[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  2006-07-07 09:26 -------
Created an attachment (id=1147)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=1147&action=view)
testing data Jim's patch

Hi Jim,
  I manually applied your patch to 2.6.17.4/ppc64, and tested on a 8-way ppc64
box. I use a multi-thread app which will create 8 threads and each thread will
call getsid() in a loop. The tests shows that the kretprobes still doesn't
scale well on SMP box. Here is the results:

<1> without being probed by stap:

root:/home/root/mt-getsid> ./app
cpu 5: loops=5000000, average=5566 ns
cpu 4: loops=5000000, average=5600 ns
cpu 0: loops=5000000, average=5868 ns
cpu 1: loops=5000000, average=5884 ns
cpu 6: loops=5000000, average=5886 ns
cpu 7: loops=5000000, average=5893 ns
cpu 3: loops=5000000, average=5937 ns
cpu 2: loops=5000000, average=5938 ns
Total cpus: loops = 40000000, average = 5822 ns

<2> probed by 'stap -e "probe syscall.getsid {}" -bM'

root:/home/root/mt-getsid> ./app
cpu 5: loops=5000000, average=7685 ns
cpu 4: loops=5000000, average=7685 ns
cpu 3: loops=5000000, average=7686 ns
cpu 1: loops=5000000, average=7685 ns
cpu 2: loops=5000000, average=7689 ns
cpu 6: loops=5000000, average=7690 ns
cpu 7: loops=5000000, average=7690 ns
cpu 0: loops=5000000, average=7694 ns
Total cpus: loops = 40000000, average = 7688 ns

<3> probed by 'stap -e "probe syscall.getsid.return {}" -bM'

root:/home/root/mt-getsid> ./app
cpu 4: loops=5000000, average=19807 ns
cpu 7: loops=5000000, average=19869 ns
cpu 1: loops=5000000, average=24322 ns
cpu 0: loops=5000000, average=24326 ns
cpu 3: loops=5000000, average=27886 ns
cpu 2: loops=5000000, average=27991 ns
cpu 6: loops=5000000, average=28986 ns
cpu 5: loops=5000000, average=29034 ns
Total cpus: loops = 40000000, average = 25277 ns

I also attached the patch I used, hope that I didn't miss something.

I ever used oprofile to sample the kretprobe, and the sample data is similar
with the data in this attachment.


-- 


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.