#!/bin/bash
#SBATCH -J Test_case_A-gpu
#SBATCH --nodes=24
#SBATCH --ntasks-per-node=4
#SBATCH --cpus-per-task=8
#SBATCH --time=00:30:00
#SBATCH --output specfem-gpu_TestCaseA-gcc-9-cuda-11-GPU-NoopenMP-NoMultithread-8cpus-distribBlock-znver2-%j.output
#SBATCH -p gpu
#SBATCH --gres=gpu:4
#SBATCH --hint=nomultithread
#SBATCH --distribution=block:block
#set -e
source ../env/env_vega-gpu
grep "^[^#;]" ../env/env_vega-gpu
cat job_vega-gpu_test_case_A.slurm 
echo "=========="
echo "config.log"
echo "=========="
cat $install_dir/TestCaseA/specfem3d_globe/config.log
echo "========"
echo "make.log"
echo "========"
cat $install_dir/TestCaseA/specfem3d_globe/make.log
cd $install_dir/TestCaseA/specfem3d_globe
grep GPU DATA/Par_file

export OMPI_MCA_pml=ucx
export OMPI_MCA_btl="^uct,tcp,openib,vader" #self,vader,openib" # with ^ucx and ^tcp -> error occurred in MPI_Bcast 
export CUDA_VISIBLE_DEVICES=0,1,2,3
#export OMP_NUM_THREADS=2

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"

time mpirun -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
#unset FORT_BUFFERED
time mpirun -n ${MPI_PROCESS}  ${SOLVER_EXE}

echo "finished successfully"
echo "====================="
echo `date`
ls -lrth $install_dir/TestCaseA/specfem3d_globe/OUTPUT_FILES/output_*.txt
cat $install_dir/TestCaseA/specfem3d_globe/OUTPUT_FILES/output_mesher.txt
cat $install_dir/TestCaseA/specfem3d_globe/OUTPUT_FILES/output_solver.txt