job.sh

#!/bin/bash
## Common code executed by the Slurm file for each configuration

# Stop the script if any part fails
set -e

# Uncomment these two lines if we ask you for debugging information
# echo "DEBUG: All Slurm environment variables for this job"
# printenv | grep SLURM

# Debug info for where the job is running
# We use srun to run this on all nodes allocated to us
srun bash -c 'echo "Job is running on $(hostname), started at $(date)"'

# Set debug level to 0 if we didn't pass any environment variables 
DEBUG=${DEBUG:-0}
# Set the "use sequential" condition to 0 if we didn't pass any env vars
SEQ=${SEQ:-0}

# Compile the code at the appropriate debug level
echo "Running make to compile your code..."
make

# Choose the right executable to run
if [ "$SEQ" -eq 1 ]; then
    # Use the sequential executable
    EXECUTABLE="distr-sched-seq"
else
    # Choose the right executable based on the DEBUG level
    case "$DEBUG" in
        0)
            EXECUTABLE="distr-sched"
            ;;
        1)
            EXECUTABLE="distr-sched-debug1"
            ;;
        2)
            EXECUTABLE="distr-sched-debug2"
            ;;
        *)
            echo "Invalid value for DEBUG: $DEBUG"
            exit 1
            ;;
    esac
fi

# Creating temp job file
TMP_JOB_FILE="/tmp/$SLURM_JOB_ID.txt"
touch $TMP_JOB_FILE
chmod 700 $TMP_JOB_FILE
echo "Created temporary file on $(hostname) for job output evaluation after this run at $TMP_JOB_FILE"

# Run your executable and pass all arguments to it
echo "Running..."
set -x
perf stat -- mpirun --map-by core --bind-to core ./$EXECUTABLE $@ | tee $TMP_JOB_FILE
set +x

# Evaluate the results, default is to evaluate
EVALUATE=${EVALUATE:-1}
if [ "$EVALUATE" -eq 1 ]; then
    echo "Evaluating your results...";
    ./check.py $@ < $TMP_JOB_FILE;
fi

# Remove the tmp file if it exists
rm -f $TMP_JOB_FILE