[Bug kprobes/2062] Return probes does not scale well on SMP box

["guanglei at cn dot ibm dot com" <sourceware-bugzilla at sourceware dot org>]

------- Additional Comments From guanglei at cn dot ibm dot com  2006-07-08 05:14 -------
(In reply to comment #18)
> 
> This looks correct -- Thanks! -- except that it appears you've introduced
> spaces, in some places, where there should be tabs.  Grep for "       " (7
> consecutive spaces).

sorry, because I used mouse to copy/paste your codes

(In reply to comment #18)
> Created an attachment (id=1149)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=1149&action=view)
> kprobes module - equivalent of 'probe syscall.getsid.return {}'
> 
> Guanglei, could you please re-run your test with this module inserted, rather
> than the SystemTap script running?  Thanks.

I tested using stap and the module generated from getsid.c separately. I ran
each test about 4 times,  here are the results:

====== 2.6.17.3/ppc64/8-way, without Jim's patch ================

no probe
  Total cpus: loops = 40000000, average = 6202 ns

kretprobe using stap:
  Total cpus: loops = 40000000, average = 43702 ns

kretprobe using getsid.c:
  Total cpus: loops = 40000000, average = 36456 ns


======= 2.6.17.4/ppc64/8-way, with Jim's patch ===================

kretprobe using stap:
  Total cpus: loops = 40000000, average = 26621 ns

kretprobe using getsid.c:
  Total cpus: loops = 40000000, average = 24975 ns

==================================================================

The tests shows that the patch could help the performance improvement( ~30%
improvement), but kretprobe is still expensive.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2062

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.