minutes of Verilog-AMS call 23 May

Attendees:
Geoffrey Coram, Analog
Jim Barby, Waterloo
Martin O'Leary, Cadence
Patrick O'Halloran, Tiburon
Peter Liebman, Xpedion
Marq Kohl, Philips
Arpad Muranyi, Intel
Sri Chandra, Freescale

Since Graham was not on the call, we discussed the updates 
to Section 3: Datatypes.


Many of the syntax boxes had change bars all through them
because the grouping of elements had changed, not because
of any change to the intrinsic syntax itself.


Martin asked about the constant_concatenation used to
initialize parameter arrays.  SystemVerilog has an "array_literal"
syntax item that includes a backtick ` for this purpose; we
need to provide a forward migration path.

Action item: Geoffrey to check SV's array literal and pass
information on to Graham for modification of Annex A.


We talked a little about conversion of strings to real;
this was previously an open issue as something that wasn't
wanted.  The 2.2 LRM says it's an error to assign a string
literal to a real or integer parameter, but this is 
somewhat "un-Verilog-like"

Action item: Geoffrey to check (with Shalom?) what SV
does when converting to/from strings.


Mantis 1405 is resolved by the modifications made to
merged_datatypes on page 37.  (Later in the call, it was
noted that the sig_flow_i discipline example on page 47
also needs to be corrected.)  Sri recalled some questions
about convergence problems with flow-only disciplines,
eg if you have only one connection at a node.  However,
this is a design problem, and nothing in the LRM prevents
the corresponding design problem of connecting two unequal 
potentials in parallel.


Ground (and wreal) declarations now accept a discipline
identifier; this was done so that a reference to V(out,gnd)
will have both nets belonging to the same discipline.


The change bars are not visible in the syntax boxes on
pages 44 and 48.


Action item: Geoffrey to review the chapter for references
to 1364-1995 and update them to -2005.


Sri talked with Johny Srouji (Accellera technical chair)
to say that LRM 2.3 will not be ready for DAC 2006 but we
still hope to be done in Q4.


Next week, we will review Graham's changes to merged_lex;
since we've already made one pass through, we hope to also
have another chapter ready for review.


-Geoffrey