We have upgraded our intel compiler to the Intel Cluster Studio suite which, along with the intel compilers, MKL libraries and other related tools, also comes with the Intel MPI library. The latter should allow users to enhance the parallel performance of their applications that use a message passing scheme.
The intel cluster studio will be incorporated in the Hoffman2 users default environment on Oct 1st 2013. Accordingly, future releases of the openmpi parallel library, currently in the Hoffman2 users default environment, will become unsupported.
We encourage our users to test and get accustom the new intel compiler (version 13.1) and Intel MPI library by loading the environmental module intel/13.cs with the command:
module load intel/13.cs
Please note that Intel MPI supports both gnu and Intel compilers, use:
mpicc/mpicxx/mpif90 for gnu compiler version of Intel MPI
mpiicc/mpiicpc/mpiifort for Intel version of Intel MPI
Parallel applications compiled and linked against Intel MPI compiler and library can be submitted to the queues using the queue script:
intelmpi.q
or, in case the users would like to create their own submission script, including, in the latter, the lines:
# for sh shell scripts:
. /u/local/Modules/default/init/modules.sh
module load intel/13.cs
# for csh shell scripts:
source /u/local/Modules/default/init/modules.csh
module load intel/13.cs
before invoking mpirun.
For Hoffman2 documentation on the intel cluster studio please see: