What I need

I want to monitor system resources (namely memory and cpu usage) by application - not just by process. Just as the Windows Task Manager is grouping resources by the "calling mother process", I like to see it like that as well. Nowadays, applications like firefox and vscode spawn many child processes and I want to get a quick and complete overview of their usage.

The solution can be a GUI or TUI, a bash script or a big one-liner. I do not really care. For it to work, I imagine I could feed it with the pid of the mother process or the name of an executable as a means of filtering.


Task Manager groupes/accumulates Chrome browser system resources

What I Tried

  • I tried htop, but it only shows me a tree where the calling process has its own memory listed - not the ones it called.
  • I tried the gnome-system-monitor, but its the same.
  • I tried a bit with ps and free but have not found the correct set of arguments / pipes to make them do what I want.

It stumped me that I could not google a solution for that. Maybe there is a reason for it?

Does anybody have an idea? I would very much appreciate it!

  • No solutions. You could write and submit patches for htop/top. Oct 25, 2021 at 13:06

1 Answer 1


This script below requires a lot of additional improvements, but I think it can serve as a basis.
I started to write comments, but for now, not able to finish it. I will use edits to my answer to add new comments and fix bugs when I get more free time.
In my environment it works fine. I called this script mytop and put it to /usr/local/bin so I have bash command tab completion on it. You can put mytop to your ~/bin directory (if ~/bin not in your $PATH, add it), or whatever place on your machine. Of course execute bit must be set, with chmod u+x mytop.

# mytop -ver 1.0

# script name (default is: 'mytop')
s_name=$(basename $0)

# version

# set default time between mytop iterations
# set default mytop repetitions/iterations

# Help function explaining syntax, options, ...
# Display Help
echo "Show Totals of %CPU and &MEM using 'top' command."
echo "Syntax:"
echo "   $s_name [-h|-V]"
echo "   $s_name [[-d <S>][-n <N>] <APP_NAME>"]
echo "Options:"
echo "     -h       Print this Help."
echo "     -d S     Delay/wait S seconds between iterations (default: 3 seconds)."
echo "     -n N     Run/iterate 'mytop' N times (default: 3 times)."
echo "     -V       Print version."
echo "Examples:"
echo "   mytop -V"
echo "   mytop -d1 -n5 chromium"
echo 'Use CTRL+C for exit!'

# Handling options from command line arguments
while getopts ":hn:d:V" option; do
case $option in
  h) # display Help
  V) # print version 
     echo "$s_name $ver"
  n) # set how many times 'mytop' will repeat/iterate
  d) # set delays in seconds
     echo "$s_name: inapropriate: '$1'."
     echo "Usage:"
     echo "  $s_name [-h|-V|-d<S> -n<N> <APP_NAME>]"

# If no arguments given just display Help function and exit
if [[ $# -eq 0 ]]; then
    # If last argument starts with '-' exit from app
    if [[ ${@:$#} =~ -+.* ]]; then
        echo ${s_name}: error: Last argument must be the name of the application that you want to track. >&2
        exit 1

# Set 'dashes' literally
# or set them with printf command
t_dsh=$(printf '%0.s-' {1..59})

# Not in use
#if [[ -z $mt_rep ]] 2>/dev/null; then
#   r_endless=1
#   mt_rep=1000
#   r_endless=0

while [[ $i -lt $mt_rep ]]; do

    #if [[ "$r_endless" == "0" ]]; then ((i++)); fi

    # Handle pids of app you want to track by removing 'mytop' pids
    # get s_name (mytop) pids
    pgrep $s_name > /tmp/mt_pids
    # get app_name pids -all of them --not desired behaviour
    pgrep -f $app_name > /tmp/app_name_pids
    # get app_name without mytop pids --desired behaviour
    for e in $(cat /tmp/mt_pids); do sed -i "/$e/d" /tmp/app_name_pids; done
    if [[ ! -s "/tmp/app_name_pids" ]]; then echo "1000000" > /tmp/app_name_pids; fi

    # top -b -n1 -p; -b for output without ANSI formating; -n1 for just one iteration of 'top'; -p for feeding processes from 'pgrep' command
    # Use LC_NUMERIC if your 'top' command outputs 'commas' instead 'dots' - with LC_NUMERIC you will get 'dots' during this script
    LC_NUMERIC=en_US.UTF-8 top -b -n1 -p $(cat /tmp/app_name_pids | xargs | tr ' ' ,) > /tmp/pstemp

    wc_l=$(wc -l < /tmp/pstemp)

    cpu_use=$(tail -n +8 /tmp/pstemp | tr -s ' ' | sed 's/^ *//' | cut -d' ' -f9 | xargs | tr ' ' + | bc)
    if [[ "$cpu_use" == "0" ]]; then
        if (( $(bc <<< "$cpu_use < 1") )); then cpu_use="0$cpu_use"; fi

    mem_use=$(tail -n +8 /tmp/pstemp | tr -s ' ' | sed 's/^ *//' | cut -d' ' -f10 | xargs | tr ' ' + | bc)
    if [[ "$mem_use" == "0" ]]; then
        if (( $(bc <<< "$mem_use < 1") )); then mem_use="0$mem_use"; fi

    echo -en "\033[2J\033[0;0f"
    # Use 'echo ...' above or 'tput ...' below (chose the one that works for you)
    #tput cup 0 0 && tput ed

    # Align Totals under %CPU and %MEM columns
    if (( $(bc <<< "$cpu_use < 1") )); then
        sed "${wc_l}a \\\n\nTotal (%CPU/%MEM): $(printf " %29s")$cpu_use  $mem_use\n${t_dsh}" /tmp/pstemp
    elif (( $(bc <<< "$cpu_use < 100") )); then
        sed "${wc_l}a \\\n\nTotal (%CPU/%MEM): $(printf " %28s")$cpu_use  $mem_use\n${t_dsh}" /tmp/pstemp
        sed "${wc_l}a \\\n\nTotal (%CPU/%MEM): $(printf " %27s")$cpu_use  $mem_use\n${t_dsh}" /tmp/pstemp

    if [[ $i -lt $mt_rep ]]; then sleep $sec_delay; fi
  • 1
    I am stomped that there is no common solution for this problem and/or nobody else is looking for this. But I am even more amazed by the beautiful and very helpful script you wrote. It perfectly addresses what I was looking for and I will use this very often. Thank you very much!
    – Y. Shallow
    Nov 1, 2021 at 13:22
  • You are welcome. I'm very glad that you found this script useful for you. When time allows I will change/add some things... I appreciate your comment/feedback.
    – Damir
    Nov 3, 2021 at 9:52
  • I tried using the script ./mytop.sh node and I get the following errors: (standard_in) 1: syntax error and sed: -e expression #1, char 2: invalid usage of line address 0. Not sure how to fix this Feb 28, 2022 at 8:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .