Hi, all,
I'm using the oprofile plug-in on Kepler. The version of the OProfile Core
Plug-in is 2.0.0.201309171939. The plug-in appears to ignore a user-specified
count value. Here are the steps to reproduce:
1. Create a simple C project to be profiled.
2. Right click on the project, select "Profiling Tools | Profiling Tools
Configurations".
3. Create a new "Profile with OProfile" configuration.
4. Under the "Global" tab, select 'operf'.
5. Under the "Events" tab, uncheck the "Use default event" box, check the
"Enabled" box in the Ctr 0 tab, select an event (like CPU_CLK_UNHALTED for
Intel systems), and then edit the "Count" field, specifying something like
"500000".
6. Click the "Apply" and "Profile" buttons.
During the profiling session, I see a bunch of messages like below spewing to
every open terminal window:
kernel:Dazed and confused, but trying to continue
This may or may not happen when trying to reproduce the problem, depending on
kernel version and processor model. My system is RHEL 6.4/Core 2 Duo.
I then went into my Eclipse workspace project directory and found the
"oprofile_data" directory where the profile data was stored. Running
'opreport' from that directory showed the profile data below [1]. Note the
'count 6000' at the end of the line where the event name and description are
given. This value of 6000 is the minimum value for the CPU_CLK_UNHALTED event.
So the oprofile plug-in ignored the '500000' value that I specified for the
'count' value. I tried other events and got the same result -- i.e., the
minimum value was always used for profiling, no matter what I specified in the
'count' field. BTW, you can see the minimum value for each event in the output
of ophelp.
Is this a bug or am I misunderstanding something about how to specify the
profile configuration?
Thanks!
-Maynard
----------------------------------------------
[1]
[maynard@oc3431575272 memcpyt]$ opreport
Using /home/maynard/workspace/memcpyt/oprofile_data/samples/ for samples
directory.
WARNING! Some of the events were throttled. Throttling occurs when
the initial sample rate is too high, causing an excessive number of
interrupts. Decrease the sampling frequency. Check the directory
/home/maynard/workspace/memcpyt/oprofile_data/samples/current/stats/throttled
for the throttled event names.
CPU: Core 2, speed 2534 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit mask
of 0x00 (Unhalted core cycles) count 6000
CPU_CLK_UNHALT...|
samples| %|
------------------
2279376 100.000 memcpyt
CPU_CLK_UNHALT...|
samples| %|
------------------
2089905 91.6876 libc-2.12.so
147437 6.4683 memcpyt
41995 1.8424 no-vmlinux
39 0.0017 ld-2.12.so
_______________________________________________
linuxtools-dev mailing list
linuxtools-dev@eclipse.org
https://dev.eclipse.org/mailman/listinfo/linuxtools-dev
