-
Getting Started with SLAC's MPI "Pinto" Cluster
The Pinto Cluster
SLAC's "pinto" cluster consists of 32 SunFire x2270 systems, each with two
quad-core Xeon processors running at 2.93 GHz. Thus there is a total
of 256 cpus
available in the cluster. The systems have an InfiniBand
low-latency interconnect. Two of the machines, available via the generic
pool name 'pinto', are open to interactive login to allow you to build and
test your code.
MPI Environment
The pinto machines are installed with RHEL5, including the OpenMPI
packages supplied by RedHat. The current version is OpenMPI 1.4.
When you log into a pinto, your environment should be set up to
use this version (unless you have used RedHat's mpi-selector script,
or your login scripts, to override the default). You can check to
see if your PATH is correct by issuing the command
which mpirun . Currently, this should return:
/usr/lib64/openmpi/1.4-gcc/bin/mpirun
Future updates to the MPI version may change the exact details of this path.
In addition, your LD_LIBRARY_PATH should include:
/usr/lib64/openmpi/1.4-gcc/lib
(or something similar).
The following assumes you are logged onto a pinto and your
environment is set up as described.
Compiling Code
MPI versions of the common gnu compilers are available, such as
mpicc, mpic++, mpif90. For example if you have the hello.c
program from the MPI Tutorial, you can say
mpicc hello.c -o hello
to compile and link the code into an executable named
hello . You can execute the program interactively by
just saying
./hello
Submitting jobs
Batch jobs can be submitted to the pinto cluster by
specifying -a mympi on your bsub command. This will
send the job to the mpi-ibq LSF queue by default. You can also
specify the queue explicitly: -q mpi-ibq , but
that should not be necessary. You specify how many CPUs your job needs
with the -n opt. So if you have the hello
program (from above) in your home directory, and want to run on 96 CPUs,
you would say:
bsub -a mympi -n 96 ~/hello
Further Information
Check the links at the top of this page for more information on
using MPI and batch at SLAC.
For more help, send email to unix-admin@slac.stanford.edu .
John Bartelt.
Last Modified: 2010 August 09.
|
