#! /bin/bash

echo "Preparing:"
set -x                          # Output commands
set -e                          # Abort on errors

cd @RUNDIR@-active

echo "Checking:"
pwd
hostname
date
cat ${PBS_NODEFILE} > SIMFACTORY/NODES

echo "Environment:"
export GMON_OUT_PREFIX=gmon.out
export OMP_NUM_THREADS=@NUM_THREADS@
export MPI_DIR=/usr/local/packages/openmpi/1.6.5/Intel-13.0.0
export MPD_NODEFILE=mpd_nodefile
export MPI_NODEFILE=mpi_nodefile
env > SIMFACTORY/ENVIRONMENT

echo "Job setup:"
echo "   Allocated:"
echo "      Nodes:                      @NODES@"
echo "      Cores per node:             @PPN@"
echo "   Running:"
echo "      MPI processes:              @NUM_PROCS@"
echo "      OpenMP threads per process: @NUM_THREADS@"
echo "      MPI processes per node:     @(1.0*@NUM_PROCS@/@NODES@)@"
echo "      OpenMP threads per core:    @(1.0*(@NUM_PROCS@*@NUM_THREADS@)/(@NODES@*@PPN@))@"
echo "      OpenMP threads per node:    @PPN_USED@"

case '@PPN_USED@:@NUM_THREADS@' in
    (16:16|8:8|4:4|2:2|1:1) options='-npernode 1 -bind-to-none';;
    (16:8|8:4|4:2|2:1)      options='-npersocket 1 -bind-to-socket';;
    (16:4|8:2|4:1)          options='-npersocket 2 -bind-to-socket';;
    (16:2|8:1)              options='-npersocket 4 -bind-to-socket';;
    (16:1|*)                options='-npersocket 8 -bind-to-core';;
esac

echo "Starting:"
uniq ${PBS_NODEFILE} > ${MPD_NODEFILE}
for node in $(cat ${MPD_NODEFILE}); do
    for ((proc=0; $proc<@(@PPN_USED@/@NUM_THREADS@)@; proc=$proc+1)); do
        echo ${node}
    done
done > ${MPI_NODEFILE}
export CACTUS_STARTTIME=$(date +%s)

time ${MPI_DIR}/bin/mpirun -n @NUM_PROCS@ -hostfile ${MPI_NODEFILE} -mca I_MPI_OFA_USE_XRC 1 -mca coll_fca_enable 0 -num-sockets 2 $options -display-map -report-bindings -x OMP_NUM_THREADS=@NUM_THREADS@ @EXECUTABLE@ -L 3 @PARFILE@

echo "Stopping:"
date

echo "Done."