Tutorial:BerkeleyGW

From OctopusWiki
Jump to navigation Jump to search

NOTE: This tutorial page is set up for the Benasque TDDFT school 2018.

  • Your instructors: David Strubbe and Adriel González.
  • Test your knowledge with a quiz!

Interacting with Cori

The BerkeleyGW tutorial is done on the Cori supercomputer (at NERSC in California). There are a few key things you need to know about how to interact with the machine:

  • To log in, run ssh trainXXX@cori.nersc.gov in your terminal, substituting the actual name of your training account for XXX.
  • Be aware that since this machine is far away, running X-Windows programs will be very slow.
  • For visualization with XCrySDen or other tools, installing and running on your laptop is recommended.
  • We submit jobs using the SLURM queue manager, using a "reservation" for Aug 22 (benasque2018_1), 23 (benasque2018_2), and 24 (benasque2018_3). The reservations are usable from 15:00 to 19:00 each day. If you want to run outside of these times, use the debug queue. The tutorials are designed to be run using an "interactive job", in which you have a node continually available for your use and then executables launched from the command line are run immediately. Start an interactive job like this (do not submit more than one):
salloc -N 1 -q regular --reservation=benasque2018_2 -t 03:00:00 -C haswell

If you are using the debug queue rather than a reservation, use this:

salloc -N 1 -q debug -t 00:30:00 -C haswell
  • Here is an example script using 32 cores. If you are using an interactive job, don't use this.
#!/bin/bash

#SBATCH -J test_pulpo
#SBATCH -N 1
#SBATCH -C haswell
#SBATCH -p debug
#SBATCH -t 00:30:00
#SBATCH --export=ALL

module load octopus/8.2
srun -n 32 octopus &> output
  • To copy files from Cori to your local machine, in a terminal on your local machine, write scp trainXXX@cori.nersc.gov:FULL_PATH_TO_YOUR_FILE . (filling in the username and filename) and enter your password when prompted. For very small ASCII files, you may find cut and paste more convenient.
  • To see if you have jobs running, do squeue -u trainXXX. You should not have more than one in the queue; if you do, cancel them with scancel JOBID, filling in the number for JOBID from the output of squeue.
  • Accounts will expire on September 11. Feel free to copy the files off the machine before that to somewhere else for your future reference.

Documentation and resources

Instructions

The first time you log in, execute these lines which will help you see color-coding for what is a link, executable, or directory:

echo 'alias ls="ls --color"' >> ~/.bashrc.ext
. ~/.bashrc

Each time you log in, you should do this:

# Load modules
module load berkeleygw/1.2
# Go to the scratch directory, where all runs should happen.
cd $SCRATCH

To begin with the examples,

# List all examples available
ls /project/projectdirs/mp149/Benasque2018
# Copy 1-boron_nitride example to your directory
cp -R /project/projectdirs/mp149/Benasque2018/1-boron_nitride .
# Go to your local folder and follow instructions
cd 1-boron_nitride
less README

Schedule

  • Day 1
    • 1-boron_nitride, GW
    • 2-benzene, GW
  • Day 2
    • 2-benzene, Bethe-Salpeter
    • 3-xct_LiCl, exciton visualization (download here or from Cori, unpack the archive with "unzip", and follow README)
    • 4-silicon, Bethe-Salpeter
  • Day 3 (optional)
    • Stretch goals from 2-benzene
    • Any of the plane-wave BerkeleyGW examples, on Cori at /project/projectdirs/mp149/BGW-2018/: silicon with PARATEC; boron nitride sheet, benzene, or sodium with Quantum ESPRESSO.

Errata

  • In 1-boron_nitride/1-mf/2.1-epsilon, the reference to ./01-calculate_wfn.qsub should be ./02-calculate_wfn.qsub, and you should run 01-get_kgrid.sh first and look at the input and output files.
  • In the output of kgrid.x, in the KPointsGrid block, the value for a small q-shift is not quite right, and should be divided by the number of k-points in that direction. The value given in the Octopus input files is the correct one.
  • In 1-boron_nitride/2-bgw/1-epsilon, the script ./01-run_epsilon.qsub should have "32" rather than "48" in the execution line, or you will get some OMP errors.
  • In 1-boron_nitride/2-bgw/2-sigma, the script ./01-run_sigma.qsub should have "32" rather than "48" in the execution line, or you will get some OMP errors.

For historical interest


Back to Tutorials