[edit] Using MPICH on the ITS HPC Cluster

MPICH is an implementation of the MPI 1.0 standard for message passing. This implementation has become known as MPICH1 since development began on the MPICH2 implementation, which is based on the MPI 2.0 standard.

On the ITS cluster, there are two installations of MPICH: one serves the Gigabit Ethernet interconnect on all cluster nodes and the other serves the Infiniband interconnect on a subset of 18 nodes. Each implementation uses its own installation of "mpiexec", the program used to launch MPI-enabled programs.

[edit] MPICH over Gigabit Ethernet

To use MPICH with the Gigabit Ethernet interconnect, you must modify your environment so that the correct MPICH installation is used when compiling or running MPI-enabled programs. An easy way to do this is by using the following commands:

   module unload mpi
   module load mpich1

These commands should be typed to the shell if the subsequent use of MPI will be interactive. They should be placed in PBS scripts for MPI-enabled jobs that will be run via PBS. Note that the mpi module is loaded into the environment automatically and should be unloaded when the mpich1 module is loaded.

[edit] MPICH over Infiniband

To use MPICH with the Infiniband interconnect, you must ensure that your environment is set up properly so that the correct MPICH installation is used when compiling or running MPI-enabled programs. An easy way to do this is by using the following command:

   module load mpi

This command should be typed to the shell if the subsequent use of MPI will be interactive. It should be placed in PBS scripts for MPI-enabled jobs that will be run via PBS.

Note that the mpi module is currently the default MPI library and is loaded into the environment automatically for both interactive shells and PBS batch jobs, so that the explicit loading of the mpi module is currently unnecessary. Note that when using MPICH over Infiniband, it is required that the user explicitly request machines that support Infiniband. Only a subset of the machines in the cluster currently have Infiniband devices. Please see "How do I request specific nodes or processors for my job?" in the Running jobs on the HPC cluster section of the FAQ.

This article is a stub. You can help by adding to it.

ITS High Performance Cluster Articles
This article is part of the ITS Cluster series of articles
FAQ | Intel compilers | GNU compilers | Portland Group compilers
Intel Math Kernel Library | MINPACK | ScaLAPACK | GSL | FFTW3 | MPICH | NAG
R | Mathematica | NAMD | GROMACS | Amber | MATLAB
FLUENT | GAMESS | Gaussian | MOLCAS | LAMMPS | APBS