Software Coordinating Committee Conference Call
		       December 20, 2001 3:00 PM EST
			  Recorder: C. DeTar

Present:

  Mawhinney, DeTar, Edwards, Mendez, Holmgren, Gottlieb, Simone, 
  Brower

Absent:
   
  Pochinski, Watson

** Action items

    1.) PR task. There is a SciDAC PI meeting in January. Bob Sugar is
        requesting a few slides on the Software Project for a mandatory
        QCD poster. 

** All to make suggestions and encouraged to construct contributions.

     2.) Improved Web page & Documentation for SciDAC project. 	

** Robert to work on setting up an lqcd.org for a web site

     3.) Level 2 Design (see Carleton's second trial ballon on data
         layout).

         Robert: Discussion of lazy shift feature.
           Mark data source and result as busy.
         Carleton: But then we must also provide a way to finish
           off the shift if a user needs to revert to Level 1.
         Carleton: What calls should we provide?  More than
           one-link parallel transport?
         Brower: Likes an atomistic approach to allow
           maximum flexibility.
         Bob M: Don't we lose the ability to optimize?
         Robert, Carleton, Richard: Level 2 is mainly for
           convenience with optimization where feasible,
           but obviously not the most efficient.         
         Bob M: Wouldn't we get terrible performance if we
           have to figure out what sites are in the subset
           on each call?
         Bob M: So on the QCDOC we use only one layout, but
           the others will still work?  Another possibility
           is to require a certain layout and implement the
           other Level 2 routines only with that assumption.
         DeTar: First pass on QCDOC could hard-wire the layout.
         Brower: What to do next?
         DeTar: How about make a more detailed list of function
           calls, data types, and objects that belong in
           Level 2, update Robert's Level 2 Conceptual Design.
** DeTar, Edwards, Brower to work on this over the next week.
        
Next call Thursday, December 27, 3 PM EST
Following Thursday, January 3, 1 PM EST ???

Conference concluded at 5 PM EST.

From detar@physics.utah.edu  Thu Dec 20 16:57:15 2001
Date: Thu, 20 Dec 2001 14:57:02 -0700 (MST)
X-Authentication-Warning: dirac.physics.utah.edu: detar set sender to detar@physics.utah.edu using -f
From: Carleton DeTar <detar@physics.utah.edu>
To: brower@lns.mit.edu
CC: cmendes@cs.uiuc.edu, detar@physics.utah.edu, djholm@fnal.gov,
   edwards@jlab.org, watson@jlab.org, rdm@cuphy5.phys.columbia.edu,
   avp@lns.mit.edu, simone@fnal.gov, sg@fuji.physics.indiana.edu,
   doug@klingon.physics.arizona.edu, gregory@klingon.physics.arizona.edu,
   osborn@physics.utah.edu
Subject: Minutes, December 20 Conference Call

	   Software Coordinating Committee Conference Call
		       December 20, 2001 3:00 PM EST
			  Recorder: C. DeTar

Present:

  Mawhinney, DeTar, Edwards, Mendez, Holmgren, Gottlieb, Simone, 
  Brower

Absent:
   
  Pochinski, Watson

** Action items

    1.) PR task. There is a SciDAC PI meeting in January. Bob Sugar is
        requesting a few slides on the Software Project for a mandatory
        QCD poster. 

** All to make suggestions and encouraged to construct contributions.

     2.) Improved Web page & Documentation for SciDAC project. 	

** Robert to work on setting up an lqcd.org for a web site

     3.) Level 2 Design (see Carleton's second trial ballon on data
         layout).

         Robert: Discussion of lazy shift feature.
           Mark data source and result as busy.
         Carleton: But then we must also provide a way to finish
           off the shift if a user needs to revert to Level 1.
         Carleton: What calls should we provide?  More than
           one-link parallel transport?
         Brower: Likes an atomistic approach to allow
           maximum flexibility.
         Bob M: Don't we lose the ability to optimize?
         Robert, Carleton, Richard: Level 2 is mainly for
           convenience with optimization where feasible,
           but obviously not the most efficient.         
         Bob M: Wouldn't we get terrible performance if we
           have to figure out what sites are in the subset
           on each call?
         Bob M: So on the QCDOC we use only one layout, but
           the others will still work?  Another possibility
           is to require a certain layout and implement the
           other Level 2 routines only with that assumption.
         DeTar: First pass on QCDOC could hard-wire the layout.
         Brower: What to do next?
         DeTar: How about make a more detailed list of function
           calls, data types, and objects that belong in
           Level 2, update Robert's Level 2 Conceptual Design.
** DeTar, Edwards, Brower to work on this over the next week.
        
Next call Thursday, December 27, 3 PM EST
Following Thursday, January 3, 1 PM EST ???

Conference concluded at 5 PM EST.