JLAB PLANS (Chip)

Here are (1) some thoughts on message passing API work,
and (2) additional details on what my group is doing

(1) Message Passing Sub-tasks
    a) implement efficient message passing for myrinet,
        aiming at minimal processor work and message latency
        the resulting low level calls may NOT have any relationship
        with MPI (i.e. NOT a subset) == task 2.1
    b) analysis / design of a lattice QCD message API which is
        able to meet need of SZIN, MILC, CPS, and expected
        future applications, which is composable from the
        message primitives of (a, 2.1).  == task 5.1
    c) implement this API on myrinet == task 2.2
    d) implement this API on QCDOC == task 2.3
    e) test performance of SZIN using optimal algorithms
        == task 1.2
     f) implement this API on MPI, to allow portability to other
        architectures already supporting MPI; this implementation
        need not guarantee high performance, only portability
        == task 2.4

As compared with your description, this flow shows explicitly
the dependencies of these subtasks.  Also note that the MPI
API is not a design constraint.

(2) JLAB / HPC group activities beyond those in Robert's list:
a) web interface to batch systems, including security issues
b) preparation for dealing with a large cluster (sysadmin issues)
c) data grid developments, including high performance WAN
    file transfers, integration with Jlab silo.  This activity overlaps
    with Jlab's participation in PPDG (Particle Physics Data Grid)





MIT PLANS (Andrew)


Andrew is developing a new tool for writing low level (site-oriented)
routines.  This is a natural evolution from his earlier work on the
optimization of QCD kernels for the Alpha in order to broaden its
scope to include the P4 and other processors.