install_darshan.md

• Ref: https://www.mcs.anl.gov/research/projects/darshan/

steps to install

• Unpack
• ./prepare.sh
• cd darshan-runtime
• ./configure --prefix=/opt/darshan/3.4.4/rt --with-log-path=/opt/logs --with-jobid-env=PBS_JOBID CC=mpicc
• make; make install
• Edit bin/darshan-mk-log-dirs.pl

use Env;
if ($DARSHAN_LOGPATH eq "") {
  $LOGDIR="/opt/logs"
} else {
  $LOGDIR=$DARSHAN_LOGPATH;
}

• Now export DARSHAN_LOGPATH=/tmp is accecpted
• cd ../darshan-util
• ./configure --prefix=/opt/darshan/3.4.4/util
• make; make installan
  • If QT visualization fails, edit dxt_analyzer.py as matplotlib.use('Agg')

steps to use

• export DARSHAN_LOGPATH=pwd
• darshan-mk-log-dirs.pl
• LD_PRELOAD=/opt/darshan/3.4.4/rt/lib/libdarshan.so mpirun -n 8 ./a.out
• darshan-job-summary.pl AAA.darshan will generate AAA.pdf
• darsha-parser AAA.darshan to produce text results