When the SCHED_DEADLINE scheduling class increases the CPU utilization,
we should not wait for the rate limit, otherwise we may miss some
deadline.
Tests using rt-app on Exynos5422 with up to 10 SCHED_DEADLINE tasks have
shown reductions of even 10% of deadline misses with a negligible
increase of energy consumption (measured through Baylibre Cape).
Signed-off-by: Claudio Scordino <[email protected]>
Acked-by: Viresh Kumar <[email protected]>
Reviewed-by: Rafael J. Wysocki <[email protected]>
CC: Rafael J. Wysocki <[email protected]>
CC: Viresh Kumar <[email protected]>
CC: Patrick Bellasi <[email protected]>
CC: Dietmar Eggemann <[email protected]>
CC: Morten Rasmussen <[email protected]>
CC: Juri Lelli <[email protected]>
CC: Vincent Guittot <[email protected]>
CC: Todd Kjos <[email protected]>
CC: Joel Fernandes <[email protected]>
CC: [email protected]
CC: [email protected]
---
Changes from v3:
- Specific routine renamed as ignore_dl_rate_limit()
---
Changes from v2:
- Rate limit ignored also in case of "fast switch"
- Specific routine added
---
Changes from v1:
- Logic moved from sugov_should_update_freq() to
sugov_update_single()/_shared() to not duplicate data structures
- Rate limit not ignored in case of "fast switch"
---
kernel/sched/cpufreq_schedutil.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)
diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
index feb5f89..2aeb1ca 100644
--- a/kernel/sched/cpufreq_schedutil.c
+++ b/kernel/sched/cpufreq_schedutil.c
@@ -257,6 +257,16 @@ static bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu)
static inline bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu) { return false; }
#endif /* CONFIG_NO_HZ_COMMON */
+/*
+ * Make sugov_should_update_freq() ignore the rate limit when DL
+ * has increased the utilization.
+ */
+static inline void ignore_dl_rate_limit(struct sugov_cpu *sg_cpu, struct sugov_policy *sg_policy)
+{
+ if (cpu_util_dl(cpu_rq(sg_cpu->cpu)) > sg_cpu->util_dl)
+ sg_policy->need_freq_update = true;
+}
+
static void sugov_update_single(struct update_util_data *hook, u64 time,
unsigned int flags)
{
@@ -270,6 +280,8 @@ static void sugov_update_single(struct update_util_data *hook, u64 time,
sugov_set_iowait_boost(sg_cpu, time);
sg_cpu->last_update = time;
+ ignore_dl_rate_limit(sg_cpu, sg_policy);
+
if (!sugov_should_update_freq(sg_policy, time))
return;
@@ -351,6 +363,8 @@ sugov_update_shared(struct update_util_data *hook, u64 time, unsigned int flags)
raw_spin_lock(&sg_policy->update_lock);
+ ignore_dl_rate_limit(sg_cpu, sg_policy);
+
sugov_get_util(sg_cpu);
sg_cpu->flags = flags;
--
2.7.4
On 13-03-18, 11:35, Claudio Scordino wrote:
> When the SCHED_DEADLINE scheduling class increases the CPU utilization,
> we should not wait for the rate limit, otherwise we may miss some
> deadline.
>
> Tests using rt-app on Exynos5422 with up to 10 SCHED_DEADLINE tasks have
> shown reductions of even 10% of deadline misses with a negligible
> increase of energy consumption (measured through Baylibre Cape).
>
> Signed-off-by: Claudio Scordino <[email protected]>
> Acked-by: Viresh Kumar <[email protected]>
> Reviewed-by: Rafael J. Wysocki <[email protected]>
> CC: Rafael J. Wysocki <[email protected]>
> CC: Viresh Kumar <[email protected]>
> CC: Patrick Bellasi <[email protected]>
> CC: Dietmar Eggemann <[email protected]>
> CC: Morten Rasmussen <[email protected]>
> CC: Juri Lelli <[email protected]>
> CC: Vincent Guittot <[email protected]>
> CC: Todd Kjos <[email protected]>
> CC: Joel Fernandes <[email protected]>
> CC: [email protected]
> CC: [email protected]
> ---
> Changes from v3:
> - Specific routine renamed as ignore_dl_rate_limit()
LGTM. Thanks.
--
viresh
On Tue, Mar 13, 2018 at 4:15 AM, Viresh Kumar <[email protected]> wrote:
> On 13-03-18, 11:35, Claudio Scordino wrote:
>> When the SCHED_DEADLINE scheduling class increases the CPU utilization,
>> we should not wait for the rate limit, otherwise we may miss some
>> deadline.
>>
>> Tests using rt-app on Exynos5422 with up to 10 SCHED_DEADLINE tasks have
>> shown reductions of even 10% of deadline misses with a negligible
>> increase of energy consumption (measured through Baylibre Cape).
>>
>> Signed-off-by: Claudio Scordino <[email protected]>
>> Acked-by: Viresh Kumar <[email protected]>
>> Reviewed-by: Rafael J. Wysocki <[email protected]>
>> CC: Rafael J. Wysocki <[email protected]>
>> CC: Viresh Kumar <[email protected]>
>> CC: Patrick Bellasi <[email protected]>
>> CC: Dietmar Eggemann <[email protected]>
>> CC: Morten Rasmussen <[email protected]>
>> CC: Juri Lelli <[email protected]>
>> CC: Vincent Guittot <[email protected]>
>> CC: Todd Kjos <[email protected]>
>> CC: Joel Fernandes <[email protected]>
>> CC: [email protected]
>> CC: [email protected]
>> ---
>> Changes from v3:
>> - Specific routine renamed as ignore_dl_rate_limit()
>
> LGTM. Thanks.
Nice! Thanks.
- Joel
On Wednesday, March 14, 2018 2:27:53 AM CET Joel Fernandes wrote:
> On Tue, Mar 13, 2018 at 4:15 AM, Viresh Kumar <[email protected]> wrote:
> > On 13-03-18, 11:35, Claudio Scordino wrote:
> >> When the SCHED_DEADLINE scheduling class increases the CPU utilization,
> >> we should not wait for the rate limit, otherwise we may miss some
> >> deadline.
> >>
> >> Tests using rt-app on Exynos5422 with up to 10 SCHED_DEADLINE tasks have
> >> shown reductions of even 10% of deadline misses with a negligible
> >> increase of energy consumption (measured through Baylibre Cape).
> >>
> >> Signed-off-by: Claudio Scordino <[email protected]>
> >> Acked-by: Viresh Kumar <[email protected]>
> >> Reviewed-by: Rafael J. Wysocki <[email protected]>
> >> CC: Rafael J. Wysocki <[email protected]>
> >> CC: Viresh Kumar <[email protected]>
> >> CC: Patrick Bellasi <[email protected]>
> >> CC: Dietmar Eggemann <[email protected]>
> >> CC: Morten Rasmussen <[email protected]>
> >> CC: Juri Lelli <[email protected]>
> >> CC: Vincent Guittot <[email protected]>
> >> CC: Todd Kjos <[email protected]>
> >> CC: Joel Fernandes <[email protected]>
> >> CC: [email protected]
> >> CC: [email protected]
> >> ---
> >> Changes from v3:
> >> - Specific routine renamed as ignore_dl_rate_limit()
> >
> > LGTM. Thanks.
>
> Nice! Thanks.
OK, the patch doesn't seem to depend on anything in -tip, so I'm going to
apply it.
Thanks!
Commit-ID: e97a90f7069b740575bcb1dae86596e0484b8957
Gitweb: https://git.kernel.org/tip/e97a90f7069b740575bcb1dae86596e0484b8957
Author: Claudio Scordino <[email protected]>
AuthorDate: Tue, 13 Mar 2018 11:35:40 +0100
Committer: Thomas Gleixner <[email protected]>
CommitDate: Fri, 23 Mar 2018 22:48:22 +0100
sched/cpufreq: Rate limits for SCHED_DEADLINE
When the SCHED_DEADLINE scheduling class increases the CPU utilization, it
should not wait for the rate limit, otherwise it may miss some deadline.
Tests using rt-app on Exynos5422 with up to 10 SCHED_DEADLINE tasks have
shown reductions of even 10% of deadline misses with a negligible
increase of energy consumption (measured through Baylibre Cape).
Signed-off-by: Claudio Scordino <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Rafael J. Wysocki <[email protected]>
Acked-by: Viresh Kumar <[email protected]>
Cc: Juri Lelli <[email protected]>
Cc: Joel Fernandes <[email protected]>
Cc: Vincent Guittot <[email protected]>
Cc: [email protected]
Cc: Peter Zijlstra <[email protected]>
Cc: Morten Rasmussen <[email protected]>
Cc: Patrick Bellasi <[email protected]>
Cc: Todd Kjos <[email protected]>
Cc: Dietmar Eggemann <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
---
kernel/sched/cpufreq_schedutil.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)
diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
index 89fe78ecb88c..2b124811947d 100644
--- a/kernel/sched/cpufreq_schedutil.c
+++ b/kernel/sched/cpufreq_schedutil.c
@@ -267,6 +267,16 @@ static bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu)
static inline bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu) { return false; }
#endif /* CONFIG_NO_HZ_COMMON */
+/*
+ * Make sugov_should_update_freq() ignore the rate limit when DL
+ * has increased the utilization.
+ */
+static inline void ignore_dl_rate_limit(struct sugov_cpu *sg_cpu, struct sugov_policy *sg_policy)
+{
+ if (cpu_util_dl(cpu_rq(sg_cpu->cpu)) > sg_cpu->util_dl)
+ sg_policy->need_freq_update = true;
+}
+
static void sugov_update_single(struct update_util_data *hook, u64 time,
unsigned int flags)
{
@@ -279,6 +289,8 @@ static void sugov_update_single(struct update_util_data *hook, u64 time,
sugov_set_iowait_boost(sg_cpu, time, flags);
sg_cpu->last_update = time;
+ ignore_dl_rate_limit(sg_cpu, sg_policy);
+
if (!sugov_should_update_freq(sg_policy, time))
return;
@@ -356,6 +368,8 @@ sugov_update_shared(struct update_util_data *hook, u64 time, unsigned int flags)
sugov_set_iowait_boost(sg_cpu, time, flags);
sg_cpu->last_update = time;
+ ignore_dl_rate_limit(sg_cpu, sg_policy);
+
if (sugov_should_update_freq(sg_policy, time)) {
next_f = sugov_next_freq_shared(sg_cpu, time);
sugov_update_commit(sg_policy, time, next_f);