Fast Track meeting 10-25-04
Attendance
Eric Marschner
John Ries
Charlie Guy
Dave Bishop



Quick overview of open action items:
  FT-22 From Eric the open issue really isn't locally static subtype. Its 
        locally static array bounds for locally static expressions that are
        composites.

  FT-23 Subtype information was to allow for implementation of jump tables. Addition 
        after meeting the issue here is only with selectors that are
        one-dimensional arrays and the requirement that they be a locally static
        subtype. The only thing really needed is the size of the index constrain
        needs to be known and this can be inferred from the choices.

Continued going over the numeric standard proposals.

N-008
Discussed how literals are now handled.  Min and Max routines are used for
bounds.  These have been modified to detect MIN_INT and return a NULL array.

Eric and David had a discussion on the use of subtype to convey
information. Eric is wondering if it really works or are we make problems for
ourselves.  Eric suggested that the proposal be for built-in.  fixed point
numbers.

Multiplication, addition and subtraction don't have overflow.  Division can have
underflow so an explicit division function is provided allow for the user to
control how underflow is handled

N-009
754 specifies 16 and 32 bit with, used for fphdl32_pkg and fphdl64_pkg
854 specifies variable widths, used for fphdl_pkg

Floating point packages are layer, fphdl_base provides actual functions that do
the work, The specific package like fphdl32_pkg provided the fixed sized real
numbers.

Eric stated that this would be good candidate for generic packages. Dave points
out that if you include the generic package multiple times then floating point
types are not directly visible.  This may not be a problem, the user would just
have to us qualified names for the types.  The operators would be selected base
on overloading because the functions in the packages would not be homographs of
each other.  Action item for Peter. Is there any other problems using generics
packages here

Base packages has two implementations. One is for synthesis and the other is
based on using REAL types.  The REAL types package can't handle some forms of
floating point numbers, IE if the floating point number is bigger than the REAL
type.

Standard contains fphdl16_pkg.vhd, fphdl32_pkgs.vhd, fphdl64_pkg.vhd. The
package fphdl128 is just an example.  The package fphdl_pkg is for variable
sized real numbers

Finish reviewing of posted Numeric standard packages.

New proposal N-010 
Reviewed.


Assuming meeting next week.