The HPC team have been making use of “module” for a while now. A few of our users are aware of its availability but it hasn’t really been disclosed as a “production” ready setup. Well, it is now. Scientific packages installed on the cluster will be included in module.
For example, executing ” module avail ” on the head node will produce the following
———– /opt/exp_soft/modules-3.2.10/Modules/3.2.10/modulefiles ————
compilers/gcc481 modules
cuda/Cuda-5.0 mpi/openmpi-1.6.5
cuda/Cuda-6.0 mpi/openmpi-1.8.2
cuda/Cuda-6.5 mpif90
juliaLang-3.02 python/anaconda-python-2.7
module-gcc481 python/anaconda-python-3.4
module-info
Lets say I needed to make use of the Cuda environment for compiling GPU code, I could simple type ” module load cuda/Cuda-6.5″ and automatically it sets up my $PATH and library paths automatically. Submitting a job to the cluster would include the same, simply add ” module load cuda/Cuda-6.5″ after your PBS directives in the job submission script and you good to go. To unload a module, simply run ” module unload cuda/Cuda-6.5″
We will be adding more and more applications so it becomes less of a hassle to remember the paths to the applications installed in /opt/exp_soft. As always, if you have any problems, please feel free to e-mail us.