Re: Removal of events for unsupported architectures or processor models

Maynard,

It would take sometimes for me to add IBS support in operf, and I don't 
want to hold back releasing the 1.0 RC. However, I would still be 
reusing the IBS event/unitmask files for operf.  But I guess you would 
rather adding the event files back when the IBS operf is ready.  If so, 
I'll provide the removal patch.

Besides the IBS events, do you also want IBS removal patches in other 
components (e.g. daemon/opd_ibs*, doc, utils/opcontrol).

Thanks,
Suravee

On Wed 30 Jul 2014 09:36:17 AM CDT, Maynard Johnson wrote:
> On 07/29/2014 06:57 PM, Suravee Suthikulanit wrote:
>> Maynard,
>>
>> IBS actually supports in perf subsystem. Let me look in to the current code and get back to you if we can extend this to operf.  However, I should be able to provide patch to remove IBS from from the opcontrol.
> Hi, Suravee.  That's good news that IBS is supported in the perf_events subsystem. Currently, opcontrol handles IBS events in do_param_setup_ibs() by writing to certain IBS-specific files in oprofilefs (for the kernel driver to interpret).  I presume you would need to add the functionality into libpe_utils/op_pe_utils.cpp.  If you need one of the extended "config" fileds from perf_event_attr (i.e., 'config1' or 'config2'), then you need to put a check into configure.ac whether that field is available in the version of perf_event.h being used (since the early versions of perf_event.h only had a 'config' field).  And, of course, you'll need some way to check whether the running kernel has the perf_events IBS support.
>
> I guess it's not been a high priority for you to add IBS support to operf, and it's a bit late now, just days before I want to put out a 1.0 release candidate.  But if you can provide a patch quickly, I'll be happy to delay the RC a bit if necessary.  Please give me an estimate of when you might be able to provide such a patch.  If you aren't able to provide such a patch, then a patch that removes the IBS events from the x86-64 events files would be appreciated.
>
> Thanks.
> -Maynard
>
>>
>> Let me know what you think?
>>
>> Thank you,
>>
>> Suravee
>>
>>
>>
>> On 7/29/2014 9:31 AM, Maynard Johnson wrote:
>>> On 07/21/2014 06:08 PM, Maynard Johnson wrote:
>>>>> Hello, oprofile community,
>>>>> Your input and feedback on the issue described below would be appreciated.
>>>>>
>>>>> As recently announced on the oprofile-list, OProfile release 1.0 will*NOT*  include the legacy opcontrol-based profiler.  Only the operf profiler -- which uses the Linux Kernel Performance Events Subsystem (aka "perf_events") -- will be included in release 1.0.  It seems logical at this time to also remove the events and unit mask files for any architecture or processor types that are not supported by perf_events.  From what I can tell, here are the architectures that do not have perf_events support:
>>>>>
>>> Suravee,
>>> When I originally posted this message, I should have mentioned AMD IBS events will no longer be supported with oprofile 1.0 either, since (AFAIK) there is no support for IBS in the perf_events kernel subsystem -- and, thus, IBS is supported only by opcontrol.
>>>
>>> I would prefer that you provide a patch to remove the IBS events from the various x86_64 events files.  I plan on rolling out a release candidate by the end of the week, so if I don't get a patch from you by then, I'll have to remove those events myself and hope I don't screw anything up.
>>>
>>> -Maynard
>>>
>>
>>
>

------------------------------------------------------------------------------
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds