recording rapl/power events

recording rapl/power events

[Adrien BAK]

Hi everyone,

I am looking into the power (rapl) events introduced in the kernel in 
version 3.14.

Right now, I can access those events through perf stat :

perf stat -a -e power/energy-pkg/ sleep 1

  Performance counter stats for 'system wide':

              17.16 Joules power/energy-pkg/

        1.002808169 seconds time elapsed

However, it seems that I can't record a sequence of power events :

perf record -a -e power/energy-pkg/ sleep 1
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) 
for event (power/energy-pkg/).
/bin/dmesg may provide additional information.
No CONFIG_PERF_EVENTS=y kernel support configured?

dmesg does not give any information, and, as far as I know, perf_events 
are fully supported by my kernel (

Is there some limitation of the power events (or kernel support) I am 
unaware of that could explain this behaviour, or is there something 
wrong with my setup/configuration ?
How would one procede to record a sequence of power events ?

Thanks in advance
A.

Re: recording rapl/power events

[Vince Weaver]

On Mon, 21 Apr 2014, Adrien BAK wrote:

> However, it seems that I can't record a sequence of power events :
> 
> perf record -a -e power/energy-pkg/ sleep 1
> Error:
> The sys_perf_event_open() syscall returned with 22 (Invalid argument) for
> event (power/energy-pkg/).
> /bin/dmesg may provide additional information.
> No CONFIG_PERF_EVENTS=y kernel support configured?

RAPL events cannot generate interrupts, so I don't think you can sample 
like this.

You can possibly sample using some other event (like one of the software 
clock events) and include the RAPL event in the list of events to record, 
not sure if that will accomplish what you're trying to do.

Vince

Re: recording rapl/power events

[Adrien BAK]

On 21/04/2014 23:07, Vince Weaver wrote:
> On Mon, 21 Apr 2014, Adrien BAK wrote:
>
>> However, it seems that I can't record a sequence of power events :
>>
>> perf record -a -e power/energy-pkg/ sleep 1
>> Error:
>> The sys_perf_event_open() syscall returned with 22 (Invalid argument) for
>> event (power/energy-pkg/).
>> /bin/dmesg may provide additional information.
>> No CONFIG_PERF_EVENTS=y kernel support configured?
> RAPL events cannot generate interrupts, so I don't think you can sample
> like this.
>
> You can possibly sample using some other event (like one of the software
> clock events) and include the RAPL event in the list of events to record,
> not sure if that will accomplish what you're trying to do.
>
> Vince

This looks promising, although I couldn't find any useful documentation 
on this. Do you know if the perf userland tool allows for this kind of 
slave-sampling or if I have to build my own tool around the perf_event API ?

Re: recording rapl/power events

[Andi Kleen]

Adrien BAK <adrien.bak@metascale.org> writes:
>
> This looks promising, although I couldn't find any useful
> documentation on this. Do you know if the perf userland tool allows
> for this kind of slave-sampling or if I have to build my own tool
> around the perf_event API ?

perf record supports it (with :S), but none of the reporting tools
unfortunately.

However keep in mind that very rapid sampling of RAPL is not
likely to give useful results. The data only really makes
sense over longer intervals.

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only

Re: recording rapl/power events

[Namhyung Kim]

Hi Andi,

On Tue, 22 Apr 2014 03:17:56 -0700, Andi Kleen wrote:
> Adrien BAK <adrien.bak@metascale.org> writes:
>>
>> This looks promising, although I couldn't find any useful
>> documentation on this. Do you know if the perf userland tool allows
>> for this kind of slave-sampling or if I have to build my own tool
>> around the perf_event API ?
>
> perf record supports it (with :S), but none of the reporting tools
> unfortunately.

It seems it's what --group option of perf report is for?

  $ perf record -e '{cycles,power/energy-pkg/}:S' -a sleep 1

  $ perf report --group --stdio


Thanks,
Namhyung

>
> However keep in mind that very rapid sampling of RAPL is not
> likely to give useful results. The data only really makes
> sense over longer intervals.
>
> -Andi

Re: recording rapl/power events

[Adrien BAK]

hi  Namhyung,

yes it seems. However, when I run what you suggest, I get the following :

perf record -a -e '{cycles,power/energy-pkg/}:S' sleep 1
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) 
for event (power/energy-pkg/).
/bin/dmesg may provide additional information.
No CONFIG_PERF_EVENTS=y kernel support configured?

Once again, I'm pretty sure my kernel is correctly configured, as I can 
get power measurement using perf stat. Is this a bug somewhere ? Can you 
confirm that this works for you and that I should investigate what can 
go wrong with my system ?

A.


On 25/04/2014 17:29, Namhyung Kim wrote:
> Hi Andi,
>
> On Tue, 22 Apr 2014 03:17:56 -0700, Andi Kleen wrote:
>> Adrien BAK <adrien.bak@metascale.org> writes:
>>> This looks promising, although I couldn't find any useful
>>> documentation on this. Do you know if the perf userland tool allows
>>> for this kind of slave-sampling or if I have to build my own tool
>>> around the perf_event API ?
>> perf record supports it (with :S), but none of the reporting tools
>> unfortunately.
> It seems it's what --group option of perf report is for?
>
>    $ perf record -e '{cycles,power/energy-pkg/}:S' -a sleep 1
>
>    $ perf report --group --stdio
>
>
> Thanks,
> Namhyung
>
>> However keep in mind that very rapid sampling of RAPL is not
>> likely to give useful results. The data only really makes
>> sense over longer intervals.
>>
>> -Andi
> --
> To unsubscribe from this list: send the line "unsubscribe linux-perf-users" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html