Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 5 February 2018 at 23:18, Valentin Schneider wrote: > On 01/30/2018 11:41 AM, Valentin Schneider wrote: >> [...] >>> I have studied a way to keep track of how many cpus still have blocked >>> load to try to minimize the number of useless ilb kick but this add >>>

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 5 February 2018 at 23:18, Valentin Schneider wrote: > On 01/30/2018 11:41 AM, Valentin Schneider wrote: >> [...] >>> I have studied a way to keep track of how many cpus still have blocked >>> load to try to minimize the number of useless ilb kick but this add >>> more atomic operations which

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 01/30/2018 11:41 AM, Valentin Schneider wrote: > [...] >> I have studied a way to keep track of how many cpus still have blocked >> load to try to minimize the number of useless ilb kick but this add >> more atomic operations which can impact the system throughput with >> heavy load and lot of

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 01/30/2018 11:41 AM, Valentin Schneider wrote: > [...] >> I have studied a way to keep track of how many cpus still have blocked >> load to try to minimize the number of useless ilb kick but this add >> more atomic operations which can impact the system throughput with >> heavy load and lot of

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 1 February 2018 at 19:10, Peter Zijlstra wrote: > On Wed, Jan 24, 2018 at 09:25:36AM +0100, Vincent Guittot wrote: >> @@ -8861,7 +8875,14 @@ static int idle_balance(struct rq *this_rq, struct >> rq_flags *rf) >> update_next_balance(sd, _balance); >>

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 1 February 2018 at 19:10, Peter Zijlstra wrote: > On Wed, Jan 24, 2018 at 09:25:36AM +0100, Vincent Guittot wrote: >> @@ -8861,7 +8875,14 @@ static int idle_balance(struct rq *this_rq, struct >> rq_flags *rf) >> update_next_balance(sd, _balance); >>

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Mon, Jan 29, 2018 at 07:31:07PM +, Valentin Schneider wrote: > But for load update via _nohz_idle_balance(), we iterate through all of the > nohz CPUS and unconditionally call update_blocked_averages(). This could be > avoided by remembering which CPUs have stale load before going idle. >

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Mon, Jan 29, 2018 at 07:31:07PM +, Valentin Schneider wrote: > But for load update via _nohz_idle_balance(), we iterate through all of the > nohz CPUS and unconditionally call update_blocked_averages(). This could be > avoided by remembering which CPUs have stale load before going idle. >

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Wed, Jan 24, 2018 at 09:25:36AM +0100, Vincent Guittot wrote: > @@ -8861,7 +8875,14 @@ static int idle_balance(struct rq *this_rq, struct > rq_flags *rf) > update_next_balance(sd, _balance); > rcu_read_unlock(); > > - if (time_after(jiffies,

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Wed, Jan 24, 2018 at 09:25:36AM +0100, Vincent Guittot wrote: > @@ -8861,7 +8875,14 @@ static int idle_balance(struct rq *this_rq, struct > rq_flags *rf) > update_next_balance(sd, _balance); > rcu_read_unlock(); > > - if (time_after(jiffies,

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 1 February 2018 at 17:57, Peter Zijlstra wrote: > On Wed, Jan 24, 2018 at 09:25:36AM +0100, Vincent Guittot wrote: >> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c >> index 898785d..ed90303 100644 >> --- a/kernel/sched/fair.c >> +++ b/kernel/sched/fair.c >> @@

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 1 February 2018 at 17:57, Peter Zijlstra wrote: > On Wed, Jan 24, 2018 at 09:25:36AM +0100, Vincent Guittot wrote: >> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c >> index 898785d..ed90303 100644 >> --- a/kernel/sched/fair.c >> +++ b/kernel/sched/fair.c >> @@ -7356,6 +7356,17 @@

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 1 February 2018 at 17:52, Peter Zijlstra wrote: > On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: > > Would've probably been easier to read if you'd not included the revert > of that timer patch... > >> @@ -9258,21 +9255,11 @@ void

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 1 February 2018 at 17:52, Peter Zijlstra wrote: > On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: > > Would've probably been easier to read if you'd not included the revert > of that timer patch... > >> @@ -9258,21 +9255,11 @@ void nohz_balance_enter_idle(int cpu) >>

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Wed, Jan 24, 2018 at 09:25:36AM +0100, Vincent Guittot wrote: > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c > index 898785d..ed90303 100644 > --- a/kernel/sched/fair.c > +++ b/kernel/sched/fair.c > @@ -7356,6 +7356,17 @@ static inline bool cfs_rq_is_decayed(struct cfs_rq > *cfs_rq)

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Wed, Jan 24, 2018 at 09:25:36AM +0100, Vincent Guittot wrote: > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c > index 898785d..ed90303 100644 > --- a/kernel/sched/fair.c > +++ b/kernel/sched/fair.c > @@ -7356,6 +7356,17 @@ static inline bool cfs_rq_is_decayed(struct cfs_rq > *cfs_rq)

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Thu, Jan 18, 2018 at 10:38:07AM +, Morten Rasmussen wrote: > It seems pointless to have a timer to update PELT if the system is > completely idle, and when it isn't we can piggy back other events to > make the updates happen. Only if we do that update before making decisions based on the

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Thu, Jan 18, 2018 at 10:38:07AM +, Morten Rasmussen wrote: > It seems pointless to have a timer to update PELT if the system is > completely idle, and when it isn't we can piggy back other events to > make the updates happen. Only if we do that update before making decisions based on the

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: Would've probably been easier to read if you'd not included the revert of that timer patch... > @@ -9258,21 +9255,11 @@ void nohz_balance_enter_idle(int cpu) > set_cpu_sd_state_idle(cpu); > > /* > - * implies a

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: Would've probably been easier to read if you'd not included the revert of that timer patch... > @@ -9258,21 +9255,11 @@ void nohz_balance_enter_idle(int cpu) > set_cpu_sd_state_idle(cpu); > > /* > - * implies a

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 30 January 2018 at 12:41, Valentin Schneider wrote: > (Resending because I snuck in some HTML... Apologies) > > On 01/30/2018 08:32 AM, Vincent Guittot wrote: >> >> On 29 January 2018 at 20:31, Valentin Schneider >> wrote: >>> >>> Hi

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 30 January 2018 at 12:41, Valentin Schneider wrote: > (Resending because I snuck in some HTML... Apologies) > > On 01/30/2018 08:32 AM, Vincent Guittot wrote: >> >> On 29 January 2018 at 20:31, Valentin Schneider >> wrote: >>> >>> Hi Vincent, Peter, >>> >>> I've been running some tests on

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

(Resending because I snuck in some HTML... Apologies) On 01/30/2018 08:32 AM, Vincent Guittot wrote: On 29 January 2018 at 20:31, Valentin Schneider wrote: Hi Vincent, Peter, I've been running some tests on your patches (Peter's base + the 2 from Vincent). The

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

(Resending because I snuck in some HTML... Apologies) On 01/30/2018 08:32 AM, Vincent Guittot wrote: On 29 January 2018 at 20:31, Valentin Schneider wrote: Hi Vincent, Peter, I've been running some tests on your patches (Peter's base + the 2 from Vincent). The results themselves are hosted

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 29 January 2018 at 20:31, Valentin Schneider wrote: > Hi Vincent, Peter, > > I've been running some tests on your patches (Peter's base + the 2 from > Vincent). The results themselves are hosted at [1]. > The base of those tests is the same: a task ("accumulator")

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 29 January 2018 at 20:31, Valentin Schneider wrote: > Hi Vincent, Peter, > > I've been running some tests on your patches (Peter's base + the 2 from > Vincent). The results themselves are hosted at [1]. > The base of those tests is the same: a task ("accumulator") is ran for 5 > seconds

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 29 January 2018 at 19:43, Dietmar Eggemann wrote: > On 01/24/2018 09:25 AM, Vincent Guittot wrote: >> >> Hi, >> >> Le Thursday 18 Jan 2018 à 10:38:07 (+), Morten Rasmussen a écrit : >>> >>> On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote:

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 29 January 2018 at 19:43, Dietmar Eggemann wrote: > On 01/24/2018 09:25 AM, Vincent Guittot wrote: >> >> Hi, >> >> Le Thursday 18 Jan 2018 à 10:38:07 (+), Morten Rasmussen a écrit : >>> >>> On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: Le Wednesday 03 Jan 2018

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Hi Vincent, Peter, I've been running some tests on your patches (Peter's base + the 2 from Vincent). The results themselves are hosted at [1]. The base of those tests is the same: a task ("accumulator") is ran for 5 seconds (arbitrary value) to accumulate some load, then goes to sleep for .5

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Hi Vincent, Peter, I've been running some tests on your patches (Peter's base + the 2 from Vincent). The results themselves are hosted at [1]. The base of those tests is the same: a task ("accumulator") is ran for 5 seconds (arbitrary value) to accumulate some load, then goes to sleep for .5

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 01/24/2018 09:25 AM, Vincent Guittot wrote: Hi, Le Thursday 18 Jan 2018 à 10:38:07 (+), Morten Rasmussen a écrit : On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : [...] Hi Peter, With the

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 01/24/2018 09:25 AM, Vincent Guittot wrote: Hi, Le Thursday 18 Jan 2018 à 10:38:07 (+), Morten Rasmussen a écrit : On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : [...] Hi Peter, With the

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Hi, Le Thursday 18 Jan 2018 à 10:38:07 (+), Morten Rasmussen a écrit : > On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: > > Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : > > > Hi Peter, > > > > > > On 22 December 2017 at 21:42, Peter Zijlstra

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Hi, Le Thursday 18 Jan 2018 à 10:38:07 (+), Morten Rasmussen a écrit : > On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: > > Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : > > > Hi Peter, > > > > > > On 22 December 2017 at 21:42, Peter Zijlstra

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 January 2018 at 10:40, Dietmar Eggemann wrote: > On 01/15/2018 08:26 AM, Vincent Guittot wrote: >> >> Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : >>> >>> Hi Peter, >>> >>> On 22 December 2017 at 21:42, Peter Zijlstra

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 January 2018 at 10:40, Dietmar Eggemann wrote: > On 01/15/2018 08:26 AM, Vincent Guittot wrote: >> >> Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : >>> >>> Hi Peter, >>> >>> On 22 December 2017 at 21:42, Peter Zijlstra >>> wrote: On Fri, Dec 22, 2017 at

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 01/15/2018 08:26 AM, Vincent Guittot wrote: Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : Hi Peter, On 22 December 2017 at 21:42, Peter Zijlstra wrote: On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: Right; but I figured we'd

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 01/15/2018 08:26 AM, Vincent Guittot wrote: Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : Hi Peter, On 22 December 2017 at 21:42, Peter Zijlstra wrote: On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: Right; but I figured we'd try and do it 'right'

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: > Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : > > Hi Peter, > > > > On 22 December 2017 at 21:42, Peter Zijlstra wrote: > > > On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Mon, Jan 15, 2018 at 09:26:09AM +0100, Vincent Guittot wrote: > Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : > > Hi Peter, > > > > On 22 December 2017 at 21:42, Peter Zijlstra wrote: > > > On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: > > >> Right;

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Mon, Jan 15, 2018 at 10:43:18AM +0100, Peter Zijlstra wrote: > On Tue, Jan 02, 2018 at 03:44:57PM +, Morten Rasmussen wrote: > > > Vincent already proposed, why can't we just modify Brendan's > > CPU_NEWLY_IDLE proposal to do a stats update from idle_balance() every > > 32ms regardless of

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Mon, Jan 15, 2018 at 10:43:18AM +0100, Peter Zijlstra wrote: > On Tue, Jan 02, 2018 at 03:44:57PM +, Morten Rasmussen wrote: > > > Vincent already proposed, why can't we just modify Brendan's > > CPU_NEWLY_IDLE proposal to do a stats update from idle_balance() every > > 32ms regardless of

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Tue, Jan 02, 2018 at 03:44:57PM +, Morten Rasmussen wrote: > Vincent already proposed, why can't we just modify Brendan's > CPU_NEWLY_IDLE proposal to do a stats update from idle_balance() every > 32ms regardless of whether we need to load-balance? I think that code is there, no?

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Tue, Jan 02, 2018 at 03:44:57PM +, Morten Rasmussen wrote: > Vincent already proposed, why can't we just modify Brendan's > CPU_NEWLY_IDLE proposal to do a stats update from idle_balance() every > 32ms regardless of whether we need to load-balance? I think that code is there, no?

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : > Hi Peter, > > On 22 December 2017 at 21:42, Peter Zijlstra wrote: > > On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: > >> Right; but I figured we'd try and do it 'right' and see how

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Le Wednesday 03 Jan 2018 à 10:16:00 (+0100), Vincent Guittot a écrit : > Hi Peter, > > On 22 December 2017 at 21:42, Peter Zijlstra wrote: > > On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: > >> Right; but I figured we'd try and do it 'right' and see how horrible it > >> is

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Hi Peter, On 22 December 2017 at 21:42, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: >> Right; but I figured we'd try and do it 'right' and see how horrible it >> is before we try and do funny things. > > So now it should have a

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Hi Peter, On 22 December 2017 at 21:42, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: >> Right; but I figured we'd try and do it 'right' and see how horrible it >> is before we try and do funny things. > > So now it should have a 32ms tick for up to .5s

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 09:42:47PM +0100, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: > > Right; but I figured we'd try and do it 'right' and see how horrible it > > is before we try and do funny things. > > So now it should have a 32ms tick for up to

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 09:42:47PM +0100, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: > > Right; but I figured we'd try and do it 'right' and see how horrible it > > is before we try and do funny things. > > So now it should have a 32ms tick for up to

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: > Right; but I figured we'd try and do it 'right' and see how horrible it > is before we try and do funny things. So now it should have a 32ms tick for up to .5s when the system goes completely idle. No idea how bad that is..

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 07:56:29PM +0100, Peter Zijlstra wrote: > Right; but I figured we'd try and do it 'right' and see how horrible it > is before we try and do funny things. So now it should have a 32ms tick for up to .5s when the system goes completely idle. No idea how bad that is..

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 03:32:53PM +0100, Vincent Guittot wrote: > > The only thing I could come up with is running a timer for this :/ That > > would keep the ILB thing running until all load is decayed (have a patch > > for that somewhere). > > IMHO running a timer doesn't sound really great I

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 03:32:53PM +0100, Vincent Guittot wrote: > > The only thing I could come up with is running a timer for this :/ That > > would keep the ILB thing running until all load is decayed (have a patch > > for that somewhere). > > IMHO running a timer doesn't sound really great I

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 December 2017 at 15:31, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 10:12:21AM +0100, Peter Zijlstra wrote: > >> The only thing I could come up with is running a timer for this :/ That >> would keep the ILB thing running until all load is decayed (have a patch >>

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 December 2017 at 15:31, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 10:12:21AM +0100, Peter Zijlstra wrote: > >> The only thing I could come up with is running a timer for this :/ That >> would keep the ILB thing running until all load is decayed (have a patch >> for that somewhere). >

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 December 2017 at 10:12, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 09:29:15AM +0100, Peter Zijlstra wrote: >> On Fri, Dec 22, 2017 at 09:05:45AM +0100, Vincent Guittot wrote: >> > On 22 December 2017 at 08:59, Peter Zijlstra wrote: >> > > On

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 December 2017 at 10:12, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 09:29:15AM +0100, Peter Zijlstra wrote: >> On Fri, Dec 22, 2017 at 09:05:45AM +0100, Vincent Guittot wrote: >> > On 22 December 2017 at 08:59, Peter Zijlstra wrote: >> > > On Thu, Dec 21, 2017 at 05:56:32PM +0100,

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 10:12:21AM +0100, Peter Zijlstra wrote: > The only thing I could come up with is running a timer for this :/ That > would keep the ILB thing running until all load is decayed (have a patch > for that somewhere). Implemented that; pushed it out, should all be at:

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 10:12:21AM +0100, Peter Zijlstra wrote: > The only thing I could come up with is running a timer for this :/ That > would keep the ILB thing running until all load is decayed (have a patch > for that somewhere). Implemented that; pushed it out, should all be at:

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 09:29:15AM +0100, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 09:05:45AM +0100, Vincent Guittot wrote: > > On 22 December 2017 at 08:59, Peter Zijlstra wrote: > > > On Thu, Dec 21, 2017 at 05:56:32PM +0100, Vincent Guittot wrote: > > >> In fact,

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 09:29:15AM +0100, Peter Zijlstra wrote: > On Fri, Dec 22, 2017 at 09:05:45AM +0100, Vincent Guittot wrote: > > On 22 December 2017 at 08:59, Peter Zijlstra wrote: > > > On Thu, Dec 21, 2017 at 05:56:32PM +0100, Vincent Guittot wrote: > > >> In fact, we can't only rely on

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 09:05:45AM +0100, Vincent Guittot wrote: > On 22 December 2017 at 08:59, Peter Zijlstra wrote: > > On Thu, Dec 21, 2017 at 05:56:32PM +0100, Vincent Guittot wrote: > >> In fact, we can't only rely on the tick and newly_idle load balance to > >> ensure

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Fri, Dec 22, 2017 at 09:05:45AM +0100, Vincent Guittot wrote: > On 22 December 2017 at 08:59, Peter Zijlstra wrote: > > On Thu, Dec 21, 2017 at 05:56:32PM +0100, Vincent Guittot wrote: > >> In fact, we can't only rely on the tick and newly_idle load balance to > >> ensure a period update of

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 December 2017 at 08:59, Peter Zijlstra wrote: > On Thu, Dec 21, 2017 at 05:56:32PM +0100, Vincent Guittot wrote: >> In fact, we can't only rely on the tick and newly_idle load balance to >> ensure a period update of the blocked load because they can never >> happen. >

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 December 2017 at 08:59, Peter Zijlstra wrote: > On Thu, Dec 21, 2017 at 05:56:32PM +0100, Vincent Guittot wrote: >> In fact, we can't only rely on the tick and newly_idle load balance to >> ensure a period update of the blocked load because they can never >> happen. > > I'm confused, why

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 December 2017 at 08:56, Peter Zijlstra wrote: > On Thu, Dec 21, 2017 at 05:23:27PM +0100, Vincent Guittot wrote: >> Hi Peter, >> >> I think that part of the proposal is missing. >> >> One goal of the patchset was to kick an update of the stats of idle >> cpu when a

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 22 December 2017 at 08:56, Peter Zijlstra wrote: > On Thu, Dec 21, 2017 at 05:23:27PM +0100, Vincent Guittot wrote: >> Hi Peter, >> >> I think that part of the proposal is missing. >> >> One goal of the patchset was to kick an update of the stats of idle >> cpu when a task wake up on a cpu but

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Thu, Dec 21, 2017 at 05:56:32PM +0100, Vincent Guittot wrote: > In fact, we can't only rely on the tick and newly_idle load balance to > ensure a period update of the blocked load because they can never > happen. I'm confused, why would the ilb not happen?

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Thu, Dec 21, 2017 at 05:56:32PM +0100, Vincent Guittot wrote: > In fact, we can't only rely on the tick and newly_idle load balance to > ensure a period update of the blocked load because they can never > happen. I'm confused, why would the ilb not happen?

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Thu, Dec 21, 2017 at 05:23:27PM +0100, Vincent Guittot wrote: > Hi Peter, > > I think that part of the proposal is missing. > > One goal of the patchset was to kick an update of the stats of idle > cpu when a task wake up on a cpu but the statistic has not been > updated for a while. > >

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On Thu, Dec 21, 2017 at 05:23:27PM +0100, Vincent Guittot wrote: > Hi Peter, > > I think that part of the proposal is missing. > > One goal of the patchset was to kick an update of the stats of idle > cpu when a task wake up on a cpu but the statistic has not been > updated for a while. > >

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 21 December 2017 at 17:23, Vincent Guittot wrote: > Hi Peter, > > I think that part of the proposal is missing. > > One goal of the patchset was to kick an update of the stats of idle > cpu when a task wake up on a cpu but the statistic has not been > updated for a

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

On 21 December 2017 at 17:23, Vincent Guittot wrote: > Hi Peter, > > I think that part of the proposal is missing. > > One goal of the patchset was to kick an update of the stats of idle > cpu when a task wake up on a cpu but the statistic has not been > updated for a while. > > That's why there

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Hi Peter, I think that part of the proposal is missing. One goal of the patchset was to kick an update of the stats of idle cpu when a task wake up on a cpu but the statistic has not been updated for a while. That's why there where a call to nohz_kick_needed in the proposal to kick ilb but only

Re: [RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Hi Peter, I think that part of the proposal is missing. One goal of the patchset was to kick an update of the stats of idle cpu when a task wake up on a cpu but the statistic has not been updated for a while. That's why there where a call to nohz_kick_needed in the proposal to kick ilb but only

[RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Suggested-by: Vincent Guittot Signed-off-by: Peter Zijlstra (Intel) --- kernel/sched/core.c |4 +-- kernel/sched/fair.c | 52 ++- kernel/sched/sched.h |4 +++ 3 files changed, 41

[RFC PATCH 2/5] sched: Add NOHZ_STATS_KICK

Suggested-by: Vincent Guittot Signed-off-by: Peter Zijlstra (Intel) --- kernel/sched/core.c |4 +-- kernel/sched/fair.c | 52 ++- kernel/sched/sched.h |4 +++ 3 files changed, 41 insertions(+), 19 deletions(-) ---