top(1)top(1)NAMEtop - display and update sorted information about processes
SYNOPSIStop [-a | -d | -e | -c <mode>]
[-F | -f]
[-h]
[-i <interval>]
[-l <samples>]
[-ncols <columns>]
[-o <key>] [-O <skey>]
[-R | -r]
[-S]
[-s <delay>]
[-n <nprocs>]
[-stats <keys>]
[-pid <processid>]
[-user <username>]
[-U <username>]
[-u]
DESCRIPTION
The top program periodically displays a sorted list of system process‐
es. The default sorting key is pid, but other keys can be used in‐
stead. Various output options are available.
OPTIONS
Command line option specifications are processed from left to right.
Options can be specified more than once. If conflicting options are
specified, later specifications override earlier ones. This makes it
viable to create a shell alias for top with preferred defaults speci‐
fied, then override those preferred defaults as desired on the command
line.
-a Equivalent to -c a.
-c <mode>
Set event counting mode to <mode>. The supported modes are:
a Accumulative mode. Count events cumulatively, starting
at the launch of top. Calculate CPU usage and CPU time
since the launch of top.
d Delta mode. Count events relative to the previous sam‐
ple. Calculate CPU usage since the previous sample.
This mode by default disables the memory object map re‐
porting. The memory object map reporting may be re-en‐
abled with the -r option or the interactive r command.
e Absolute mode. Count events using absolute counters.
n Non-event mode (default). Calculate CPU usage since the
previous sample.
-d Equivalent to -c d.
-e Equivalent to -c e.
-F Do not calculate statistics on shared libraries, also known as
frameworks.
-f Calculate statistics on shared libraries, also known as frame‐
works (default).
-h Print command line usage information and exit.
-i <interval>
Update framework (-f) info every <interval> samples; see the
PERFORMANCE vs. ACCURACY section below for more details.
-l <samples>
Use logging mode and display <samples> samples, even if standard
output is a terminal. 0 is treated as infinity. Rather than
redisplaying, output is periodically printed in raw form. Note
that the first sample displayed will have an invalid %CPU dis‐
played for each process, as it is calculated using the delta be‐
tween samples.
-ncols <columns>
Display <columns> when using logging mode. The default is infi‐
nite. The number must be > 0 or an error will occur.
-n <nprocs>
Only display up to <nprocs> processes.
-O <skey>
Use <skey> as a secondary key when ordering the process display.
See -o for key names (pid is the default).
-o <key>
Order the process display by sorting on <key> in descending or‐
der. A + or - can be prefixed to the key name to specify as‐
cending or descending order, respectively. The supported keys
are:
pid Process ID (default).
command
Command name.
cpu CPU usage.
csw Number of context switches.
time Execution time.
threads
alias: th
Number of threads (total/running).
ports alias: prt
Number of Mach ports.
mregion
alias: mreg, reg
Number of memory regions.
mem Internal memory size.
rprvt Resident private address space size.
purg Purgeable memory size.
vsize Total memory size.
vprvt Private address space size.
kprvt Private kernel memory size.
kshrd Shared kernel memory size.
pgrp Process group id.
ppid Parent process id.
state alias: pstate
Process state.
uid User ID.
wq alias: #wq, workqueue
The workqueue total/running.
faults alias: fault
The number of page faults.
cow alias: cow_faults
The copy-on-write faults.
user alias: username Username.
msgsent
Total number of mach messages sent.
msgrecv
Total number of mach messages received.
sysbsd Total BSD syscalls.
sysmach
Total Mach syscalls.
pageins
Total pageins.
-R Do not traverse and report the memory object map for each
process.
-r Traverse and report the memory object map for each process (de‐
fault).
-S Display the global statistics for swap and purgeable memory.
-s <delay>
Set the delay between updates to <delay> seconds. The default
delay between updates is 1 second.
-stats <keys>
Only display the comma separated statistics. See the -o flag
for the valid <keys>.
-pid <processid>
Only display <processid> in top.
-user <user>
Only display processes owned by <user>.
-U <user>
This is an alias for -user.
-u This is an alias equivalent to: -o cpu -O time.
DISPLAY
The first several lines of the top display show various global state.
All of the information is labeled. Following is an alphabetical list
of global state fields and their descriptions.
CPU Percentage of processor usage, broken into user, system,
and idle components. The time period for which these per‐
centages are calculated depends on the event counting mode.
Disks Number and total size of disk reads and writes.
LoadAvg Load average over 1, 5, and 15 minutes. The load average
is the average number of jobs in the run queue.
MemRegions Number and total size of memory regions, and total size of
memory regions broken into private (broken into non-library
and library) and shared components.
Networks Number and total size of input and output network packets.
PhysMem Physical memory usage, broken into wired, active, inactive,
used, and free components.
Procs Total number of processes and number of processes in each
process state.
SharedLibs Resident sizes of code and data segments, and link editor
memory usage.
Threads Number of threads.
Time Time, in H:MM:SS format. When running in logging mode Time
is in YYYY/MM/DD HH:MM:SS format by default, but may be
overridden with accumulative mode. When running in accumu‐
lative event counting mode, the Time is in HH:MM:SS since
the beginning of the top process.
VirtMem Total virtual memory, virtual memory consumed by shared li‐
braries, and number of pageins and pageouts.
Swap Swap usage: total size of swap areas, amount of swap space
in use and amount of swap space available.
Purgeable Number of pages purged and number of pages currently purge‐
able.
Below the global state fields, a list of processes is displayed. The
fields that are displayed depend on the options that are set. The pid
field displays the following for the architecture:
+ for 64-bit native architecture, or - for 32-bit native
architecture, or * for a non-native architecture.
INTERACTION
When top is run in interactive (non-logging) mode, it is possi‐
ble to control the output of top, as well as interactively send
signals to processes. The interactive command syntax is terse.
Each command is one character, followed by 0 to 2 arguments.
Commands that take arguments prompt interactively for the argu‐
ments, and where applicable, the default value is shown in
square brackets. The default value can be selected by leaving
the input field blank and pressing enter. ^G escapes the inter‐
active argument prompt, and has the same effect as leaving the
input field blank and pressing enter.
The following commands are supported:
? Display the help screen. Any character exits help screen mode.
This command always works, even in the middle of a command.
^L Redraw the screen.
c<mode>
Set output mode to <mode>. The supported modes are:
a Accumulative mode.
d Delta mode.
e Event mode.
n Non-event mode.
O<skey>
Use <skey> as a secondary key when ordering the process display.
See the -o option for key names.
o<key>
Order the process display by sorting on <key> in descending or‐
der. A + or - can be prefixed to the key name to specify as‐
cending or descending order, respectively. The supported keys
and alises are listed with the -o option above.
q Quit.
r Toggle traversal and reporting of the memory object map for each
process.
S<signal><pid>
Send <sig> to <pid>. <sig> can be specified either as a number
or as a name (for example, HUP). The default signal starts out
as TERM. Each time a signal is successfully sent, the default
signal is updated to be that signal. <pid> is a process id.
s<delay>
Set the delay between updates to <delay> seconds.
U<user>
Only display processes owned by <user>. Either the username or
uid number can be specified. To display all processes, press
enter without entering a username or uid number.
PERFORMANCE vs. ACCURACY
Calculating detailed memory statistics is fundamentally resource-inten‐
sive. To reduce the cpu usage in top, the -i parameter has been intro‐
duced to allow the user to tune this tradeoff. With the default value
of 10, framework stats will be updated once every 10 samples. Specify‐
ing -i 1 will result in the most accurate display, at the expense of
system resources.
N/A - Not Available
When this occurs in a stat it's caused by the memory object map report‐
ing being disabled. Memory object map reporting is disabled by default
in delta mode, but may be optionally enabled via -r or the interactive
r command. To enable the -r option use it after any -c mode options.
EXAMPLEStop-o cpu -O +rsize -s 5 -n 20
Sort the processes according to CPU usage (descending) and resi‐
dent memory size (ascending), sample and update the display at 5
second intervals, and limit the display to 20 processes.
top-c d
Run top in delta mode.
top-stats pid,command,cpu,th,pstate,time
Display only the specified statistics, regardless of any growth
of the terminal. If the terminal is too small, only the statis‐
tics that fit will be displayed.
SEE ALSOkill(2), vm_stat(1), signal(3), vmmap(1)toptop(1)