#!/bin/bash
#
#PBS -l nodes=1:ppn=4:sl,walltime=20:00:00
#PBS -o /home/woody/capn/mppi033h/logs/orcasong/submit_data_to_images_${PBS_JOBID}_${PBS_ARRAYID}.out -e /home/woody/capn/mppi033h/logs/orcasong/submit_data_to_images_${PBS_JOBID}_${PBS_ARRAYID}.err
# first non-empty non-comment line ends PBS options

# Submit with 'qsub -t 1-x submit_data_to_images.sh'
# This script uses the data_to_images.py file in order to convert all .h5 raw MC files to .h5 event "images" (CNN input).
# Currently available ORCA 115l sim files:
# neutrinos: 600 files each for 1-5 GeV prod (muon-CC, elec-CC/NC) number of jobs needed = 5 with files_per_job=120,
#            else (3-100 GeV prod)
#            muon-CC = 2400 files, number of jobs needed = 20 with files_per_job=120
#            elec-CC = 1200 files, number of jobs needed = 10 with files_per_job=120
#            elec-NC = 1200 files, number of jobs needed = 10 with files_per_job=120
#            tau-CC = 1800 files (half the n_evts of other interaction channels), number of jobs needed = 15 with files_per_job=120 and half walltime
# mupage: 20000 files, with files_per_job=250, 80 jobs needed with 20h walltime.
# random_noise: 500 files, with files_per_job=20 , 25 jobs needed with 20h walltime.


#--- USER INPUT ---#

# load env, only working for conda env as of now
python_env_folder=/home/hpc/capn/mppi033h/.virtualenv/python_3_env/
job_logs_folder=/home/woody/capn/mppi033h/logs/orcasong/cout

detx_filepath=/home/woody/capn/mppi033h/Code/OrcaSong/user/detx_files/orca_115strings_av23min20mhorizontal_18OMs_alt9mvertical_v1.detx
config_file=/home/woody/capn/mppi033h/Code/OrcaSong/user/config/orca_115l_mupage_rn_neutr_classifier/conf_ORCA_115l_random_noise_xyz-c.toml

particle_type=random_noise
mc_prod=random_noise

# total number of files per job
# For neutrinos 3-100GeV:
# muon-CC/elec-CC/elec-NC/tau-CC n=120 with PBS -l nodes=1:ppn=4:sl,walltime=5:00:00
# For neutrinos 1-5GeV:
# muon-CC/elec-CC/elec-NC n=120 with PBS -l nodes=1:ppn=4:sl,walltime=5:00:00
# For mupage: n=250 with PBS -l nodes=1:ppn=4:sl,walltime=5:00:00
# For random_noise: n=20 with PBS -l nodes=1:ppn=4:sl,walltime=5:00:00
files_per_job=20 # must be dividible by 4!

#--- USER INPUT ---#

# setup

n=${PBS_ARRAYID}
source activate ${python_env_folder}

declare -A filename_arr
declare -A folder_ip_files_arr

if [ ${mc_prod} == "neutr_3-100GeV" ]
then
filename_arr=( ["muon-CC"]="JTE.KM3Sim.gseagen.muon-CC.3-100GeV-9.1E7-1bin-3.0gspec.ORCA115_9m_2016"
               ["elec-CC"]="JTE.KM3Sim.gseagen.elec-CC.3-100GeV-1.1E6-1bin-3.0gspec.ORCA115_9m_2016"
               ["elec-NC"]="JTE.KM3Sim.gseagen.elec-NC.3-100GeV-3.4E6-1bin-3.0gspec.ORCA115_9m_2016"
               ["tau-CC"]="JTE.KM3Sim.gseagen.tau-CC.3.4-100GeV-2.0E8-1bin-3.0gspec.ORCA115_9m_2016")
elif [ ${mc_prod} == "neutr_1-5GeV" ]
then
filename_arr=( ["muon-CC"]="JTE.KM3Sim.gseagen.muon-CC.1-5GeV-9.2E5-1bin-1.0gspec.ORCA115_9m_2016"
               ["elec-CC"]="JTE.KM3Sim.gseagen.elec-CC.1-5GeV-2.7E5-1bin-1.0gspec.ORCA115_9m_2016"
               ["elec-NC"]="JTE.KM3Sim.gseagen.elec-NC.1-5GeV-2.2E6-1bin-1.0gspec.ORCA115_9m_2016")
else
filename_arr=( ["mupage"]="JTE.ph.ph.mupage.ph.ph.ph.ORCA115_9m_2016"
               ["random_noise"]="JTE.ph.ph.random_noise.ph.ph.ph.ORCA115_9m_2016")
fi

folder_ip_files_arr=( ["neutr_3-100GeV"]="/home/saturn/capn/mppi033h/Data/raw_data/ORCA_JTE_NEMOWATER/calibrated/with_jte_times/3-100GeV/${particle_type}"
                      ["neutr_1-5GeV"]="/home/saturn/capn/mppi033h/Data/raw_data/ORCA_JTE_NEMOWATER/calibrated/with_jte_times/1-5GeV/${particle_type}"
                      ["mupage"]="/home/saturn/capn/mppi033h/Data/raw_data/mupage"
                      ["random_noise"]="/home/saturn/capn/mppi033h/Data/raw_data/random_noise")

filename="${filename_arr[${particle_type}]}"
folder="${folder_ip_files_arr[${mc_prod}]}"

# run

no_of_loops=$((${files_per_job}/4)) # divide by 4 cores -> e.g, 15 4-core loops needed for files_per_job=60
file_no_start=$((1+((${n}-1) * ${files_per_job}))) # filenumber of the first file that is being processed by this script (depends on JobArray variable 'n')

# currently only working for 4 cores

for (( k=1; k<=${no_of_loops}; k++ ))
do
    file_no_loop_start=$((${file_no_start}+(k-1)*4))
    thread1=${file_no_loop_start}
    thread2=$((${file_no_loop_start} + 1))
    thread3=$((${file_no_loop_start} + 2))
    thread4=$((${file_no_loop_start} + 3))

    (time taskset -c 0  make_nn_images -c ${config_file} ${folder}/${filename}.${thread1}.h5 ${detx_filepath} > ${job_logs_folder}/${filename}.${thread1}.txt) &
    (time taskset -c 1  make_nn_images -c ${config_file} ${folder}/${filename}.${thread2}.h5 ${detx_filepath} > ${job_logs_folder}/${filename}.${thread2}.txt) &
    (time taskset -c 2  make_nn_images -c ${config_file} ${folder}/${filename}.${thread3}.h5 ${detx_filepath} > ${job_logs_folder}/${filename}.${thread3}.txt) &
    (time taskset -c 3  make_nn_images -c ${config_file} ${folder}/${filename}.${thread4}.h5 ${detx_filepath} > ${job_logs_folder}/${filename}.${thread4}.txt) &
    wait
done