----------------------------------------------------------------------

H   H  PPP    CCC                           UK  NATIONAL  HPC  SERVICE
H   H  P  P  C   C                          --------------------------
HHHHH  PPP   C      x  x                    provided by
H   H  P     C   C   xx                     EPCC and
H   H  P      CCC   x  x                    CCLRC Daresbury Laboratory

----------------------------------------------------------------------
HPCx User Mailing 015                                      14 May 2003
----------------------------------------------------------------------

Contents

** Interactive Parallel Jobs

** Memory Limits for Parallel Batch Jobs

----------------------------------------------------------------------
Greetings--


INTERACTIVE PARALLEL JOBS

Interactive execution is now available for parallel jobs. For example,
it is now possible to run an MPI job directly from the command line.
Details on how to use this facility are available in the HPCx User's
Guide:

http://www.hpcx.ac.uk/support/documentation/UserGuide/HPCxuser/

in the section 'Interactive Execution'.

In order to use this facility, you will need to create a LoadLeveler
script. Example scripts for running interactive parallel jobs are
now included in the standard templates at:

http://www.hpcx.ac.uk/support/FAQ/template.tar


----------------------------------------------------------------------

MEMORY LIMITS FOR PARALLEL BATCH JOBS

Some users had been experiencing problems with the memory limits for
parallel batch jobs on HPCx. At the last maintenance session we made
changes to allow users to use the 7.2GB of available memory on an LPAR
more efficiently. In particular, this now enables OpenMP and mixed MPI
/ OpenMP jobs to utilise the full memory on each LPAR.

For a parallel batch job (excluding interactive jobs), the amount of
memory allocated per node is now automatically determined from the
requested tasks-per-node in your LoadLeveler script. The amount of
memory available per process is simply:

    total available memory (7.2GB) / number of tasks_per_node

So, for example, a job that requests 4 tasks per node will be allowed
up to  ~ 1.8GB (7.2 / 4) per process.


----------------------------------------------------------------------

Regards

-- David

----------------------------------------------------------------------
Earlier mailings: http://www.hpcx.ac.uk/support/notices/index.html

--
David Henty                               d.henty@epcc.ed.ac.uk
HPCx User Administration and Helpdesk
HPCx:  http://www.hpcx.ac.uk              Helpdesk: support@hpcx.ac.uk
Phone: +44 131 650 5029                   Fax: +44 131 650 6555