(r1) 2013_MeetingJanuary31 < P1076

P1076 Web>Vhdl2019MeetingMinutes>2013_MeetingJanuary31 (revision 1) (raw view)~~EditAttach~~
<!--
   * Set ALLOWTOPICCHANGE =  P1076AdminGroup
-->
---+ P1076 January 31, 2013 Meeting Minutes
---++ Attendees:
   * Main.CliffordWalinsky * Main.PeterFlake * Main.JarekKaczynski * Main.JimLewis

---++ Agenda:
   * Review VHDL-2008-Performance remaining list items
   * Review VHDL-2008-MP remaining list items
   * Review Accellera TC remaining list items
   * See these items in the [[http://www.eda.org/twiki/bin/view.cgi/P1076/Vhdl2019CollectedRequirements][collected requirements list]]

---++ DPI & SCE-MI
   * Peter met with SCE-MI 
      * For dual languages, they want a DPI to SystemVerilog Data Types
      * Wanted to know time line for it to ge in the standard
   * DPI to ADA?
---++ 64 bit Integers and Beyond, and Real
   * Extend integer to full 2's complement range
   * Need bigger integers for coverage modeling and randomization
   * Separate type or subtypes?
   * integer subtypes & subprogram elaboration will need to be considered
   * Need someone to start a proposal page and flow above into this
   * What does Ada do?
---++ Performance Proposals
   * Perf 1: lots of concern about removal of delta cycles 
      * NN much performance gain
      * reject
   * Perf 2: Expressions in sensitivity lists (posedge) 
      * Isn't a compiler smart enough to extract this anyway - especially if it can handle process (all).
      * reject
   * Perf 3: 2 & 4 state semantics 
      * Simulate std_logic using 2 states (0, 1) or 4 states (0,1,X,Z)
      * Could this buy us anything?
      * Where would it be specified to do this? Is it a tool thing?
      * Does bit simulate faster / use less memory that std_logic?
      * Why can't a compiler/simulator do this automatically for RTL code? 
         * If I cannot do it, can I mark it somehow?
         * Simulator flag or marker in code?
      * Further Action Required
   * Perf 4: Light-weight signals 
      * Should be handled as a simulator optimization!
      * Is this shared variable with regular types?
      * Does it have a delta cycle?
      * Need to have a use model that interests us
      * reject
   * Perf 5: Atomic Composite Signals 
      * Normal composites track events on each subelement
      * Atomic composite only tracks events on the composite
      * With Atomic do I get element resolution functions or do require a resolution function for the entire composite?
      * Can simulators already do this? 
         * To some degree. If elements not accessed
      * Are there use cases that the compiler needs help with? 
         * What sort of help does the compiler need?
         * Will this provide any benefit?
---++ Action Items:
   * none
---++ Review and Approve Meeting Minutes:
   * Motion: Cliff 2nd: Peter
---++ Next Meeting Date (proposed):

Thursday February 21, 8 am Pacific
Topic revision: r1 - 2020-02-17 - 15:36:12 - TWikiGuest
P1076