Notes for NCI (Gadi)

November 2019

Quickstart

# Instructions for Gadi

#
# Load module for netcdf
#
module load intel-compiler/2019.3.199
module load netcdf/4.7.1

#
# Clone the git repository
#
git clone https://github.com/chiaraholgate/QIBT_shared.git
cd QIBT_shared

#
# Compile with OpenMP
#
# gfortran equivalent:
# gfortran -ffree-line-length-512 -fopenmp -O3 -c QIBT_exp10.f90
ifort -03 -fopenmp -c QIBT_exp10.f90

#
# Link the executable
#
# gfortran equivalent
# gfortran -fopenmp -O3 -L/apps/netcdf/4.7.1/lib -lnetcdff QIBT_exp10.o -o main
ifort -O3 -fopenmp -L/apps/netcdf/4.7.1/lib -lnetcdff QIBT_exp10.o -o main

#
# Run QIBT from 11 Jan 1980 to 13 Jan 1980, storing the results in subdirectory ./outputs
#
chmod u+x main
mkdir outputs

#
# PBS jobscript for Gadi
#
cat - <<EOF >191119-qibt.sh
#!/bin/bash
#PBS -q normal
#PBS -l walltime=10:00:00
#PBS -l ncpus=48
#PBS -l mem=64GB
#PBS -l storage=gdata/hh5 
cd $PBS_O_WORKDIR
ulimit -s unlimited
./main 11 01 1980 13 01 1980 ./outputs

EOF

#
# Submit the job to Gadi normal queue
#
qsub 191119-qibt.sh

#
# Example resource usage via qstat
#
Every 60.0s: qstat -f 149802                                     gadi-login-04.gadi.nci.org.au: Wed Nov 20 10:19:04 2019

Job Id: 149802.gadi-pbs
    Job_Name = gadicascade.sh
    Job_Owner = [email protected]
    resources_used.cpupercent = 4252
    resources_used.cput = 23:11:41
    resources_used.jobfs = 0b
    resources_used.mem = 15460196kb
    resources_used.ncpus = 48
    resources_used.vmem = 15460196kb
    resources_used.walltime = 00:32:29
    job_state = R
    queue = normal-exec

Data sources

$ ls -lh /g/data/hh5/tmp/w28/jpe561/back_traj/
total 4.7M
drwxrwsr-x+ 2 jpe561 hh5 4.0K Oct  5  2018 watershed
drwxrwsr-x+ 2 jpe561 hh5 3.6M Mar 15  2019 wrfhrly
drwxrwsr-x+ 2 jpe561 hh5 1.1M Dec 21  2018 wrfout

** Required: a brief introduction to what is the format and contents of these files

Verify that scaling with OpenMP gives consistent output

Visualise with Panoply

First, download and install Java

Then download and extract Panoply