Software Coordinating Committee Conference Call
		       October 12, 2001 11:00 AM EDT
			  Recorder: C. DeTar

Present:

  DeTar, Brower, Mendez, Holmgren, Watson, Edwards, Pochinsky

Late:

  Mawhinney 11:50 AM

** = action required

**   members to send topics to be included in software presentation

  1.  Software meeting invitation list

      Look over Rich's list of names -  want people who are
      being paid by SciDAC and who are writing code.
**    DeTar to get Toussaint postdoc on list.

# 2.  Complete Specification of Level 1 MP_API (Chip)

      Agree this is mostly settled now.  
      DeTar: But feels the logical machine part needs further discussion.

      DeTar: We need to have a reply from Columbia about the suitability
      of the MP API.
      Brower: Status of Columbia implementation of MP-API on QCDOC?

      Mawhinney (with Chulwoo): Implementing basic QCDOC calls so they
      are as close as possible.  Some additional features - support
      aggregate call to start multiple (up to 24) sends and receives.
**    Need to look more closely at current MP API to see that it
      conforms.

      Watson: Hope we are not building in impaired performance
      of the API on QCDOC.

      Mawhinney: No, the API will get the very best performance.
      But we may need to add some extra features - e.g. a priority
      flag for a communication channel.  

**    Mawhinney: Hopes to have these issues in concrete form by meeting.

# 3.  Status of Massaging API implementation:

      (a) MPI implementation of Level 1 MP_API   (Edwards  due Sept 27 )
**    80% done.

      (b) MILC com_XXX.c on Level 1 MP_API       (Carleton due Oct 11  )
      This is done as far as we can go now without an MPI implementation.
**    DeTar will send Rich the current com_XXX.c within a week.

**    DeTar: James Osborn could help with the MPI implementation.

      What about SMP and threads?  
      Edwards: Let's punt for now - MPI and GM already handle this
      for us.
          
      (c) QCDOC implementation of Level 1 MP_API (Bob M due Oct 18     ) 
      Put on future level.
      
# 4.  Major new task for committee: Use of "vertical slice" as a prototype 
      of Level 1 and 2 Linear Algebra API.  (Robert & Andrew)
      Robert: Discussion of Level 1 V 0.1
      Pochinsky: Discussion of sample vertical slice code.  See
      http://www-lns.mit.edu/~avp/QCDAPI-sample/sample.c
      http://www-lns.mit.edu/~avp/QCDAPI-sample/conj.c
      http://www-lns.mit.edu/~avp/QCDAPI-sample/dirac.c
      etc.
      These routines call the level 2 variants of Robert's
      level 1.
      
Next call: Thurs Oct 18 3 PM EDT
Call concluded at 12:30 PM EDT.