* [PATCH] sched/fair: clean up asym packing
@ 2019-06-03 14:11 Vincent Guittot
2019-06-03 14:17 ` Vincent Guittot
0 siblings, 1 reply; 5+ messages in thread
From: Vincent Guittot @ 2019-06-03 14:11 UTC (permalink / raw)
To: mingo, peterz, linux-kernel; +Cc: Vincent Guittot
Clean up asym packing to follow the default load balance behavior:
- classify the group by creating a group_asym_capacity field.
- calculate the imbalance in calculate_imbalance() instead of bypassing it.
We don't need to test twice same conditions anymore to detect asym packing
and we consolidate the calculation of imbalance in calculate_imbalance().
There is no functional changes.
Signed-off-by: Vincent Guittot <vincent.guitto@linaro.org>
---
This is a simple cleanup to gather all imbalance calculations in calculate_imbalance()
before a deeper rework of the load_balance.
kernel/sched/fair.c | 63 ++++++++++++++---------------------------------------
1 file changed, 16 insertions(+), 47 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index f35930f..93c2447 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -7888,6 +7888,7 @@ struct sg_lb_stats {
unsigned int group_weight;
enum group_type group_type;
int group_no_capacity;
+ int group_asym_capacity;
unsigned long group_misfit_task_load; /* A CPU has a task too big for its capacity */
#ifdef CONFIG_NUMA_BALANCING
unsigned int nr_numa_running;
@@ -8382,9 +8383,17 @@ static bool update_sd_pick_busiest(struct lb_env *env,
* ASYM_PACKING needs to move all the work to the highest
* prority CPUs in the group, therefore mark all groups
* of lower priority than ourself as busy.
+ *
+ * This is primarily intended to used at the sibling level. Some
+ * cores like POWER7 prefer to use lower numbered SMT threads. In the
+ * case of POWER7, it can move to lower SMT modes only when higher
+ * threads are idle. When in lower SMT modes, the threads will
+ * perform better since they share less core resources. Hence when we
+ * have idle threads, we want them to be the higher ones.
*/
if (sgs->sum_nr_running &&
sched_asym_prefer(env->dst_cpu, sg->asym_prefer_cpu)) {
+ sgs->group_asym_capacity = 1;
if (!sds->busiest)
return true;
@@ -8522,51 +8531,6 @@ static inline void update_sd_lb_stats(struct lb_env *env, struct sd_lb_stats *sd
}
/**
- * check_asym_packing - Check to see if the group is packed into the
- * sched domain.
- *
- * This is primarily intended to used at the sibling level. Some
- * cores like POWER7 prefer to use lower numbered SMT threads. In the
- * case of POWER7, it can move to lower SMT modes only when higher
- * threads are idle. When in lower SMT modes, the threads will
- * perform better since they share less core resources. Hence when we
- * have idle threads, we want them to be the higher ones.
- *
- * This packing function is run on idle threads. It checks to see if
- * the busiest CPU in this domain (core in the P7 case) has a higher
- * CPU number than the packing function is being run on. Here we are
- * assuming lower CPU number will be equivalent to lower a SMT thread
- * number.
- *
- * Return: 1 when packing is required and a task should be moved to
- * this CPU. The amount of the imbalance is returned in env->imbalance.
- *
- * @env: The load balancing environment.
- * @sds: Statistics of the sched_domain which is to be packed
- */
-static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds)
-{
- int busiest_cpu;
-
- if (!(env->sd->flags & SD_ASYM_PACKING))
- return 0;
-
- if (env->idle == CPU_NOT_IDLE)
- return 0;
-
- if (!sds->busiest)
- return 0;
-
- busiest_cpu = sds->busiest->asym_prefer_cpu;
- if (sched_asym_prefer(busiest_cpu, env->dst_cpu))
- return 0;
-
- env->imbalance = sds->busiest_stat.group_load;
-
- return 1;
-}
-
-/**
* fix_small_imbalance - Calculate the minor imbalance that exists
* amongst the groups of a sched_domain, during
* load balancing.
@@ -8650,6 +8614,11 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s
local = &sds->local_stat;
busiest = &sds->busiest_stat;
+ if (busiest->group_asym_capacity) {
+ env->imbalance = busiest->group_load;
+ return;
+ }
+
if (busiest->group_type == group_imbalanced) {
/*
* In the group_imb case we cannot rely on group-wide averages
@@ -8754,8 +8723,8 @@ static struct sched_group *find_busiest_group(struct lb_env *env)
busiest = &sds.busiest_stat;
/* ASYM feature bypasses nice load balance check */
- if (check_asym_packing(env, &sds))
- return sds.busiest;
+ if (busiest->group_asym_capacity)
+ goto force_balance;
/* There is no busy sibling group to pull tasks from */
if (!sds.busiest || busiest->sum_nr_running == 0)
--
2.7.4
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH] sched/fair: clean up asym packing
2019-06-03 14:11 [PATCH] sched/fair: clean up asym packing Vincent Guittot
@ 2019-06-03 14:17 ` Vincent Guittot
2019-06-03 18:15 ` Valentin Schneider
0 siblings, 1 reply; 5+ messages in thread
From: Vincent Guittot @ 2019-06-03 14:17 UTC (permalink / raw)
To: mingo, peterz, linux-kernel; +Cc: Vincent Guittot
Clean up asym packing to follow the default load balance behavior:
- classify the group by creating a group_asym_capacity field.
- calculate the imbalance in calculate_imbalance() instead of bypassing it.
We don't need to test twice same conditions anymore to detect asym packing
and we consolidate the calculation of imbalance in calculate_imbalance().
There is no functional changes.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
---
kernel/sched/fair.c | 63 ++++++++++++++---------------------------------------
1 file changed, 16 insertions(+), 47 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index f35930f..93c2447 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -7888,6 +7888,7 @@ struct sg_lb_stats {
unsigned int group_weight;
enum group_type group_type;
int group_no_capacity;
+ int group_asym_capacity;
unsigned long group_misfit_task_load; /* A CPU has a task too big for its capacity */
#ifdef CONFIG_NUMA_BALANCING
unsigned int nr_numa_running;
@@ -8382,9 +8383,17 @@ static bool update_sd_pick_busiest(struct lb_env *env,
* ASYM_PACKING needs to move all the work to the highest
* prority CPUs in the group, therefore mark all groups
* of lower priority than ourself as busy.
+ *
+ * This is primarily intended to used at the sibling level. Some
+ * cores like POWER7 prefer to use lower numbered SMT threads. In the
+ * case of POWER7, it can move to lower SMT modes only when higher
+ * threads are idle. When in lower SMT modes, the threads will
+ * perform better since they share less core resources. Hence when we
+ * have idle threads, we want them to be the higher ones.
*/
if (sgs->sum_nr_running &&
sched_asym_prefer(env->dst_cpu, sg->asym_prefer_cpu)) {
+ sgs->group_asym_capacity = 1;
if (!sds->busiest)
return true;
@@ -8522,51 +8531,6 @@ static inline void update_sd_lb_stats(struct lb_env *env, struct sd_lb_stats *sd
}
/**
- * check_asym_packing - Check to see if the group is packed into the
- * sched domain.
- *
- * This is primarily intended to used at the sibling level. Some
- * cores like POWER7 prefer to use lower numbered SMT threads. In the
- * case of POWER7, it can move to lower SMT modes only when higher
- * threads are idle. When in lower SMT modes, the threads will
- * perform better since they share less core resources. Hence when we
- * have idle threads, we want them to be the higher ones.
- *
- * This packing function is run on idle threads. It checks to see if
- * the busiest CPU in this domain (core in the P7 case) has a higher
- * CPU number than the packing function is being run on. Here we are
- * assuming lower CPU number will be equivalent to lower a SMT thread
- * number.
- *
- * Return: 1 when packing is required and a task should be moved to
- * this CPU. The amount of the imbalance is returned in env->imbalance.
- *
- * @env: The load balancing environment.
- * @sds: Statistics of the sched_domain which is to be packed
- */
-static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds)
-{
- int busiest_cpu;
-
- if (!(env->sd->flags & SD_ASYM_PACKING))
- return 0;
-
- if (env->idle == CPU_NOT_IDLE)
- return 0;
-
- if (!sds->busiest)
- return 0;
-
- busiest_cpu = sds->busiest->asym_prefer_cpu;
- if (sched_asym_prefer(busiest_cpu, env->dst_cpu))
- return 0;
-
- env->imbalance = sds->busiest_stat.group_load;
-
- return 1;
-}
-
-/**
* fix_small_imbalance - Calculate the minor imbalance that exists
* amongst the groups of a sched_domain, during
* load balancing.
@@ -8650,6 +8614,11 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s
local = &sds->local_stat;
busiest = &sds->busiest_stat;
+ if (busiest->group_asym_capacity) {
+ env->imbalance = busiest->group_load;
+ return;
+ }
+
if (busiest->group_type == group_imbalanced) {
/*
* In the group_imb case we cannot rely on group-wide averages
@@ -8754,8 +8723,8 @@ static struct sched_group *find_busiest_group(struct lb_env *env)
busiest = &sds.busiest_stat;
/* ASYM feature bypasses nice load balance check */
- if (check_asym_packing(env, &sds))
- return sds.busiest;
+ if (busiest->group_asym_capacity)
+ goto force_balance;
/* There is no busy sibling group to pull tasks from */
if (!sds.busiest || busiest->sum_nr_running == 0)
--
2.7.4
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] sched/fair: clean up asym packing
2019-06-03 14:17 ` Vincent Guittot
@ 2019-06-03 18:15 ` Valentin Schneider
2019-06-03 18:32 ` Vincent Guittot
0 siblings, 1 reply; 5+ messages in thread
From: Valentin Schneider @ 2019-06-03 18:15 UTC (permalink / raw)
To: Vincent Guittot, mingo, peterz, linux-kernel
Hi,
On 03/06/2019 15:17, Vincent Guittot wrote:
> Clean up asym packing to follow the default load balance behavior:
> - classify the group by creating a group_asym_capacity field.
Being nitpicky here, this doesn't classify the group in the usual way
- it doesn't get a specific group_type value (group_classify()). So maybe
"classify" isn't the best term here.
Also, why tag this group in update_sd_pick_busiest()? It would make more
sense to do so in update_sg_lb_stats() like with the other sg_lb_stats fields:
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 93c24473c8a0..537710026c3a 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -8298,6 +8298,10 @@ static inline void update_sg_lb_stats(struct lb_env *env,
}
}
+ if (sgs->sum_nr_running &&
+ sched_asym_prefer(env->dst_cpu, group->asym_prefer_cpu))
+ sgs->group_asym_capacity = 1;
+
/* Adjust by relative CPU capacity of the group */
sgs->group_capacity = group->sgc->capacity;
sgs->avg_load = (sgs->group_load*SCHED_CAPACITY_SCALE) / sgs->group_capacity;
@@ -8391,9 +8395,7 @@ static bool update_sd_pick_busiest(struct lb_env *env,
* perform better since they share less core resources. Hence when we
* have idle threads, we want them to be the higher ones.
*/
- if (sgs->sum_nr_running &&
- sched_asym_prefer(env->dst_cpu, sg->asym_prefer_cpu)) {
- sgs->group_asym_capacity = 1;
+ if (sgs->group_asym_capacity) {
if (!sds->busiest)
return true;
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] sched/fair: clean up asym packing
2019-06-03 18:15 ` Valentin Schneider
@ 2019-06-03 18:32 ` Vincent Guittot
2019-06-04 17:24 ` Valentin Schneider
0 siblings, 1 reply; 5+ messages in thread
From: Vincent Guittot @ 2019-06-03 18:32 UTC (permalink / raw)
To: Valentin Schneider; +Cc: Ingo Molnar, Peter Zijlstra, linux-kernel
On Mon, 3 Jun 2019 at 20:15, Valentin Schneider
<valentin.schneider@arm.com> wrote:
>
> Hi,
>
> On 03/06/2019 15:17, Vincent Guittot wrote:
> > Clean up asym packing to follow the default load balance behavior:
> > - classify the group by creating a group_asym_capacity field.
>
> Being nitpicky here, this doesn't classify the group in the usual way
> - it doesn't get a specific group_type value (group_classify()). So maybe
> "classify" isn't the best term here.
My original goal was to add a group type to classify the group but
this would have broken the current behavior whereas I only want to
move code
>
> Also, why tag this group in update_sd_pick_busiest()? It would make more
> sense to do so in update_sg_lb_stats() like with the other sg_lb_stats fields:
With your proposal below, the test is called for every groups'
statistic update whereas it is only done lastly after checking other
rules in the current code and I don't want to modify the current
behavior but only move code to set imbalance in calculate imbalance.
A bigger cleanup will come in next steps
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 93c24473c8a0..537710026c3a 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -8298,6 +8298,10 @@ static inline void update_sg_lb_stats(struct lb_env *env,
> }
> }
>
> + if (sgs->sum_nr_running &&
> + sched_asym_prefer(env->dst_cpu, group->asym_prefer_cpu))
> + sgs->group_asym_capacity = 1;
> +
> /* Adjust by relative CPU capacity of the group */
> sgs->group_capacity = group->sgc->capacity;
> sgs->avg_load = (sgs->group_load*SCHED_CAPACITY_SCALE) / sgs->group_capacity;
> @@ -8391,9 +8395,7 @@ static bool update_sd_pick_busiest(struct lb_env *env,
> * perform better since they share less core resources. Hence when we
> * have idle threads, we want them to be the higher ones.
> */
> - if (sgs->sum_nr_running &&
> - sched_asym_prefer(env->dst_cpu, sg->asym_prefer_cpu)) {
> - sgs->group_asym_capacity = 1;
> + if (sgs->group_asym_capacity) {
> if (!sds->busiest)
> return true;
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] sched/fair: clean up asym packing
2019-06-03 18:32 ` Vincent Guittot
@ 2019-06-04 17:24 ` Valentin Schneider
0 siblings, 0 replies; 5+ messages in thread
From: Valentin Schneider @ 2019-06-04 17:24 UTC (permalink / raw)
To: Vincent Guittot; +Cc: Ingo Molnar, Peter Zijlstra, linux-kernel
On 03/06/2019 19:32, Vincent Guittot wrote:
> On Mon, 3 Jun 2019 at 20:15, Valentin Schneider
[...]
> My original goal was to add a group type to classify the group but
> this would have broken the current behavior whereas I only want to
> move code
>
>>
>> Also, why tag this group in update_sd_pick_busiest()? It would make more
>> sense to do so in update_sg_lb_stats() like with the other sg_lb_stats fields:
>
> With your proposal below, the test is called for every groups'
> statistic update whereas it is only done lastly after checking other
> rules in the current code and I don't want to modify the current
> behavior but only move code to set imbalance in calculate imbalance.
>
Adding a new group_type would make sense. From a behavioral point of view
your change is fine, but from a logical one it sits halfway between being
a new stat and being a new group_type. I'd rather see a new group_type,
though as you said that's a different topic than cleaning up duplicate
operations.
> A bigger cleanup will come in next steps
>
[...]
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2019-06-04 17:24 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-06-03 14:11 [PATCH] sched/fair: clean up asym packing Vincent Guittot
2019-06-03 14:17 ` Vincent Guittot
2019-06-03 18:15 ` Valentin Schneider
2019-06-03 18:32 ` Vincent Guittot
2019-06-04 17:24 ` Valentin Schneider
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.