Re: [PATCH 2/3] sched: idle: Add sched balance option

[Daniel Lezcano <daniel.lezcano@linaro.org>]

On 04/29/2014 01:11 AM, Rafael J. Wysocki wrote:
> On Monday, April 28, 2014 01:07:31 PM Daniel Lezcano wrote:
>> On 04/28/2014 12:28 PM, Peter Zijlstra wrote:
>>> On Mon, Apr 28, 2014 at 12:09:20PM +0200, Daniel Lezcano wrote:
>>>> I agree a numerical value is not flexible. But it sounds weird to put a
>>>> scheduler option in the sysfs and maybe more options will follow.
>>>>
>>>> I am wondering if we shouldn't create a new cgroup for 'energy' and put
>>>> everything in there. So we will have more flexibility for extension and we
>>>> will be able to create a group of tasks for performance and a group of tasks
>>>> for energy saving.
>>>>
>>>> Does it make sense ?
>>>
>>> The old knobs used to live here:
>>>
>>> -What:          /sys/devices/system/cpu/sched_mc_power_savings
>>> -               /sys/devices/system/cpu/sched_smt_power_savings
>>
>> Ah right.
>>
>>> Not entirely sure that's a fine place, but it has precedent.
>>
>> I share your doubts about the right place.
>>
>> I'm really wondering if the cgroup couldn't be a good solution:
>>
>> Amit pointed the conflict about the power vs performance with some
>> applications. We want to have for example a game to run fast performance
>> and some other application to save power.
>
> You can't save power.
>
> Power is the energy flow *rate*.  It's like speed, so how can you save it?
>
> If you talk about saving in this context, please always talk about energy as
> well, because that's what we want to save.

Hi Rafael,

yeah, I think there is an abuse when talking about 'power'. I thought I 
took care of talking about energy but 'power' comes always in my mind. I 
believe the confusion is coming from the meaning of 'power' in French 
where one translation is 'energy'.

Anyway, thanks for the clarification, I will try to use the term 
'energy' and 'power' conveniently next time.

> This means that positioning power against performance doesn't make any sense
> whatsoever.  You could try to position energy efficiency (that is, the relative
> cost of doing work in terms of energy) against performance, but even that is
> questionable, because, as I said in one of the previous messages, what is good
> for performance is often good for energy efficiency too (think about race to
> idle for example).
> In other words, you want to have a knob whose both ends may happen to mean
> the same thing.  Wouldn't that be a little odd?

Yes, I share this point of view. I believe we won't care about adding 
any knobs for this situation.

> In my opinion it would be much better to have a knob representing the current
> relative value of energy to the user (which may depend on things like whether
> or not the system is on battery etc) and meaning how far we need to go with
> energy saving efforts.
>
> So if that knob is 0, we'll do things that are known-good for performance.
> If it is 1, we'll do some extra effort to save enery as well possibly at
> a small expense of performance if that's necessary.  If it is 100, we'll do
> all we can to save as much energy as possible without caring about performance
> at all.
>
> And it doesn't even have to be scheduler-specific, it very well may be global.

That would be very nice but I don't see how we can quantify this energy 
and handle that generically from the kernel for all the hardware.

I am pretty sure we will discover for some kind of hardware a specific 
option will consume more power, argh ! energy I mean, than another 
hardware because of the architecture.

 From my personal experience, when we are facing this kind of complexity 
and heuristic, it is the sign the userspace has some work to do.

What I am proposing is not in contradiction with your approach, it is 
about exporting a lot of knobs to userspace, and the userspace decide 
how to map what is '0' <--> '100' regarding these options. Nothing 
prevent the different platform to set a default value for these options.

 From my POV, the cgroup could be a good solution for that for different 
reasons. Especially one good reason is we can stick the energy policy 
per task instead of the entire system.

Let's imagine the following scenario:

An user has a laptop running a mailer looking for the email every 5 
minutes. The system switched to 'power'. The user wants to play a video 
game but due to the 'power' policy, the game is not playable so it 
forces the policy to 'performance'. All the tasks will use the 
'performance' policy, thus consuming more energy.

If we do per task, the video game will use the 'performance' policy and 
the other tasks on the system will use the 'power' policy. The userspace 
can take the decision to freeze the application running 'performance' if 
we reach a critical battery level.

The cgroup is a good framework to do that and gives a lot of flexibility 
to userspace. I understood Peter does not like the cgroup but I did not 
give up to convince him, the cgroup could be good solution :)

Looking forward, if the energy policy is tied with the task, in the 
future we can normalize the energy consumption and stick to an 'energy 
load' per task and reuse the load tracking for energy, do per task 
energy accounting, nice per energy, etc ...

Going back to reality, concretely this sysctl patch did not reach a 
consensus. So I will resend the two other patches, hoping the discussion 
will lead to an agreement.



-- 
  <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog