2024-02-14 21:17:58

by Thomas Gleixner

[permalink] [raw]
Subject: [RFD] Remove irq_timings

Daniel!

It's 7 years now that we merged irq timings into the kernel, but we
still have zero users for this.

I'm tempted to declare this experiment failed and remove the whole thing
for good.

Comments?

Thanks,

tglx


2024-02-14 21:39:20

by Daniel Lezcano

[permalink] [raw]
Subject: Re: [RFD] Remove irq_timings


Hi Thomas,


On 14/02/2024 22:17, Thomas Gleixner wrote:
> Daniel!
>
> It's 7 years now that we merged irq timings into the kernel, but we
> still have zero users for this.
>
> I'm tempted to declare this experiment failed and remove the whole thing
> for good.
>
> Comments?

I worked on an irq cpuidle governor which had better results than the
menu governor and equal than the teo governor. But I never succeed to
have better results without putting some arbitrary when computing the
next event.

At one moment, Daniel Bristot de Oliveira (Cc'ed) was thinking to may be
use it for the deadline scheduler.

Ulf (Cc'ed) may be has a plan for the next event for the CPU cluster.

But if no one has plan to use it, there is no good reason to keep it and
I'm fine if we remove it.



--
<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


2024-02-15 12:25:25

by Ulf Hansson

[permalink] [raw]
Subject: Re: [RFD] Remove irq_timings

On Wed, 14 Feb 2024 at 22:39, Daniel Lezcano <[email protected]> wrote:
>
>
> Hi Thomas,
>
>
> On 14/02/2024 22:17, Thomas Gleixner wrote:
> > Daniel!
> >
> > It's 7 years now that we merged irq timings into the kernel, but we
> > still have zero users for this.

Wow, is it really 7 years since then. :-)

> >
> > I'm tempted to declare this experiment failed and remove the whole thing
> > for good.
> >
> > Comments?
>
> I worked on an irq cpuidle governor which had better results than the
> menu governor and equal than the teo governor. But I never succeed to
> have better results without putting some arbitrary when computing the
> next event.
>
> At one moment, Daniel Bristot de Oliveira (Cc'ed) was thinking to may be
> use it for the deadline scheduler.
>
> Ulf (Cc'ed) may be has a plan for the next event for the CPU cluster.

Yes, I still have that plan, but haven't been able to run some real tests yet.

>
> But if no one has plan to use it, there is no good reason to keep it and
> I'm fine if we remove it.

Besides that the code isn't really used at the moment, is it also
blocking us from doing some cleanup/refactoring or other related code?

Kind regards
Uffe

2024-02-15 14:51:21

by Thomas Gleixner

[permalink] [raw]
Subject: Re: [RFD] Remove irq_timings

On Thu, Feb 15 2024 at 12:23, Ulf Hansson wrote:
> On Wed, 14 Feb 2024 at 22:39, Daniel Lezcano <[email protected]> wrote:
>> On 14/02/2024 22:17, Thomas Gleixner wrote:
>> > Daniel!
>> >
>> > It's 7 years now that we merged irq timings into the kernel, but we
>> > still have zero users for this.
>
> Wow, is it really 7 years since then. :-)
>
>> >
>> > I'm tempted to declare this experiment failed and remove the whole thing
>> > for good.
>> >
>> > Comments?
>>
>> I worked on an irq cpuidle governor which had better results than the
>> menu governor and equal than the teo governor. But I never succeed to
>> have better results without putting some arbitrary when computing the
>> next event.
>>
>> At one moment, Daniel Bristot de Oliveira (Cc'ed) was thinking to may be
>> use it for the deadline scheduler.
>>
>> Ulf (Cc'ed) may be has a plan for the next event for the CPU cluster.
>
> Yes, I still have that plan, but haven't been able to run some real tests yet.
>
>>
>> But if no one has plan to use it, there is no good reason to keep it and
>> I'm fine if we remove it.
>
> Besides that the code isn't really used at the moment, is it also
> blocking us from doing some cleanup/refactoring or other related code?

No. I just stumbled over it (again) and wondered whether it is ever
going to be used or not. If not there is no point to carry dead weight
around, but there is no hurry.

Thanks,

tglx


2024-03-07 11:55:04

by Ulf Hansson

[permalink] [raw]
Subject: Re: [RFD] Remove irq_timings

On Thu, 15 Feb 2024 at 15:46, Thomas Gleixner <[email protected]> wrote:
>
> On Thu, Feb 15 2024 at 12:23, Ulf Hansson wrote:
> > On Wed, 14 Feb 2024 at 22:39, Daniel Lezcano <[email protected]> wrote:
> >> On 14/02/2024 22:17, Thomas Gleixner wrote:
> >> > Daniel!
> >> >
> >> > It's 7 years now that we merged irq timings into the kernel, but we
> >> > still have zero users for this.
> >
> > Wow, is it really 7 years since then. :-)
> >
> >> >
> >> > I'm tempted to declare this experiment failed and remove the whole thing
> >> > for good.
> >> >
> >> > Comments?
> >>
> >> I worked on an irq cpuidle governor which had better results than the
> >> menu governor and equal than the teo governor. But I never succeed to
> >> have better results without putting some arbitrary when computing the
> >> next event.
> >>
> >> At one moment, Daniel Bristot de Oliveira (Cc'ed) was thinking to may be
> >> use it for the deadline scheduler.
> >>
> >> Ulf (Cc'ed) may be has a plan for the next event for the CPU cluster.
> >
> > Yes, I still have that plan, but haven't been able to run some real tests yet.
> >
> >>
> >> But if no one has plan to use it, there is no good reason to keep it and
> >> I'm fine if we remove it.
> >
> > Besides that the code isn't really used at the moment, is it also
> > blocking us from doing some cleanup/refactoring or other related code?
>
> No. I just stumbled over it (again) and wondered whether it is ever
> going to be used or not. If not there is no point to carry dead weight
> around, but there is no hurry.

Maybe a small chat about it at the next LPC? Then we can see if I have
been able to play more with it.

Kind regards
Uffe

2024-03-07 21:32:53

by Thomas Gleixner

[permalink] [raw]
Subject: Re: [RFD] Remove irq_timings

On Thu, Mar 07 2024 at 12:54, Ulf Hansson wrote:
> On Thu, 15 Feb 2024 at 15:46, Thomas Gleixner <[email protected]> wrote:
>> No. I just stumbled over it (again) and wondered whether it is ever
>> going to be used or not. If not there is no point to carry dead weight
>> around, but there is no hurry.
>
> Maybe a small chat about it at the next LPC? Then we can see if I have
> been able to play more with it.

Works for me.

Thanks,

tglx