JLAB PLANS (Robert)

1) General improvement of szin code support for cluster
level computing. Improving the code library to be more
readable and I'm even attempting documentation.
E.g., a new user can more easily find their way around
the code. Some of the UKQCD people (Tony Kennedy and co.)
want to use it. One of the things that has been done is
there exists a very easy way to drop in calls to optimized
routines (QCD lib.?) in place of my current macro generated
versions based solely on the names/types of the operands.
The code currently supports threaded nodes with master control
call to MPI or whatever (this is the only QCD code I know
that can do this).

2) With a (very bright) high school student Chris McClendon,
am extending AVP's current scheme code for producing
optimized alpha assembly for single node code fragments.
Following the discussion of a message I sent you before,
these single node fragments are used to build a parallel
Dirac operator. Andrew must have understood that previous
mail message enough to also be targeting the same primitives,
but in a much more sophisticated manner. I don't know how
long AVP will take, but we are moving quickly.

2.5) As a side note, this same student is graduating now.
He has a summer internship at the lab and should be working
with David and I. He is an available resource - if there are
suggestions for priorities, let me know.

3) Using my current parallel Dirac operator framework,
making a new P4 version that utilizes the essential bits
of Luscher's single node P4 Dirac inverter.

4) A side note, all of my cluster level parallel code goes
through my own small (lightweight) generic communication
primitives (all in one short file). It currently wraps on
top of MPI.

5) Developing a more optimized communication myrinet driver
utilizing the BIP code.

6) Development effort with Tony K., Balint Joo, etc. to develop
a Aldor QCD code. This is the IBM/NAG computer math language.