Qlogic Fast Fabric User Manual page 146

6 – MPI Sample Applications
Pallas
Prior to running this application, a HPL.dat file must be installed into
/opt/iba/src/mpi_apps/hpl/bin/ICS/HPL.dat on all nodes. The
config_hpl script and some sample configurations are included.
The config_hpl script can select from one of the assorted HPL.dat files in
hpl-config. For assorted cluster sizes (by number of CPUs). Assorted sample
HPL.dat files are provided in /opt/iba/src/mpi_apps/hpl-config. These
files are a good starting point for most clusters and should get within 10-20% of the
optimal performance for the cluster. The problem sizes used assume a cluster with
1GB of physical memory per processor (e.g., for a 2 processor node, 2GB of node
memory is assumed). For each cluster size, 4 files are provided:
t - a very small test run (5000 problem size)
s - a small problem size on the low end of optimal problem sizes
m - a medium problem size
l - a large problem size
These can be selected using config_hpl. The following command displays the
preconfigured problem sizes available:
./config_hpl
For example, to quickly confirm that HPL will run on the 16 nodes in the
/opt/iba/src/mpi_apps/mpi_hosts file, enter ./config_hpl 16t. This
will edit the HPL.dat file on the local host for a 16 host "very small" test, and copy
that HPL.dat file to all hosts in the mpi_hosts file.
Once the HPL.dat has been configured and copied, HPL can be run using the
script:
1. cd /opt/iba/src/mpi_apps
2. ./run_hpl NP
where NP is the number of processors for the run. For example:
./run_hpl 16
For more information about HPL, consult the README, TUNING and assorted
HTML files in /opt/iba/src/mpi_apps/hpl.
During this benchmark the /opt/iba/src/mpi_apps/mpi.param.hpl
config file is used.
6.7
Pallas
The Pallas benchmark does exhaustive benchmarking of latency and bandwidth
for assorted message sizes for many MPI primitives. This benchmark is a good
tool to evaluate and tune small clusters or a subset of a large cluster.
6-4
Q
D000006-000 Rev A