Date:   Wed, 22 Nov 2017 16:04:06 +0000
From:   Sudeep Holla <sudeep.holla@....com>
To:     Eduardo Valentin <edubezval@...il.com>
Cc:     Sudeep Holla <sudeep.holla@....com>,
        Javi Merino <javi.merino@...nel.org>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Lukasz Luba <llu.ker.dev@...il.com>,
        Daniel Lezcano <daniel.lezcano@...aro.org>,
        Ionela Voinescu <ionela.voinescu@....com>,
        Punit Agrawal <punit.agrawal@....com>,
        "Rafael J. Wysocki" <rjw@...ysocki.net>,
        Viresh Kumar <viresh.kumar@...aro.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Amit Daniel Kachhap <amit.kachhap@...il.com>,
        Zhang Rui <rui.zhang@...el.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ingo Molnar <mingo@...hat.com>,
        Linux PM <linux-pm@...r.kernel.org>,
        Lukasz Luba <lukasz.luba@....com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 4/4] cpu_cooling: Drop static-power related stuff



On 22/11/17 15:34, Eduardo Valentin wrote:
> On Wed, Nov 22, 2017 at 10:59:21AM +0000, Sudeep Holla wrote:
>> (sorry for chiming in quite late)
>>
>> On 21/11/17 18:12, Eduardo Valentin wrote:
>>> On Tue, Nov 21, 2017 at 06:00:07PM +0000, Javi Merino wrote:
>>>> Hi,
>>>>
>>>> On Tue, Nov 21, 2017 at 08:57:06AM -0800, Eduardo Valentin wrote:
>>>>
>>
>> [...]
>>
>>>>
>>>> In a nutshell, mainline does not want platform specific code, but we
>>>> haven't figured out how to calculate static power without platform
>>>> specific code.>
>>> To, that is still fine to have it as a callback, as long as you have at
>>> least one user! I still do not understand why Juno static power cannot
>>> go as platform code that register the callback to implement the static
>>> power model.
>>>
>>
>> 1. It was proved not so useful(anyone can prove otherwise ?)
> 
> Can anyone prove it does not have static power?
> 

I didn't claim that. I said it has been found that it's negligible based
on experiments.

>> 2. I am told static power is negligible compared to dynamic power with
>>    new fab processes.
> 
> I am told quantum computer is out there :-), does it mean we should drop the
> maintenance of everything else?
> 

Good comparison :)

Anyways all I told is if this code expects Juno to be user, then no as
there's no reliable way to test it.

It's entirely up to you if you want to support it or delete it as you
are the maintainer. I am bit confused here as you seem to imply you want
to continue supporting it here with you quantum comparison but say you
are fine to delete in other thread.

> 
>> 3. It's very hard to even test IPA on Juno as it doesn't reach the
>>    required critical temperature easily. So as Juno platform maintainer
>>    I want a test case to test regression before we merge anything.
>>
>> IMO, if the $subject code is expected to be used on Juno, then my answer
>> is no if one can't test it reliably and also prove that static power
>> really matters on Juno. So far, I have heard both the above is not
>> possible. So please delete the code if Juno is the only user in
>> short and mid term. We can get the code back if we find any users in
>> longer term.
> 
> Yeah, the fact that Juno takes time to reach crit temperature does not
> necessarily imply it does not have static power consumption, or that its
> static power consumption is negligible. Now, if you want to ignore it,
> because it is not the best example to show usefulness of IPA, that is a
> different story.
> 

OK, I agree that I want to ignore the usefulness of static power on Juno
as no one provides a reliable way to see that and test that regularly. I
am open to change my mind if circumstances change.

-- 
Regards,
Sudeep

Powered by blists - more mailing lists