Computational Science Community Wiki

Revision 14 as of 2012-09-26 17:02:40

Clear message

Training Background Materials


Information useful for training courses on research computing provided by IT Services

CSF and Course Materials

For CSF-based courses we have provided a Quick Start Guide (pdf). You may also follow the links from the current list of courses for the particular course you want.

Getting the exercises and solutions

CSF in-depth

Most courses are run on the Computation Shared Facility (CSF). Course attendees have access to the CSF (and help using it) until the end of the course session. At which point you may wish to move your files from the CSF to your local machine eg

The below links give further information to back up the Quick Start Guide

  1. Accessing CSF:
    1. CSF User Info

    2. Introduction to UNIX course includes a slide on ssh

    3. Introduction to Research Computing course includes a fews slides & an example on ssh

  2. Using Linux
    1. UNIX/Linux links

    2. Introduction to UNIX course

  3. Using modules:
    1. CSF Modules

    2. Course module on Modules, covered in the Introduction to UNIX course

  4. Accessing Compilers on CSF
    1. See CSF Software, in particular

      1. Intel compiler - please use this for all courses

      2. GNU compiler

      3. PGI compiler

    2. Programming Techniques for Research Computing course includes a fews slides & an example on using the Intel compiler and comparing its optimisation values for -O

  5. Compiling and Running Parallel Codes on CSF
    1. All parallel codes should be run in batch (see below). Also see the CSF MPI Implementation (openMPI) and CSF OpenMP pages.

    2. MPI (1 or more nodes). Nodes may be connected by gigEthernet (1Gb/sec) or InfiniBand (4xQDR on CSF which equates to 32Gb/sec). The MPI module required depends on the compiler you wish to use (eg Intel or GCC) and the interconnect between nodes. For the courses, we have 2 non-InfiniBand connected nodes and ask people to use the Intel compilers (v12.0 is current) and the only MPI implementation is v1.6 of OpenMPI, therefore the relevant MPI module to load is mpi/intel-12.0/openmpi/1.6. This gives access to the compilers (mpif90/mpicc) and the MPI launcher mpirun. The number of MPI processes is controlled by the -np flag to mpirun.

    3. OpenMP (max of 1 node). The Intel compiler supports OpenMP by use of the -openmp flag. No additional commands are required to run a threaded executable. The number of threads is controlled, essentially, by the OMP_NUM_THREADS environment variable.

  6. Makefiles (to follow)
    1. Programming Techniques for Research Computing course includes a fews slides & an example

  7. Submitting Jobs on CSF:
    1. CSF Batch Information

    2. Introduction to Research Computing course includes a fews slides & an example on using SGE to submit jobs

  8. Editors available on Linux Desktop and CSF:
    1. emacs

    2. gedit

    3. vi

    4. IDEs available on Linux Desktop and CSF: (to be confirmed)

Back to the Community Learning page