Software Coordinating Committee Conference Call
		       July 1, 2004 3:00 PM EDT
			  Recorder: C. DeTar

Present:

Guests:
  Brower, DeTar, Holmgren, Simone, Gregory, Mendes, Edwards, Watson
  Neff, Jung, Osborn, Mawhinney (joined late), 

Absent:

  Pochinsky, 
  

** action items
=====================================================================

We were having problems with the phone lines.  It seems only 7 lines
were open, so some callers couldn't join.

1) Credit for contributed software

   Brower: Peter Boyle was anxious to get credit if his
      software is used.  We should certainly help young
      researchers this way.  Perhaps some day we will
      write an article describing the code with a full
      author list to include in a citation policy.

*** Watson give input on Software "Copyleft" policy.

2) Priorities for Level 3

**
   Odd numbers of sites per axis.

   DeTar: The immediate need is to be able to put a 40^3 x 96 lattice
          on the QCDOC
   Brower: Toussaint and DeTar should work directly with Columbia
   [ Discussion was continued after Mawhinney joined.]
   Mawhinney: The existing conventionally cabled machine will run
      40^3 x 96 on 4800 nodes.  The machine has to be split into four pieces.
      Allowing 40^3 x 96 is a physics question as well.  Does the
      program committee agree on running this dimension?
      I am working on a note explaining the partitioning and
      will send it before next Monday.
   DeTar: It makes no sense to wait for the program committee,
      which, in any case, will more likely go with what
      the machine can do at that point.
      We shouldn't be building in artificial constraints on
      the physics we can do.
   Mawhinney: I disagree that we are imposing constraints on physics.
   Brower: How much time would it take to generalize the inverter dimensions?
   Mawhinney: It is more a question of who will do it.
      Perhaps another graduate student.
      It is easier to recable physically than to rewrite the
      software at this point.
   Brower: It would help to have a general discussion of the
      constraints on the dimensions.  Can you include that in your note?
  
3) Fall meeting

   Brower: We need to have a meeting not piggy-backing on other
      meetings. I am willing to host at BU.  September?  Later?
      Please suggest a place and time.
      I'll send a note requesting suggestions.

4) Task priorities
   [see Brower e-mail Wed 30 Jun 2004]

   DeTar and Osborn: 
     (1) QMP 2 for all platforms
     (2) Level 3 interfaces:
           Asqtad inverter
           DW inverter (not my task, but next in priority)
     Osborn: Pochinski was asking for a mapping interface
     Edwards: We need a uniform API for all modules.
     DeTar: That probably means offering both mapping and standard
     Osborn: That shouldn't be much work
     Jung: It adds up, though.
     Brower: Chulwoo, James, and Eric should make a suggestion.
     (3) QDP/XML from Jim Simone
     (4) Packaging we need to put everything on CVS, automake and autoconf
         and put the release in a common location
     (5) QIO
     (6) MILC on QCDOC

   Watson:
     (1) Cluster
     (2) SRM testing/grid work (ILDG)
     (3) Morphable jobs - movable to alternative machine if need be.

   Edwards
     (1) New communications/overlap strategies for QDP++
     (2) New I/O support for mixed mode (binary/XML)
     (3) New QDP parallel-transport support.

     Edwards: Craig McNeile reports a wedding of QDP++ and CPS
     Brower: I was discussing the possibility of a QDP/C and QDP++
       wedding.

   Jung (Mawhinney)
     (1) QMP

     Edwards: What is the status of implementation of QDP 2 on
       the QCDOC
     Jung: Not quite finished.
     Brower: How long to completion?
     Jung: End of summer.
     Brower: Can anyone help you? Kostya?
     Jung: I'll speak with him.
     Osborn: Celso, have you been able to test QMP 2, yet?
     Mendes: Not yet.
     Edwards: As for the GigE version, Jie is out until July 12.
     Brower: QMP 2 seems to be a bottleneck.
     Edwards: So we should work on Level 3 

    (2) Batch system and OS

   Pochinsky:
     (1) Finish optimization of DWF and interfact with QCDOC
     (2) Opteron optimization
     (3) GigE communications - lowering latency
     Edwards: This needs more work

   
=====================================================================
Committee conference concluded at 4:30 PM EDT.  
Next plenary call July 8 at 3 PM EDT.