64

Lexical vs Language structural

  Adding the protect constructs to the grammar makes the grammar very messy and hard to do,
  because of all the place the constructs are allowed.
  Lexical has an issue with where pragma occurs.

  What does lexical means.  Means encryption can start and stop at any lexical boundary.
  In proposal pragma is allowed anyplace a comment is allowed.

  Peter: We don't specify how new lines are specified. Does this cause problems.
  The decryption code may need to handle conversion of NL to CR/NL or CR/NL to
  NL.

  When specifying the encryption algorithm does it remain in effect for multiple
  protect blocks?

  Steve: does the protection in context unit apply to all units that use that
  context unit?  No, applies only to the file.

  Peter: pragma definition is -- paragma protect, do we want to use a separate
  pragma construct.  Should we use ` or % and define a pre-process directives
  instead. Should we use something not in verilog so we can have common
  directives. Directive doesn't have to be a single character. Other pragmas
  could use this mechanism.

  Need to specify default encryption and algorithm.

  Charlie: are we specifying the standard info in the VHDL LRM or someplace
  else?  Short term it is not possible to put together is separate standard that
  Verilog and VHDL reference.  The issue here is inter-operable between Verilog,
  VHDL, PSL.

  Need to identify all algorithms that can be used as part of the standard
  support.  Need to allow implementations to add algorithms.  For encryption and
  encoding algorithms specify at least one version that what everyone must
  support. Provide enough information that it can be implemented.
  I.E. supply reference to algorithm.

Type generics:

  Any open issues here?

  Bhasker is to busy to do any more testbench work.  We need to code up versions
  of testbench work that work with type generics.

  Could provide these as freeware or public domain packages.

  Peter will look at implementing some of these data structures.

Other Business

  Peter is confused about real status.

  Some items like FT10A was marked done, but we did review of it today.

  Jim please review status.  Can't be "Done & Reviewed" or "Done & Being
  Revisit".  Need to have a statuses of Proposed, Review, Deferred, and Done,
  might have sub statuses.  Like reviewed, but needs modification. and so forth.