Software Coordinating Committee Conference Call
		       December 9, 2004 3:00 PM EST
			  Recorder: C. DeTar

Present:

  Brower, Osborn, DeTar, Maresca, Zhang, Jung, Edwards, 
  Mawhinney, Renner, Holmgren, Watson

Absent:

  Petrov, Pochinsky, Simone, 

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


	1. The usual update of Software Tasks: QMP/QIO/Level 3/...

QCDOC/QIO

   DeTar:     James, you were having trouble with QIO and MILC on QCDOC.
              Any progress?
   Osborn:    I am working with Chulwoo.  I will focus on QIO.
** Renner:    I will work on the MILC problem.

   DeTar:     Any progress on the host file conversion?  Is it Kostya's job?
   Mawhinney: I believe so.  I don't know where it stands.
** DeTar:     I'll check with Kostya.
   Brower:    How is UK doing it?
   Mawhinney: Each node does its own I/O.
   Mawhinney: We have found that 4 proc IBM 615 can handle 1024 NFS
              mounts without problem.
              8 proc IBM 655 will be used for QCDOC Riken and BNL
              We have all 1024 nodes reading from and writing to a
              single file.  We are hoping to do it with Linux boxes.
              The 655 is expensive.

QCDOC/Level 3

   Brower:    How about Asqtad Level 3?
   Jung:      I am cleaning up -- converting from CPS to standalone.
              How to test?  Can I have a standalone test routine?
** Renner:    I can modify Eric's code to call Level 3.  I can work
              with Chulwoo.

1024 node test machine:

   Brower:    When will we have 1K node machines for alpha testers.
   Mawhinney: Maybe for short tests (hour or so) in early Jan on a 
              RBRC machine
              End of January for a BNL machine.
   Brower:    It would be very good to have the alpha testers run.
   Mawhinney: There is no DOE-purchased 1K machine right now.

Testing/Benchmarking:

   Zhang:     Morton will be collecting data from QMP profiling library.
              I am communicating with Ken Roche.
   Brower:    Why don't you forward it to me as well?
   Zhang:     OK
   Zhang:     I am benchmarking MILC on Ultrix/ORNL Everything seems to be
              fine, except that I couldn't malloc a 16^4 per node problem

	2. Recommended focus for Software Workshop 

** Brower:    I will compile an agenda for comment

     [ Some discussion of start times led to
       deciding to start Saturday afternoon with the main
       topic Sat afternoon to be running on the QCDOC. 
       Then a long day on Sunday. ]

Plenary call concluded at 4:00 PM.

Subcommittee discussion of MILC benchmarking followed:

   (Holmgren, Zhang, Osborn, Brower)

  Zhang:    I started working on SvPablo profiling on 
            Pentium 4 and Xeon.
  DeTar:    Good.  Celso started with that, but was held
            up because of lack of proper hardware support.
  Holmgren: There was no PAPI support for perfcounter 
            so we couldn't get the numbers we needed.
  Zhang:    I am running on Altrix for Roche.
  Holmgren: We have a small Itanium cluster, so I would be
            interested in the Ultrix results.
  Holmgren: You could come up to FNAL for a face-to-face meeting.
  Zhang:    Perhaps I should spend some time in Jan meeting
            with James.
  Brower:   We could set something up for Saturday morning.
  Zhang:    What platforms other than FNAL?
  Osborn:   JLab.  Also seaborg.

=====================================================================
Committee conference concluded at 4:15 PM EST.  
Next plenary call December 9 at 3 PM EST.