Ctrl+K

GNU Parallel#

  • Tool for executing jobs in parallel on one or more computer

  • Jobs can be single commands, scripts or parallelized jobs themselves

  • Helpful for running repetitive tasks or utilizing computing resources better

Running Parallel on HPCs#

  • Most clusters should have one installed, check with module av or module spider parallel, if available then module load

  • You can also install your own. Here is an example, installed into ~/Programs

    1. Vist the GNU website https://www.gnu.org/software/parallel/

    2. Find the download links HTTPS, HTTP, FTP

    3. Right-click and copy the link

    4. In Terminal, login to the cluster and navigate to cd ~/.Programs

    5. Copy the link here with wget WWW....

    6. Decompress the folder with tar xf parallel-202...

    7. Go to the directory and configure cd  parallel-202... and configure ./configure --prefix=~/Programs && make && make install

Parallel Umbrella Sampling Windows for MD Simulations#

#!/bin/bash
#SBATCH -p long
#SBATCH -t 5-00:00:00
#SBATCH -n 1
#SBATCH --ntasks-per-node=32
#SBATCH --output=%j.out
#SBATCH --error=%j.err
#SBATCH --job-name=${name}${job}

date
module load parallel 

w_i=0   # First window 
w_f=42  # last window
jobs=8  # number of cores for each simulation

parallel --jobs ${JOBS} bash runmd.sh {} ::: $(seq -f"%02g" w_i w_f)

date
Show Source