These are the same as trace-cmd-record(1) with the --profile
option.
-p tracer
Set a tracer plugin to run instead of function graph tracing
set to depth of 1. To not run any tracer, use -p nop.
-S
Only enable the tracer or events speficied on the command
line. With this option, the function_graph tracer is not
enabled, nor are any events (like sched_switch), unless they
are specifically specified on the command line (i.e. -p
function -e sched_switch -e sched_wakeup)
-G
Set interrupt (soft and hard) events as global (associated to
CPU instead of tasks).
-o file
Write the output of the profile to file. This supersedes
--stderr
-H event-hooks
Add custom event matching to connect any two events together.
Format is:
[<start_system>:]<start_event>,<start_match>[,<start_pid>]/
[<end_system>:]<end_event>,<end_match>[,<flags>]
The start_system:start_event (start_system is optional), is the event that
starts the timing.
start_match is the field in the start event that is to match with the
end_match in the end event.
start_pid is optional, as matches are attached to the tasks that run
the events, if another field should be used to find that task, then
it is specified with start_pid.
end_system:end_event is the event that ends the timing (end_system is
optional).
end_match is the field in end_match that wil match the start event field
start_match.
flags are optional and can be the following (case insensitive):
p : The two events are pinned to the same CPU (start and end happen
on the same CPU always).
s : The event should have a stack traced with it (enable stack tracing
for the start event).
g : The event is global (not associated to a task). start_pid is
not applicable with this flag.
--stderr
Redirect the output to stderr. The output of the command
being executed is not changed. This allows watching the
command execute and saving the output of the profile to
another file.
