P1076 October 2, 2014 Meeting Minutes

Attendees:

• PeterFlake * CliffordWalinsky * MartinThompson * JimLewis * StevenDovich * RyanHinton * DavidBishop

Agenda:

Action Item Review

Meeting Discussion

• Done: Peter: Subtypes and Functions - add issues from previous meeting to proposals- in progress
• Jim: Package based implementation of mod (record with modulus and integer value) - in progress
• Martin: Integers and Logic Operators - in progress
• Jim: Add warning to ?? on meta value
• Trace back information from warning in package (such as numeric_std) to line number in entity that caused it.
  • VHPI has this information.
  • 'instance_name does not work since it goes through the library rather than the call stack
  • Define stack trace to be a string produced by VHPI full name
  • Could be a standard DPI library to call VHPI applications.
• Jim: Prepare ranking sheet
• Martin: BigNum Proposal
  • A BigNum library type based on integer_vector (or other array type) cannot be used in for loops or indcies of arrays, where as a built in type potentially could.
  • Addressing a large RAM or disk drive in simulation.
  • What type of uses are BigNums used for: Are indexing and looping important?
  • How do you express a literal for a BigNum? As a string value or as an Universal Integer?
  • Python integer type is BigNum and uses regular integer values
  • Java has a BigNum package and uses string values or arrays of bytes
  • see: Wikipedia Arbitrary Precision Arithmetic
  • BigNum libraries are typically dynamically sized.
  • Integer_vector would not allow for dynamic sizing
  • Want to be able to either constrain an object or not - with a range constraint
  • In simulation a BigNum can be unconstrained
  • BigNum'high and 'low will error or return a value if the object is unconstrained
  • Will need something like ieee.math_real to help translate between BigNum
  • Should conversions between BigNum and Integer be implicit or explicit (type conversion like integer and real)?
    • Implicit conversions may cause issues with package overloading
• Arbitrary width real: see Nov 8 2012 meeting
  • Arbitrary width reals? Not built into any language.
  • IEEE 754-2006 allows for arbitrary width float
  • Currently real is a double and this is ok.
• Tabled - Multi dimensional Array restructuring: Array <--> Matrix Transformations 2012_MeetingJuly19 MeetingDecember15
  • Matlab has built-in reshape functions
  • In VHDL - explicit vs implicit defined operator
  • David Bishop: Requiring an implicit operator could delay implementations
  • Can this be handled by anonymous types on interfaces or generics on a package?
  • Language facility for creating implicit operators?
• Ryan - Slicing Multidimensional Arrays (FT15) maybe related to above 2012_MeetingMay24
• Jim - Mark bugzillas for which we have a proposal with that status
• Jim - Table driven modeling via protected types 2012_MeetingMay10
• Steven: P1735 submitted to RevCom. VHDL work ready to start

Ranking Criteria

• Rank based on need
  • Item needs to be fixed
  • Useful for an application
• Rank based on complexity
  • How hard to implement
• Noted several items in Collected Requirements as Must Do
• AI: Jim: Create ranking sheet.

Review and Approve Meeting Minutes:

• Motion: Steven 2nd: Cliff

Next Meeting: Thursday October 16, 2014 8 am Pacific

Previous Meeting: Thursday September 25, 2014