Software Coordinating Committee Conference Call
               October 17, 2002 3:00 PM EDT
                  Recorder: R. Edwards

Present:
  Osborn, Mendes, Edwards, Petrov, Mawhinney, Watson

DeTar, Simone (attended at 4pm)

Absent:

  Brower, Holmgren

** Action items

----------------------------------------------------------------------
Agenda:

Richard is away and asked RGE to shepard the usual conf. call.
The agenda (taken from Richard's previous mailings):

1) Update on QDP C Level 2 API (DeTar)

2) Update on QDP C++ API  (Edwards)

3) Update on test code for QMP (Simone and Mendes)

4) Update on I/O.
----------------------------------------------------------------------

3) Update on test code for QMP (Mendes)
   Mendes: Reported on using DejaGNU for software testing of QMP. Overhead
     appears high. Looking at other approaches.
  
2) Update on QDP C++ API

   More discussion on syntax issues of communication/subsets
  
   Mawhinney: would like to see discussion of implementation strategies
     for the Wilson-Dirac operator and plaquette
  
** Edwards: will try to put together a sample (not functional)
     implementation strategy for next(?) conf. call

** Edwards: will send out in another mail a summary of syntax for current
     debate on C++

3) More on QMP

   Mawhinney: discussed potential problem in QMP. Lack-of description in
     spec about a send from "A" to "B" followed by another send from
     "A" to "B" without first checking for completion of first spec.
     Prevalent in MILC code

   DeTar: can simply insert a hold or block till first comm. finishes on
     QCDOC - not performant.

   Osborn: a performant bypass of the problem is to combine messages into
     a multi-message.

** DeTar: send out a further discussion on the issue

----------------------------------------------------------------------
Conference concluded at  5:00 PM EDT. 
Next plenary call Oct 25 at 3 PM EDT.