Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Friday 28 Apr 2017 à 16:52:59 (+0100), Morten Rasmussen a écrit : > Hi Vincent, > > Sorry for crashing the party this late. As you know, it takes a long > period of uninterrupted review time to properly review PELT stuff. > > Disclaimer: I haven't read the rest of the thread yet. > > On Mon,

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Friday 28 Apr 2017 à 16:52:59 (+0100), Morten Rasmussen a écrit : > Hi Vincent, > > Sorry for crashing the party this late. As you know, it takes a long > period of uninterrupted review time to properly review PELT stuff. > > Disclaimer: I haven't read the rest of the thread yet. > > On Mon,

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 1 May 2017 at 11:00, Peter Zijlstra wrote: > On Sat, Apr 29, 2017 at 12:09:24AM +0200, Peter Zijlstra wrote: >> On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: >> > +++ b/include/linux/sched.h >> > @@ -313,6 +313,7 @@ struct load_weight { >> > */ >> >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 1 May 2017 at 11:00, Peter Zijlstra wrote: > On Sat, Apr 29, 2017 at 12:09:24AM +0200, Peter Zijlstra wrote: >> On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: >> > +++ b/include/linux/sched.h >> > @@ -313,6 +313,7 @@ struct load_weight { >> > */ >> > struct sched_avg { >>

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Sat, Apr 29, 2017 at 12:09:24AM +0200, Peter Zijlstra wrote: > On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > > +++ b/include/linux/sched.h > > @@ -313,6 +313,7 @@ struct load_weight { > > */ > > struct sched_avg { > > u64 last_update_time;

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Sat, Apr 29, 2017 at 12:09:24AM +0200, Peter Zijlstra wrote: > On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > > +++ b/include/linux/sched.h > > @@ -313,6 +313,7 @@ struct load_weight { > > */ > > struct sched_avg { > > u64 last_update_time;

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > +++ b/include/linux/sched.h > @@ -313,6 +313,7 @@ struct load_weight { > */ > struct sched_avg { > u64 last_update_time; > + u64 stolen_idle_time; > u64

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > +++ b/include/linux/sched.h > @@ -313,6 +313,7 @@ struct load_weight { > */ > struct sched_avg { > u64 last_update_time; > + u64 stolen_idle_time; > u64

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 28/04/17 16:52, Morten Rasmussen wrote: > Hi Vincent, [...] > As mentioned above, waiting time, i.e. !running && weight, is not > scaled, which causes trouble for load. I ran some rt-app-based tests on a system with frequency and cpu invariance. (1) Two periodic 20% tasks with 12ms period

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 28/04/17 16:52, Morten Rasmussen wrote: > Hi Vincent, [...] > As mentioned above, waiting time, i.e. !running && weight, is not > scaled, which causes trouble for load. I ran some rt-app-based tests on a system with frequency and cpu invariance. (1) Two periodic 20% tasks with 12ms period

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Hi Vincent, Sorry for crashing the party this late. As you know, it takes a long period of uninterrupted review time to properly review PELT stuff. Disclaimer: I haven't read the rest of the thread yet. On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > The current

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Hi Vincent, Sorry for crashing the party this late. As you know, it takes a long period of uninterrupted review time to properly review PELT stuff. Disclaimer: I haven't read the rest of the thread yet. On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > The current

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 14 April 2017 at 10:49, Vincent Guittot wrote: > On 13 April 2017 at 18:13, Peter Zijlstra wrote: >> On Thu, Apr 13, 2017 at 05:16:20PM +0200, Vincent Guittot wrote: >>> On 13 April 2017 at 15:39, Peter Zijlstra wrote:

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 14 April 2017 at 10:49, Vincent Guittot wrote: > On 13 April 2017 at 18:13, Peter Zijlstra wrote: >> On Thu, Apr 13, 2017 at 05:16:20PM +0200, Vincent Guittot wrote: >>> On 13 April 2017 at 15:39, Peter Zijlstra wrote: >> >>> > OK, so the reason util_avg varies is because we compute it

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 13 April 2017 at 18:13, Peter Zijlstra wrote: > On Thu, Apr 13, 2017 at 05:16:20PM +0200, Vincent Guittot wrote: >> On 13 April 2017 at 15:39, Peter Zijlstra wrote: > >> > OK, so the reason util_avg varies is because we compute it wrong. And I >> >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 13 April 2017 at 18:13, Peter Zijlstra wrote: > On Thu, Apr 13, 2017 at 05:16:20PM +0200, Vincent Guittot wrote: >> On 13 April 2017 at 15:39, Peter Zijlstra wrote: > >> > OK, so the reason util_avg varies is because we compute it wrong. And I >> > think we can easily fix that once we pull

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 13 April 2017 at 20:06, Peter Zijlstra wrote: > On Thu, Apr 13, 2017 at 04:59:15PM +0200, Vincent Guittot wrote: >> On 13 April 2017 at 15:32, Peter Zijlstra wrote: >> > On Wed, Apr 12, 2017 at 01:28:58PM +0200, Peter Zijlstra wrote: >> > >> >> I

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 13 April 2017 at 20:06, Peter Zijlstra wrote: > On Thu, Apr 13, 2017 at 04:59:15PM +0200, Vincent Guittot wrote: >> On 13 April 2017 at 15:32, Peter Zijlstra wrote: >> > On Wed, Apr 12, 2017 at 01:28:58PM +0200, Peter Zijlstra wrote: >> > >> >> I still wonder about the whole !running vs

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Thu, Apr 13, 2017 at 04:59:15PM +0200, Vincent Guittot wrote: > On 13 April 2017 at 15:32, Peter Zijlstra wrote: > > On Wed, Apr 12, 2017 at 01:28:58PM +0200, Peter Zijlstra wrote: > > > >> I still wonder about the whole !running vs !weight thing., > > > > Ah, since we

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Thu, Apr 13, 2017 at 04:59:15PM +0200, Vincent Guittot wrote: > On 13 April 2017 at 15:32, Peter Zijlstra wrote: > > On Wed, Apr 12, 2017 at 01:28:58PM +0200, Peter Zijlstra wrote: > > > >> I still wonder about the whole !running vs !weight thing., > > > > Ah, since we use this for both util

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Thu, Apr 13, 2017 at 05:16:20PM +0200, Vincent Guittot wrote: > On 13 April 2017 at 15:39, Peter Zijlstra wrote: > > OK, so the reason util_avg varies is because we compute it wrong. And I > > think we can easily fix that once we pull out all the factors (which > > would

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Thu, Apr 13, 2017 at 05:16:20PM +0200, Vincent Guittot wrote: > On 13 April 2017 at 15:39, Peter Zijlstra wrote: > > OK, so the reason util_avg varies is because we compute it wrong. And I > > think we can easily fix that once we pull out all the factors (which > > would mean your patch and

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 13 April 2017 at 15:39, Peter Zijlstra wrote: > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > >> > Secondly, what's up with the util_sum < LOAD_AVG_MAX * 1000 thing? >> >> The lost idle time makes sense only if the task can also be "idle" >> when

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 13 April 2017 at 15:39, Peter Zijlstra wrote: > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > >> > Secondly, what's up with the util_sum < LOAD_AVG_MAX * 1000 thing? >> >> The lost idle time makes sense only if the task can also be "idle" >> when running at max capacity.

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 13 April 2017 at 15:32, Peter Zijlstra wrote: > On Wed, Apr 12, 2017 at 01:28:58PM +0200, Peter Zijlstra wrote: > >> I still wonder about the whole !running vs !weight thing., > > Ah, since we use this for both util _and_ load, we need !running && > !weight, and it so

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 13 April 2017 at 15:32, Peter Zijlstra wrote: > On Wed, Apr 12, 2017 at 01:28:58PM +0200, Peter Zijlstra wrote: > >> I still wonder about the whole !running vs !weight thing., > > Ah, since we use this for both util _and_ load, we need !running && > !weight, and it so happens that !weight

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > Secondly, what's up with the util_sum < LOAD_AVG_MAX * 1000 thing? > > The lost idle time makes sense only if the task can also be "idle" > when running at max capacity. When util_sum reaches the >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > Secondly, what's up with the util_sum < LOAD_AVG_MAX * 1000 thing? > > The lost idle time makes sense only if the task can also be "idle" > when running at max capacity. When util_sum reaches the >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Wed, Apr 12, 2017 at 01:28:58PM +0200, Peter Zijlstra wrote: > I still wonder about the whole !running vs !weight thing., Ah, since we use this for both util _and_ load, we need !running && !weight, and it so happens that !weight implies !running. Is that it?

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Wed, Apr 12, 2017 at 01:28:58PM +0200, Peter Zijlstra wrote: > I still wonder about the whole !running vs !weight thing., Ah, since we use this for both util _and_ load, we need !running && !weight, and it so happens that !weight implies !running. Is that it?

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 12 April 2017 at 17:44, Peter Zijlstra wrote: > On Wed, Apr 12, 2017 at 04:50:47PM +0200, Vincent Guittot wrote: >> Le Wednesday 12 Apr 2017 à 13:28:58 (+0200), Peter Zijlstra a écrit : > >> > >> > |-|-| (wall-time) >> > -

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 12 April 2017 at 17:44, Peter Zijlstra wrote: > On Wed, Apr 12, 2017 at 04:50:47PM +0200, Vincent Guittot wrote: >> Le Wednesday 12 Apr 2017 à 13:28:58 (+0200), Peter Zijlstra a écrit : > >> > >> > |-|-| (wall-time) >> > - F=100% >> >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Wed, Apr 12, 2017 at 04:50:47PM +0200, Vincent Guittot wrote: > Le Wednesday 12 Apr 2017 à 13:28:58 (+0200), Peter Zijlstra a écrit : > > > > |-|-| (wall-time) > > - F=100% > > **--- F= 66% > > |--||

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Wed, Apr 12, 2017 at 04:50:47PM +0200, Vincent Guittot wrote: > Le Wednesday 12 Apr 2017 à 13:28:58 (+0200), Peter Zijlstra a écrit : > > > > |-|-| (wall-time) > > - F=100% > > **--- F= 66% > > |--||

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Wednesday 12 Apr 2017 à 13:28:58 (+0200), Peter Zijlstra a écrit : > On Tue, Apr 11, 2017 at 03:09:20PM +0200, Vincent Guittot wrote: > > Le Tuesday 11 Apr 2017 à 12:49:49 (+0200), Peter Zijlstra a écrit : > > > > > > Lets go back to the unscaled version: > > > > > > running idle > > >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Wednesday 12 Apr 2017 à 13:28:58 (+0200), Peter Zijlstra a écrit : > On Tue, Apr 11, 2017 at 03:09:20PM +0200, Vincent Guittot wrote: > > Le Tuesday 11 Apr 2017 à 12:49:49 (+0200), Peter Zijlstra a écrit : > > > > > > Lets go back to the unscaled version: > > > > > > running idle > > >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 03:09:20PM +0200, Vincent Guittot wrote: > Le Tuesday 11 Apr 2017 à 12:49:49 (+0200), Peter Zijlstra a écrit : > > > > Lets go back to the unscaled version: > > > > running idle > >|*|-| > > > > With the current code, that would effectively end

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 03:09:20PM +0200, Vincent Guittot wrote: > Le Tuesday 11 Apr 2017 à 12:49:49 (+0200), Peter Zijlstra a écrit : > > > > Lets go back to the unscaled version: > > > > running idle > >|*|-| > > > > With the current code, that would effectively end

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Tuesday 11 Apr 2017 à 12:49:49 (+0200), Peter Zijlstra a écrit : > > Lets go back to the unscaled version: > > running idle >|*|-| > > With the current code, that would effectively end up like (again > assuming 50%): > > running idle >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Tuesday 11 Apr 2017 à 12:49:49 (+0200), Peter Zijlstra a écrit : > > Lets go back to the unscaled version: > > running idle >|*|-| > > With the current code, that would effectively end up like (again > assuming 50%): > > running idle >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Tuesday 11 Apr 2017 à 12:41:36 (+0200), Peter Zijlstra a écrit : > On Tue, Apr 11, 2017 at 11:40:21AM +0200, Vincent Guittot wrote: > > Le Tuesday 11 Apr 2017 à 10:53:05 (+0200), Peter Zijlstra a écrit : > > > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > > > Le Monday 10

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Tuesday 11 Apr 2017 à 12:41:36 (+0200), Peter Zijlstra a écrit : > On Tue, Apr 11, 2017 at 11:40:21AM +0200, Vincent Guittot wrote: > > Le Tuesday 11 Apr 2017 à 10:53:05 (+0200), Peter Zijlstra a écrit : > > > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > > > Le Monday 10

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Lets go back to the unscaled version: running idle |*|-| With the current code, that would effectively end up like (again assuming 50%): running idle |*_*_*_*_*|-| Time stays the same, but we add extra idle cycles.

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Lets go back to the unscaled version: running idle |*|-| With the current code, that would effectively end up like (again assuming 50%): running idle |*_*_*_*_*|-| Time stays the same, but we add extra idle cycles.

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 11:40:21AM +0200, Vincent Guittot wrote: > Le Tuesday 11 Apr 2017 à 10:53:05 (+0200), Peter Zijlstra a écrit : > > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > > Le Monday 10 Apr 2017 à 19:38:02 (+0200), Peter Zijlstra a écrit : > > > > > > > >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 11:40:21AM +0200, Vincent Guittot wrote: > Le Tuesday 11 Apr 2017 à 10:53:05 (+0200), Peter Zijlstra a écrit : > > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > > Le Monday 10 Apr 2017 à 19:38:02 (+0200), Peter Zijlstra a écrit : > > > > > > > >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 11 April 2017 at 11:12, Peter Zijlstra wrote: > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > >> > > + } else if (!weight) { >> > > + if (sa->util_sum < (LOAD_AVG_MAX * 1000)) { >> > >> > But here I'm completely lost. WTF just happened ;-) >>

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On 11 April 2017 at 11:12, Peter Zijlstra wrote: > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > >> > > + } else if (!weight) { >> > > + if (sa->util_sum < (LOAD_AVG_MAX * 1000)) { >> > >> > But here I'm completely lost. WTF just happened ;-) >> > >> > Firstly, I

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Tuesday 11 Apr 2017 à 10:53:05 (+0200), Peter Zijlstra a écrit : > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > Le Monday 10 Apr 2017 à 19:38:02 (+0200), Peter Zijlstra a écrit : > > > > > > Thanks for the rebase. > > > > > > On Mon, Apr 10, 2017 at 11:18:29AM +0200,

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Tuesday 11 Apr 2017 à 10:53:05 (+0200), Peter Zijlstra a écrit : > On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > Le Monday 10 Apr 2017 à 19:38:02 (+0200), Peter Zijlstra a écrit : > > > > > > Thanks for the rebase. > > > > > > On Mon, Apr 10, 2017 at 11:18:29AM +0200,

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > > + } else if (!weight) { > > > + if (sa->util_sum < (LOAD_AVG_MAX * 1000)) { > > > > But here I'm completely lost. WTF just happened ;-) > > > > Firstly, I think we want a comment on why we care about the !weight > >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > > > + } else if (!weight) { > > > + if (sa->util_sum < (LOAD_AVG_MAX * 1000)) { > > > > But here I'm completely lost. WTF just happened ;-) > > > > Firstly, I think we want a comment on why we care about the !weight > >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > Le Monday 10 Apr 2017 à 19:38:02 (+0200), Peter Zijlstra a écrit : > > > > Thanks for the rebase. > > > > On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > > > > Ok, so let me try and paraphrase what this patch

Re: [PATCH v2] sched/fair: update scale invariance of PELT

On Tue, Apr 11, 2017 at 09:52:21AM +0200, Vincent Guittot wrote: > Le Monday 10 Apr 2017 à 19:38:02 (+0200), Peter Zijlstra a écrit : > > > > Thanks for the rebase. > > > > On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > > > > Ok, so let me try and paraphrase what this patch

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Monday 10 Apr 2017 à 19:38:02 (+0200), Peter Zijlstra a écrit : > > Thanks for the rebase. > > On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > > Ok, so let me try and paraphrase what this patch does. > > So consider a task that runs 16 out of our 32ms window: > >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Le Monday 10 Apr 2017 à 19:38:02 (+0200), Peter Zijlstra a écrit : > > Thanks for the rebase. > > On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: > > Ok, so let me try and paraphrase what this patch does. > > So consider a task that runs 16 out of our 32ms window: > >

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Thanks for the rebase. On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: Ok, so let me try and paraphrase what this patch does. So consider a task that runs 16 out of our 32ms window: running idle |-|-| You're saying that when we scale running with the

Re: [PATCH v2] sched/fair: update scale invariance of PELT

Thanks for the rebase. On Mon, Apr 10, 2017 at 11:18:29AM +0200, Vincent Guittot wrote: Ok, so let me try and paraphrase what this patch does. So consider a task that runs 16 out of our 32ms window: running idle |-|-| You're saying that when we scale running with the

[PATCH v2] sched/fair: update scale invariance of PELT

The current implementation of load tracking invariance scales the contribution with current frequency and uarch performance (only for utilization) of the CPU. One main result of this formula is that the figures are capped by current capacity of CPU. Another one is that the load_avg is not

[PATCH v2] sched/fair: update scale invariance of PELT

The current implementation of load tracking invariance scales the contribution with current frequency and uarch performance (only for utilization) of the CPU. One main result of this formula is that the figures are capped by current capacity of CPU. Another one is that the load_avg is not