BIND 10 #1563: document how to profile - oprofile

BIND 10 Development do-not-reply at isc.org
Wed Jan 18 14:16:52 UTC 2012


#1563: document how to profile - oprofile
-------------------------------------+-------------------------------------
                   Reporter:  jelte  |                 Owner:  jelte
                       Type:  task   |                Status:  reviewing
                   Priority:  major  |             Milestone:
                  Component:         |  Sprint-20120124
  Unclassified                       |            Resolution:
                   Keywords:         |             Sensitive:  0
            Defect Severity:  N/A    |           Sub-Project:  DNS
Feature Depending on Ticket:         |  Estimated Difficulty:  0
        Add Hours to Ticket:  0      |           Total Hours:  0
                  Internal?:  0      |
-------------------------------------+-------------------------------------
Changes (by vorner):

 * owner:  vorner => jelte


Comment:

 Hello

 Replying to [comment:6 jelte]:
 > - opcontrol needs to be run as root (with sudo in my case), I'm guessing
 it is setuid on your system?

 No, I did say it needs to run as root in the introduction as one of the
 disadvantages. But I added the note in parentheses before the opcontrol
 prompts.

 > - as noted, IBS_OP_ALL is an AMD-specific setting. It does not work on
 my i5, and it is kind of half mentioned, but if one changes the -e values,
 the settings are retained until changed again (even through reboots). I
 was playing around with some random ones from the i7 list, and could not
 get any results at all anymore (no samples error) :) Probably worth
 mentioning that you can revert them with '-e default'

 I added a note about it and about the almost equivalent counter.

 > - at least for the run with query_bench, I had to call opreport and
 opannotate on .libs/lt-query_bench, not .lib/query_bench

 Hmm, this might be system dependant. I added a note.

 > Not sure if we should just mention the systemspecific points or perhaps
 think of some architecture-dependent notes as a sort of appendix.

 That would be a little bit short I think.

 Thanks

-- 
Ticket URL: <http://bind10.isc.org/ticket/1563#comment:7>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development


More information about the bind10-tickets mailing list