sample - Process Profiling

Overview

sample will allow you to profile a process during runtime.

How it works

sample will retrieve the stack trace of all threads of a given process every 10 ms during a delay that was given by the user. This uses ptrace to get the stack trace information. These stack trace will be grouped by thread, then by callers to get a call tree . sample will count how many times a given call stack has been reported: this is the number on the left of the function name.

Limitations

- This currently works only on Linux x86 and x86_64.
- The process must be alive from the beginning to the end of the sample.

Usage

$ sample [pid] [delay-in-seconds] > sample-result.txt

When finished, the data will be shown as a call tree . The number on the left will show how much time was spent in each function.

By using the following command, the results will be shown in a GUI.

$ gtk-sample sample-result.txt

How to read

• You can see that etpan_lep_fetch_msg_list() call (33 units of time, unit of time is 10 ms) consists mainly in calling mailfolder_get_envelopes_list() (33 units of time).
  
  
• mailfolder_get_envelopes_list() call (33 units of time) consists mainly in calling mailsession_get_envelopes_list() (33 units of time).
  
  
• mailsession_get_envelopes_list() (33 units of time) splits time usage in 6 parts:
  • 13 units of time is spent in imapdriver_cached_get_envelopes_list(), get_flags_list().
  • 11 units of time is spent in imapdriver_cached_get_envelopes_list(), imapdriver_get_cached_envelope().
  • 6 units of time is spent in imapdriver_cached_get_envelopes_list(), maildriver_message_cached_clean_up().
  • The other parts can be neglected.
    

DINH V. Hoà - Sat Jul 7 16:37:36 2007