#!/bin/bash -x
#SBATCH -J Test_case_B-gpu
#SBATCH --account=prpb85
#SBATCH --nodes=24
#SBATCH --ntasks-per-node=4
#SBATCH --cpus-per-task=12
#SBATCH --time=00:29:59
#SBATCH --partition=booster
#SBATCH --output=specfem_%x_juwels-booster-%j.output
#SBATCH --gres=gpu:4
##SBATCH --acctg-freq=task=1
set -e

source ../env/env_juwels-booster
grep "^[^#;]" ../env/env_juwels-booster
cat job_juwels-booster_test_case_B.slurm
echo "=========="
echo "config.log"
echo "=========="
cat $install_dir/TestCaseB/specfem3d_globe/config.log
echo "========"
echo "make.log"
echo "========"
cat $install_dir/TestCaseB/specfem3d_globe/make.log
cd $install_dir/TestCaseB/specfem3d_globe
export I_MPI_PMI_VALUE_LENGTH_MAX=1800

#Make sure that OMP_NUM_THREADS / KMP_HW_SUBSET = cpus-per-task
export KMP_HW_SUBSET=1T
export OMP_NUM_THREADS=${SLURM_CPUS_PER_TASK}
export KMP_AFFINITY=granularity=core,compact
export FORT_BUFFERED=true

ulimit -s unlimited

MESHER_EXE=./bin/xmeshfem3D
SOLVER_EXE=./bin/xspecfem3D

# backup files used for this simulation
cp DATA/Par_file OUTPUT_FILES/
cp DATA/STATIONS OUTPUT_FILES/
cp DATA/CMTSOLUTION OUTPUT_FILES/

##
## mesh generation
##
sleep 2

echo
echo `date`
echo "starting MPI mesher"
echo

MPI_PROCESS=` echo "$SLURM_NNODES*$SLURM_NTASKS_PER_NODE" | bc -l` 
echo "SLURM_NTASKS_PER_NODE = " $SLURM_NTASKS_PER_NODE
echo "SLURM_CPUS_PER_TASKS = " $SLURM_CPUS_PER_TASK
echo "SLURM_NNODES=" $SLURM_NNODES
echo "MPI_PROCESS $MPI_PROCESS"
echo "OMP_NUM_THREADS=$OMP_NUM_THREADS"

time srun -n ${MPI_PROCESS} ${MESHER_EXE}
echo "  mesher done: `date`"
echo

##
## forward simulation
##
sleep 2

echo
echo `date`
echo starting run in current directory $PWD
echo
time srun -n ${MPI_PROCESS}  ${SOLVER_EXE}
echo "====================="
echo `date`
ls -lrth $install_dir/TestCaseB/specfem3d_globe/OUTPUT_FILES/output_*.txt
cat $install_dir/TestCaseB/specfem3d_globe/OUTPUT_FILES/output_mesher.txt
cat $install_dir/TestCaseB/specfem3d_globe/OUTPUT_FILES/output_solver.txt