[Powertop] [powertop] ::wiggle() - what does that do?

[Powertop] [powertop] ::wiggle() - what does that do?

[Ivan Shapovalov]

[-- Attachment #1: Type: text/plain, Size: 381 bytes --]

Hello all,

I've been wondering on what does the abstract_cpu::wiggle() function does and 
why it is needed?.. All the code says is that it is needed to "have a record 
of CPU at the start and end of the perf trace".

I'm asking just because I feel somewhat uncomfortable to have powertop 
changing my scaling frequency back and forth each N seconds. :)

Thanks,
- Ivan

Re: [Powertop] [powertop] ::wiggle() - what does that do?

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 681 bytes --]

On 3/19/2013 12:51 PM, Ivan Shapovalov wrote:
> Hello all,
>
> I've been wondering on what does the abstract_cpu::wiggle() function does and
> why it is needed?.. All the code says is that it is needed to "have a record
> of CPU at the start and end of the perf trace".
>
> I'm asking just because I feel somewhat uncomfortable to have powertop
> changing my scaling frequency back and forth each N seconds. :)

the reason this exists is that powertop wants to track what frequency the cpu is at for the duration...
... and there's no way to ask for "current", only to get change notifications

so wiggle() forces a change... and from then on we know what it's at.

Re: [Powertop] [powertop] ::wiggle() - what does that do?

[Ivan Shapovalov]

[-- Attachment #1: Type: text/plain, Size: 942 bytes --]

On 19 March 2013 12:53:26 Arjan van de Ven wrote:
> On 3/19/2013 12:51 PM, Ivan Shapovalov wrote:
> > Hello all,
> > 
> > I've been wondering on what does the abstract_cpu::wiggle() function does
> > and why it is needed?.. All the code says is that it is needed to "have a
> > record of CPU at the start and end of the perf trace".
> > 
> > I'm asking just because I feel somewhat uncomfortable to have powertop
> > changing my scaling frequency back and forth each N seconds. :)
> 
> the reason this exists is that powertop wants to track what frequency the
> cpu is at for the duration... ... and there's no way to ask for "current",
> only to get change notifications
> 
> so wiggle() forces a change... and from then on we know what it's at.

...And why isn't it possible to keep the value from last notification?
Because time can pass between measurement end and new measurement start?

Sorry for stupid questions.

Re: [Powertop] [powertop] ::wiggle() - what does that do?

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 1015 bytes --]

On 3/19/2013 12:57 PM, Ivan Shapovalov wrote:
> On 19 March 2013 12:53:26 Arjan van de Ven wrote:
>> On 3/19/2013 12:51 PM, Ivan Shapovalov wrote:
>>> Hello all,
>>>
>>> I've been wondering on what does the abstract_cpu::wiggle() function does
>>> and why it is needed?.. All the code says is that it is needed to "have a
>>> record of CPU at the start and end of the perf trace".
>>>
>>> I'm asking just because I feel somewhat uncomfortable to have powertop
>>> changing my scaling frequency back and forth each N seconds. :)
>>
>> the reason this exists is that powertop wants to track what frequency the
>> cpu is at for the duration... ... and there's no way to ask for "current",
>> only to get change notifications
>>
>> so wiggle() forces a change... and from then on we know what it's at.
>
> ...And why isn't it possible to keep the value from last notification?

there may not have been one...

> Because time can pass between measurement end and new measurement start?

yep

Re: [Powertop] [powertop] ::wiggle() - what does that do?

[Ivan Shapovalov]

[-- Attachment #1: Type: text/plain, Size: 1170 bytes --]

On 19 March 2013 12:58:49 Arjan van de Ven wrote:
> On 3/19/2013 12:57 PM, Ivan Shapovalov wrote:
> > On 19 March 2013 12:53:26 Arjan van de Ven wrote:
> >> On 3/19/2013 12:51 PM, Ivan Shapovalov wrote:
> >>> Hello all,
> >>> 
> >>> I've been wondering on what does the abstract_cpu::wiggle() function
> >>> does
> >>> and why it is needed?.. All the code says is that it is needed to "have
> >>> a
> >>> record of CPU at the start and end of the perf trace".
> >>> 
> >>> I'm asking just because I feel somewhat uncomfortable to have powertop
> >>> changing my scaling frequency back and forth each N seconds. :)
> >> 
> >> the reason this exists is that powertop wants to track what frequency the
> >> cpu is at for the duration... ... and there's no way to ask for
> >> "current",
> >> only to get change notifications
> >> 
> >> so wiggle() forces a change... and from then on we know what it's at.
> > 
> > ...And why isn't it possible to keep the value from last notification?
> 
> there may not have been one...
> 
> > Because time can pass between measurement end and new measurement start?
> 
> yep

Thanks. It's clear to me now.