This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
[Bug runtime/10665] backtrace() is too slow
- From: "alexl at redhat dot com" <sourceware-bugzilla at sourceware dot org>
- To: systemtap at sources dot redhat dot com
- Date: 19 Jan 2010 10:54:05 -0000
- Subject: [Bug runtime/10665] backtrace() is too slow
- References: <20090918213942.10665.wcohen@redhat.com>
- Reply-to: sourceware-bugzilla at sourceware dot org
------- Additional Comments From alexl at redhat dot com 2010-01-19 10:54 -------
ubacktrace() is no speed demon either, hopefully performance fixes to
backtrace() fixes ubacktrace() too.
Also, moving as much work as possible to the translator is very important. I had
a test script that at each GObject construction time the stored the backtrace in
an array based on the pointer, then freed it at finalize time. Then at program
shutdown we can print the stacktraces of all leaked objects. This means almost
no backtraces are actually translated, so any computation that can be moved to
translation is a huge gain.
I *was* able to run the above script, but i had to disable/increase all sort of
system limitations for it to work.
--
http://sourceware.org/bugzilla/show_bug.cgi?id=10665
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.