Adding inside

examples/parallel/KTH_PDC/Dardel_runSnudda_inside.job

@@ -0,0 +1,79 @@
+#!/bin/bash -l
+#SBATCH --partition=main
+#SBATCH -o log/runSnudda-%j-output.txt
+#SBATCH -e log/runSnudda-%j-error.txt
+#SBATCH -t 00:30:00
+#SBATCH -J Snudda
+#SBATCH -A naiss2023-5-231
+#SBATCH --nodes=2
+#SBATCH -n 256
+#SBATCH --cpus-per-task=2
+#SBATCH --mem-per-cpu=930M
+#SBATCH --mail-type=ALL                                                         
+module load snic-env
+
+
+#..
+#export OMP_STACKSIZE=128G
+ulimit -s unlimited
+
+
+#let NWORKERS="$SLURM_NTASKS-2"
+let NWORKERS="100"
+
+export IPNWORKERS=$NWORKERS
+
+
+export IPYTHONDIR="/cfs/klemming/scratch/${USER:0:1}/$USER/.ipython"
+rm -r $IPYTHONDIR
+export IPYTHON_PROFILE=default
+source $HOME/Snudda/snudda_env/bin/activate
+
+
+#.. Start the ipcontroller
+export FI_CXI_DEFAULT_VNI=$(od -vAn -N4 -tu < /dev/urandom)
+srun -n 1 -N 1 -c 2 --exact --overlap --mem=0 ./ipcontroller_new.sh &
+
+
+echo ">>> waiting 60s for controller to start"
+sleep 60 
+
+#.. Read in CONTROLLERIP
+CONTROLLERIP=$(<controller_ip.txt)
+
+
+##.. Start the engines
+echo ">>> starting ${IPNWORKERS} engines "
+#srun -n ${IPNWORKERS} -c 2 --exact --overlap ipengine --location=${CONTROLLERIP} --profile=${IPYTHON_PROFILE} --mpi \
+#--ipython-dir=${IPYTHONDIR}  --timeout=30.0 --log-level=DEBUG \
+#--BaseParallelApplication.verbose_crash=True --IPEngine.verbose_crash=True \
+#--Kernel.stop_on_error_timeout=1.0 --IPythonKernel.stop_on_error_timeout=1.0 \
+#Session.buffer_threshold=4096 Session.copy_threshold=250000 \
+#Session.digest_history_size=250000 c.EngineFactory.max_heartbeat_misses=10  c.MPI.use='mpi4py' \
+#1> ipe_${SLURM_JOBID}.out 2> ipe_${SLURM_JOBID}.err &
+
+#srun -n ${IPNWORKERS} -c 2 --exact --overlap valgrind --leak-check=full --show-leak-kinds=all \
+#ipengine --location=${CONTROLLERIP} --profile=${IPYTHON_PROFILE} --mpi \
+#--ipython-dir=${IPYTHONDIR}  --timeout=30.0 c.EngineFactory.max_heartbeat_misses=10  c.MPI.use='mpi4py' \
+#1> ipe_${SLURM_JOBID}.out 2> ipe_${SLURM_JOBID}.err &
+
+export FI_CXI_DEFAULT_VNI=$(od -vAn -N4 -tu < /dev/urandom)
+srun -n ${IPNWORKERS} -c 2 -N ${SLURM_JOB_NUM_NODES} --exact --overlap --mem=0 ipengine \
+--location=${CONTROLLERIP} --profile=${IPYTHON_PROFILE} --mpi \
+--ipython-dir=${IPYTHONDIR}  --timeout=30.0 c.EngineFactory.max_heartbeat_misses=10  c.MPI.use='mpi4py' \
+1> ipe_${SLURM_JOBID}.out 2> ipe_${SLURM_JOBID}.err &
+
+
+echo ">>> waiting 60s for engines to start"
+sleep 30
+
+export FI_CXI_DEFAULT_VNI=$(od -vAn -N4 -tu < /dev/urandom)
+srun -n 1 -N 1 --exact --overlap --mem=0 ./Dardel_runSnudda_inside.sh
+
+
+echo " "
+
+echo "JOB END "`date` start_time_network_connect.txt
+
+wait
+

examples/parallel/KTH_PDC/Dardel_runSnudda_inside.sh

@@ -0,0 +1,120 @@
+#!/bin/bash
+
+
+
+SNUDDA_DIR=$HOME/Snudda/snudda
+JOBDIR=networks/test_10k
+
+SIMSIZE=10000
+
+# If the BasalGangliaData directory exists, then use that for our data
+#/cfs/klemming/scratch/${USER:0:1}/$USER/BasalGangliaData/data
+#BasalGangliaData/Parkinson/PD0
+if [[ -d "$HOME/BasalGangliaData/data" ]]; then
+    export SNUDDA_DATA="$HOME/BasalGangliaData/data"
+    echo "Setting SNUDDA_DATA to $SNUDDA_DATA"
+else
+    echo "SNUDDA_DATA environment variable not changed (may be empty): $SNUDDA_DATA"
+fi
+
+mkdir -p $JOBDIR
+
+echo "Dardel_runSnudda.sh should be started with srun -n 1, to only get one process"
+
+echo "SLURM_PROCID = $SLURM_PROCID"
+
+if [ "$SLURM_PROCID" -gt 0 ]; then
+    mock_string="Not main process"
+else
+
+    # For debug purposes:                                                         
+    echo "PATH: "$PATH
+    echo "IPYTHONDIR: "$IPYTHONDIR
+    echo "PYTHONPATH: "$PYTHONPATH
+    echo "LD_LIBRARY_PATH: "$LD_LIBRARY_PATH
+
+    echo ">>>>>> Main process starting ipcluster"
+    echo
+
+    echo "Start time: " > start_time_network_connect.txt
+    date >> start_time_network_connect.txt
+
+    echo ">>> Init: "`date`
+    snudda init ${JOBDIR} --size ${SIMSIZE} --overwrite --randomseed 1234 --stayInside
+
+    if [ $? != 0 ]; then
+	echo "Something went wrong during init, aborting!"
+	ipcluster stop	
+	exit -1
+    fi
+
+# WE NOW START IPCLUSTER USING ipcontroller.sh INSTEAD...
+#
+#    echo "SLURM_NODELIST = $SLURM_NODELIST"
+#    let NWORKERS="$SLURM_NTASKS - 1"
+#
+#    echo ">>> NWORKERS " $NWORKERS
+#    echo ">>> Starting ipcluster `date`"
+#    
+#    #.. Start the ipcluster
+#    ipcluster start -n ${NWORKERS} \
+#	      --ip='*' \
+#	      --HeartMonitor.max_heartmonitor_misses=1000 \
+#	      --HubFactory.registration_timeout=600 \
+#	      --HeartMonitor.period=10000 & 
+#
+#    
+#    #.. Sleep to allow engines to start
+#    echo ">>> Wait 120s to allow engines to start"
+#    sleep 120 #60
+
+    echo ">>> Place: "`date`
+    snudda place ${JOBDIR} --verbose
+
+    if [ $? != 0 ]; then
+	echo "Something went wrong during placement, aborting!"
+	# ipcluster stop	
+	exit -1
+    fi
+
+    echo ">>> Detect: "`date`
+    snudda detect ${JOBDIR} --hvsize 50 --parallel
+
+    if [ $? != 0 ]; then
+	echo "Something went wrong during detection, aborting!"
+	# ipcluster stop	
+	exit -1
+    fi
+
+    echo ">>> Prune: "`date`
+    snudda prune ${JOBDIR} --parallel
+
+    if [ $? != 0 ]; then
+	echo "Something went wrong during pruning, aborting!"
+	# ipcluster stop	
+	exit -1
+    fi
+
+    # Disable input generation at the moment
+
+    #echo ">>> Input: "`date`
+    # cp -a $SNUDDA_DIR/data/input_config/input-v10-scaled.json ${JOBDIR}/input.json
+    cp -a $SNUDDA_DIR/data/input_config/external-input-dSTR-scaled-v4.json ${JOBDIR}/input.json
+
+    snudda input ${JOBDIR} --parallel --time 5
+
+
+    #.. Shut down cluster
+    # ipcluster stop	
+    #.. Shutdown ipcontroller
+    echo "Shutting down ipcontroller"
+
+    python ipcontroller_shutdown.py
+
+
+    date
+    #echo "JOB END "`date` start_time_network_connect.txt
+
+    echo "EXITING Dardel_runjob.sh"
+
+fi