Intel Core i3 support

------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security
threats, fraudulent activity, and more. Splunk takes this data and makes
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2dcopy2

_______________________________________________
oprofile-list mailing list
oprofile-list <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oprofile-list

[PATCH] Remove old unused kernel module code

On 09/22/2011 02:08 PM, William Cohen wrote:
> On 09/21/2011 12:30 PM, Maynard Johnson wrote:
>> William Cohen wrote:
>>> Hi All,
>>>
>>> I have been thinking list of things that should be addressed in the
>>> next version of oprofile.  This is just a quick list of things that
>>> come to mind. I am sure that other people have additional ideas for
>>> oprofile. This is a start to the discussion to improve oprofile
>>> rather than an exact list.
>>
>> Thanks for the great suggestions!  Comments below.
>>
>> -Maynard
> 
> Hi Maynard,
> 
>>
>>>
>>> -Will
>>>
>>> Eliminate the Linux module in oprofile user-space package
>>>
>>> OProfile support has been in the Linux kernel for a very long
>>> time. Carrying around the kernel module in the user-space package
>>> doesn't make much sense.  Getting rid of it will remove about 260KB of
>>> dead code.
>> Agree.  Can you open a bug report for this?  Perhaps you could assign it to yourself if you have time.
> 
> Added Bugs item #3412940 for this.

I have a patch for this. It is a bit large (190KB) to send on the mailing list. It is attached to the bug #3412940:

0001-Remove-the-old-oprofile-linux-kernel-drivers-in-OPro.patch

If it looks okay, I can check it in. That will make OProfile a bit more compact.

-Will

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

RE: Kernel compilation and call graph

I can do nothing concerning my Linux kernel version. This is a given condition I must adapt to.

I tried to dig out what this issue might be related to (e.g. --with-kernel-support flag for the user mode
module, kernel compilation various flags etc.) and tried everything, with no success unfortunately.
Although my Pleistocene scenario I could backport stuff from 2.6.31 for supporting Nehalem CPU (thanks
guys for the assistance some months ago) so it seems 2.6.18 is still a bit capable... Yet no success with
callgraph for the kernel.

I you guys can direct me to any relevant idea this would be a great help. We must be able to analyze the kernel
performance in a resolution more elaborate then sample count.

Thanks for your help,

Avi

-----Original Message-----
From: Andi Kleen [mailto:andi <at> firstfloor.org] 
Sent: Friday, September 30, 2011 2:03 AM
To: Maynard Johnson
Cc: Andi Kleen; Avi Gozlan; 'oprofile-list <at> lists.sourceforge.net'
Subject: Re: Kernel compilation and call graph

> Do you have any clues why Avi gets no kernel callgraph info for his Intel 
> processors running on kernel version 2.6.18?

No, I've long forgotten most things about such old kernels.
My recommendation would be to update to mainline and not engage
in software archeology.

-Andi

Scanned by Check Point Total Security Gateway.

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

Re: Kernel compilation and call graph

> I you guys can direct me to any relevant idea this would be a great help. We must be able to analyze the kernel
performance in a resolution more elaborate then sample count.

Debug the backtracer the hard way. Add printks, find
out what goes wrong. Fix it. Repeat.

-Andi

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

Re: [PATCH] Remove old unused kernel module code

On 10/03/2011 4:58 PM, William Cohen wrote:
> On 09/22/2011 02:08 PM, William Cohen wrote:
>> On 09/21/2011 12:30 PM, Maynard Johnson wrote:
>>> William Cohen wrote:
>>>> Hi All,
>>>>
>>>> I have been thinking list of things that should be addressed in the
>>>> next version of oprofile.  This is just a quick list of things that
>>>> come to mind. I am sure that other people have additional ideas for
>>>> oprofile. This is a start to the discussion to improve oprofile
>>>> rather than an exact list.
>>>
>>> Thanks for the great suggestions!  Comments below.
>>>
>>> -Maynard
>>
>> Hi Maynard,
>>
>>>
>>>>
>>>> -Will
>>>>
>>>> Eliminate the Linux module in oprofile user-space package
>>>>
>>>> OProfile support has been in the Linux kernel for a very long
>>>> time. Carrying around the kernel module in the user-space package
>>>> doesn't make much sense.  Getting rid of it will remove about 260KB of
>>>> dead code.
>>> Agree.  Can you open a bug report for this?  Perhaps you could assign it to yourself if you have time.
>>
>> Added Bugs item #3412940 for this.
>
> I have a patch for this. It is a bit large (190KB) to send on the mailing list. It is attached to the bug #3412940:
>
> 0001-Remove-the-old-oprofile-linux-kernel-drivers-in-OPro.patch
>
> If it looks okay, I can check it in. That will make OProfile a bit more compact.

Thanks for the patch, Will.  I'll review it today.

-Maynard

>
> -Will
>

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

[ oprofile-Bugs-3412940 ] Remove old oprofile kernel module code

Bugs item #3412940, was opened at 2011-09-22 10:47
Message generated for change (Comment added) made by maynardj
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=116191&aid=3412940&group_id=16191

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
Status: Open
Resolution: None
Priority: 3
Private: No
Submitted By: William Cohen (wcohen)
Assigned to: Nobody/Anonymous (nobody)
Summary: Remove old oprofile kernel module code

Initial Comment:
The upstream linux kernels have had oprofile support for many years. At this point there is little reason
keeping the oprofile/module code in the user-space oprofile. 

----------------------------------------------------------------------

>Comment By: Maynard Johnson (maynardj)
Date: 2011-10-04 10:03

Message:
Will, thanks again for putting this patch together.  The attached patch
looks fine, but it's not quite complete. There are a lot of doc changes
that are needed, too (mostly in the manual, but the oprofile man page as
well).  As a matter of fact, there are so many changes needed, that it's
probably easier for me to make the changes myself rather than write them
down here and ask you to make them.  I'll work up a doc patch and attach it
to this bug for you to review.

Also, regarding the image-path stuff for finding 2.6 kernel modules. . .
If we're not supporting older kernels anymore, should we default to
/lib/modules/`uname -r` unless a different path is passed in by the user? 
The downside is that if the user of an oprofile post-processing tool is
booted with a different kernel than what was used to collect the raw
profile data, we would silently attribute samples to (possibly) wrong
kernel modules.  This probably doesn't happen very often.  If we highlight
this in the man pages, do you think such a default behavior would be OK?  I
could go either way.

If we decide to make such a change in default behavior for locating kernel
modules, we would need to scrub the man pages for various tools, which
currently say something to the effect of "This is needed to find modules in
kernels 2.6 and upwards".  This would have to be changed to read something
like "This is needed to find the correct kernel modules in use at the time
of profiling."   I think this whole "image-path" issue could be handled via
a separate bug since it's not really dependent on the old kernel module
code -- it's just tangentially related.  What do you think?

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=116191&aid=3412940&group_id=16191

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

Re: Intel Core i3 support

On 10/02/2011 3:03 PM, Jeff Eastlack wrote:
> Hi Guys,
>
> Thanks for your help on the last question! ... I have a 2 more questions with
> using oprofile in my core-i3 processor.
>
> 1)oprofile states that it doesn't know my processor very well. The core-i3
> processor is a dual core Nehalem architecture. There are several more events
> then what is listed below, but I get this message "This is a limited set of
> fallback events because oprofile doesn't know your CPU"

I *think* that processor is supported with oprofile 0.9.7, which you can 
download at:
http://oprofile.sourceforge.net/download/.
If that doesn't work, then Andi (on cc) can probably provide a simple patch to 
make it work.

>
> 2) my second question is how do I have different counters counting different
> events during the same sampling run? For example the Nehalem has more than one
> counter but when I use the following command it assigns to counter:0 by default
>
> sudo opcontrol--event=CPU_CLK_UNHALTED:10000:0x0:0:1
>
> sudo opcontrol--event=INST_RETIRED:10000:0x0:0:1

Please see the man page for opcontrol.  The description for the --event option says:
           Note  that  this  over-rides  all  previous events  selected;
           if  you  want two or more counters used simultaneously, you
           must specify them on the same opcontrol invocation.

-Maynard

>
> How do I get both of these events counted at the same time by different counters?
>
>
> ---------------------the limited Nehalem events
> supported-----------------------------------------------
> jeff <at> east:~$ sudo opcontrol --list-events
> oprofile: available events for CPU type "Intel Architectural Perfmon"
>
> See Intel 64 and IA-32 Architectures Software Developer's Manual
> Volume 3B (Document 253669) Chapter 18 for architectural perfmon events
> This is a limited set of fallback events because oprofile doesn't know your CPU
> CPU_CLK_UNHALTED: (counter: all)
> Clock cycles when not halted (min count: 6000)
> INST_RETIRED: (counter: all)
> number of instructions retired (min count: 6000)
> LLC_MISSES: (counter: all)
> Last level cache demand requests from this core that missed the LLC (min count:
> 6000)
> Unit masks (default 0x41)
> ----------
> 0x41: No unit mask
> LLC_REFS: (counter: all)
> Last level cache demand requests from this core (min count: 6000)
> Unit masks (default 0x4f)
> ----------
> 0x4f: No unit mask
> BR_INST_RETIRED: (counter: all)
> number of branch instructions retired (min count: 500)
> BR_MISS_PRED_RETIRED: (counter: all)
> number of mispredicted branches retired (precise) (min count: 500)
>
>
> ------------------------------------------------------------------------------
> All of the data generated in your IT infrastructure is seriously valuable.
> Why? It contains a definitive record of application performance, security
> threats, fraudulent activity, and more. Splunk takes this data and makes
> sense of it. IT sense. And common sense.
> http://p.sf.net/sfu/splunk-d2dcopy2
>
>
>
> _______________________________________________
> oprofile-list mailing list
> oprofile-list <at> lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/oprofile-list

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

Re: Intel Core i3 support

> I *think* that processor is supported with oprofile 0.9.7, which you can 
> download at:
> http://oprofile.sourceforge.net/download/.
> If that doesn't work, then Andi (on cc) can probably provide a simple patch 
> to make it work.

Unless he has a IvyBridge he should be ok.

-Andi

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

Re: [PATCH] Remove old unused kernel module code

To all oprofile community members:

The patch referenced below will remove support for pre-2.6 kernels.  There may 
not be any oprofile users still stuck in the stone age, but if there are, you 
could still continue to get current oprofile userspace support by 
reverse-applying the patch that Will added to bug #3412940 
(https://sourceforge.net/tracker/?func=detail&atid=116191&aid=3412940&group_id=16191). 
  But no formal support is promised for such a patched oprofile -- i.e., use as 
is.  Speak up now if you have any issue with this plan.

Regards,
-Maynard

On 10/03/2011 4:58 PM, William Cohen wrote:
> On 09/22/2011 02:08 PM, William Cohen wrote:
>> On 09/21/2011 12:30 PM, Maynard Johnson wrote:
>>> William Cohen wrote:
>>>> Hi All,
>>>>
>>>> I have been thinking list of things that should be addressed in the
>>>> next version of oprofile.  This is just a quick list of things that
>>>> come to mind. I am sure that other people have additional ideas for
>>>> oprofile. This is a start to the discussion to improve oprofile
>>>> rather than an exact list.
>>>
>>> Thanks for the great suggestions!  Comments below.
>>>
>>> -Maynard
>>
>> Hi Maynard,
>>
>>>
>>>>
>>>> -Will
>>>>
>>>> Eliminate the Linux module in oprofile user-space package
>>>>
>>>> OProfile support has been in the Linux kernel for a very long
>>>> time. Carrying around the kernel module in the user-space package
>>>> doesn't make much sense.  Getting rid of it will remove about 260KB of
>>>> dead code.
>>> Agree.  Can you open a bug report for this?  Perhaps you could assign it to yourself if you have time.
>>
>> Added Bugs item #3412940 for this.
>
> I have a patch for this. It is a bit large (190KB) to send on the mailing list. It is attached to the bug #3412940:
>
> 0001-Remove-the-old-oprofile-linux-kernel-drivers-in-OPro.patch
>
> If it looks okay, I can check it in. That will make OProfile a bit more compact.
>
> -Will
>

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

Re: Intel Core i3 support

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2dcopy1

_______________________________________________
oprofile-list mailing list
oprofile-list <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oprofile-list