Asap on Niflheim

Enabling the pre-installed Asap

XXXX Will be updated primo February 2020. XXXX

For more information about modules on Niflheim, read the description on the Niflheim wiki.

Check the installation

Please check that you can load Asap, and that you get the version you expect:

[thul] ~$python
Python 2.4.3 (#1, Jul 27 2009, 17:56:30)
[GCC 4.1.2 20080704 (Red Hat 4.1.2-44)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import asap3
>>> asap3.print_version()
ASAP version 3.2.2 serial, compiled 2009-08-31 15:10 on thul.fysik.dtu.dk using 'icpc -O3 -g -xS -fPIC -vec_report0 -DSTACKTRACE=linux'
>>>
>>>
>>> asap3.print_version(1)
ASAP version 3.2.2 serial, compiled 2009-08-31 15:10 on thul.fysik.dtu.dk using 'icpc -O3 -g -xS -fPIC -vec_report0 -DSTACKTRACE=linux'
  Python module: /opt/campos-asap3/3.2.2/1.el5.fys.ifort.11.0.python2.4.openmpi.1.3.3/lib64/python2.4/site-packages/asap3/__init__.pyc
  C++ module:    /opt/campos-asap3/3.2.2/1.el5.fys.ifort.11.0.python2.4.openmpi.1.3.3/lib64/python2.4/site-packages/asapserial3.so
  ase module:    /opt/campos-ase3/3.1.0.846/1.el5.fys.python2.4/lib64/python2.4/site-packages/ase/__init__.pyc
>>>

The second form also shows where Asap is installed.

Since the developers use their own installation, there is a risk that the default installation is outdated. If you discover that the installed version is ancient, complain to Jakob Schiøtz.

Developer installation on Niflheim

If you need the newest (unreleased) version of Asap, or if you contribute to Asap, you need a developer installation. It is recommended to do this in a Virtual Environment (venv), together with a developer installation of ASE and possibly of GPAW.

You can use both the Intel and the FOSS toolchains, but the Intel toolchain is strongly recommended as it gives around a factor two in performance boost for common Asap operations.

Developer installation with GPAW

  1. Install GPAW in a venv (if you have not already done so): See the corresponding GPAW installation guide. Remember to choose the Intel toolchain, if desired.
  2. Edit the file bin/activate (inside the virtual environment folder), and add these lines to the bottom of the file:

    export FYS_PLATFORM=Nifl7_${CPU_ARCH}_${TCHAIN}
    export PYTHONPATH=$VIRTUAL_ENV/asap/Python:$VIRTUAL_ENV/asap/$FYS_PLATFORM:$PYTHONPATH
    export PATH=$VIRTUAL_ENV/asap/${FYS_PLATFORM}:$PATH
  3. Activate the virtual environment (assuming that you have already changed into the directory of the virtual environment):

    source bin/activate
  4. Go to point 6 in the next section.

Developer installation without GPAW

If you do not need GPAW, then there is no need to install it.

  1. Select the right toolchain, by importing the right SciPy-bundle module. That will import Python and the toolchain:

    module purge
    module load SciPy-bundle/2020.11-intel-2020b

    Use SciPy-bundle/2020.11-foss-2020b if you want the FOSS toolchain.

  2. Create the virtual environment, giving it a sensible name (here: asap-intel):

    python -m venv --system-site-packages asap-intel
  3. Edit the file bin/activate (in the asap-intel folder). Add these lines to the top of the file:

    TCHAIN=intel
    module purge
    unset PYTHONPATH
    module load matplotlib/3.3.3-${TCHAIN}-2020b
    module load openkim-models/20210128-GCCcore-10.2.0

    and the following lines to the bottom of the file:

    export FYS_PLATFORM=Nifl7_${CPU_ARCH}_${TCHAIN}
    export PYTHONPATH=$VIRTUAL_ENV/asap/Python:$VIRTUAL_ENV/asap/$FYS_PLATFORM:$PYTHONPATH
    export PATH=$VIRTUAL_ENV/asap/${FYS_PLATFORM}:$PATH

    If you use the foss toolchain, the first line should read TCHAIN=foss instead.

  4. Activate the virtual environment:

    cd asap-intel
    source bin/activate
  5. Install ASE and prerequisites:

    pip install --upgrade pip
    git clone git@gitlab.com:ase/ase.git
    pip install -e ase

    Note that the git command assumes that you already have an account on gitlab.com, and have set up your user for command line access with an SSH key (if not, see https://gitlab.com/help/ssh/README). If you do not desire to use a gitlab.com account, you can make an anonymous checkout with the command git clone https://gitlab.com/ase/ase.git .

  6. Clone Asap, and check that the right makefile is used:

    git clone git@gitlab.com:asap/asap.git
    cd asap
    make version

    You should see output like this. It is important that the FYS_PLATFORM detected line is present, it indicates that the Niflheim architecture is being detected.:

    (asap-intel) 15:50 [slid] asap$ make version
    Getting configuration from makefile-Nifl7_broadwell_intel
    FYS_PLATFORM detected - building for CAMd/CINF/Niflheim with OBJDIR=Nifl7_broadwell_intel.
    ASAP version 3.12.1
  7. Compile asap for all the Niflheim architectures (currently four, so this takes some time):

    ./compile-niflheim.sh
  8. Test that it works:

    cd Test
    python TestAll.py
    ./TestParallel.sh

Submitting ASAP jobs to Niflheim

A special program, asap-sbatch, is used to submit ASAP jobs to Niflheim. You specify Slurm options on the command line or in comments beginning with #SBATCH, just as when submitting with the ordinary sbatch command. The special asap-sbatch command detects if you are submitting serial or parallel jobs, and submits a helper script starting your simulation in the right way.

Example:

(asap-intel) 08:49 [slid] Test$ asap-sbatch TestAll.py
Virtual environment detected: /home/niflheim/schiotz/development/asap-intel

Submitted batch job 3397275

The TestAll.py begins like this:

#SBATCH --partition=xeon8
#SBATCH -N 1
#SBATCH -n 4
#SBATCH --time=1:00:00
#SBATCH --mem=2G
#SBATCH --output=Timing-%J.out

The first comments specify the Slurm queue, the next two that the job should run on a single node and uses a single cpu core. Then the maximal wall time is specified, and the maximal memory usage per cpu core. The output (both stdout and stderr) goes to a file named Timing-NNNNN.out, where NNNNN is the job number. See the Slurm section on the Niflheim wiki for more information. Options specified on the command line will overrule options in the Python script.

Serial simulations

Serial simulations are submitted as described above. Please submit serial scripts to the xeon-8 partition, as the machines with more cores should preferably be used for parallel jobs.

Parallel simulations

Parallel simulations are submitted in exactly the same way. Your script will be started on all the requested CPUs. Just be sure to request the right number of nodes and processors using the -N nodes and -n processors sbatch options, either on the asap-sbatch command line or as a #SLURM comment in the Python script.

Aim to use a number of processors (cpu cores) that is divisible by the number of cpu cores in the partition of Niflheim you are using. For example, on the xeon-40 partitions, the number of cores should be a multiple of 40.

Multithreaded simulations

Multithreaded simulations are currently unsupported by ``asap-sbatch``

Asap: Niflheim (last edited 2021-10-07 10:04:07 by JakobSchiøtz)