Modern interactive systems, such as recent Android phones, tend to have
power efficient shallow idle states. Selecting deeper idle states on a
device while a latency-sensitive workload is running can adversely
impact performance due to increased latency. Additionally, if the CPU
wakes up from a deeper sleep before its target residency as is often the
case, it results in a waste of energy on top of that.
At the moment, none of the available idle governors take any scheduling
information into account. They also tend to overestimate the idle
duration quite often, which causes them to select excessively deep idle
states, thus leading to increased wakeup latency and lower performance
with no power saving. For 'menu' while web browsing on Android for
instance, those types of wakeups ('too deep') account for over 24% of
all wakeups.
At the same time, on some platforms idle state 0 can be power efficient
enough to warrant wanting to prefer it over idle state 1. This is
because the power usage of the two states can be so close that
sufficient amounts of too deep state 1 sleeps can completely offset the
state 1 power saving to the point where it would've been more power
efficient to just use state 0 instead. This is, of course, for systems
where state 0 is not a polling state, such as arm-based devices.
Sleeps that happened in state 0 while they could have used state 1 ('too
shallow') only save less power than they otherwise could have. Too deep
sleeps, on the other hand, harm performance and nullify the potential
power saving from using state 1 in the first place. While taking this
into account, it is clear that on balance it is preferable for an idle
governor to have more too shallow sleeps instead of more too deep sleeps
on those kinds of platforms.
This patch specifically tunes TEO to prefer shallower idle states in
order to reduce wakeup latency and achieve better performance.
To this end, before selecting the next idle state it uses the avg_util
signal of a CPU's runqueue in order to determine to what extent the CPU
is being utilized. This util value is then compared to a threshold
defined as a percentage of the CPU's capacity (capacity >> 6 ie. ~1.5%
in the current implementation). If the util is above the threshold, the
index of the idle state selected by TEO metrics will be reduced by 1,
thus selecting a shallower state. If the util is below the threshold,
the governor defaults to the TEO metrics mechanism to try to select the
deepest available idle state based on the closest timer event and its
own correctness.
The main goal of this is to reduce latency and increase performance for
some workloads. Under some workloads it will result in an increase in
power usage (Geekbench 5) while for other workloads it will also result
in a decrease in power usage compared to TEO (PCMark Web, Jankbench,
Speedometer).
It can provide drastically decreased latency and performance benefits in
certain types of workloads that are sensitive to latency.
Example test results:
1. GB5 (better score, latency & more power usage)
| metric | menu | teo | teo-util-aware |
| ------------------------------------- | -------------- | ----------------- | ----------------- |
| gmean score | 2826.5 (0.0%) | 2764.8 (-2.18%) | 2865 (1.36%) |
| gmean power usage [mW] | 2551.4 (0.0%) | 2606.8 (2.17%) | 2722.3 (6.7%) |
| gmean too deep % | 14.99% | 9.65% | 4.02% |
| gmean too shallow % | 2.5% | 5.96% | 14.59% |
| gmean task wakeup latency (asynctask) | 78.16μs (0.0%) | 61.60μs (-21.19%) | 54.45μs (-30.34%) |
2. Jankbench (better score, latency & less power usage)
| metric | menu | teo | teo-util-aware |
| ------------------------------------- | -------------- | ----------------- | ----------------- |
| gmean frame duration | 13.9 (0.0%) | 14.7 (6.0%) | 12.6 (-9.0%) |
| gmean jank percentage | 1.5 (0.0%) | 2.1 (36.99%) | 1.3 (-17.37%) |
| gmean power usage [mW] | 144.6 (0.0%) | 136.9 (-5.27%) | 121.3 (-16.08%) |
| gmean too deep % | 26.00% | 11.00% | 2.54% |
| gmean too shallow % | 4.74% | 11.89% | 21.93% |
| gmean wakeup latency (RenderThread) | 139.5μs (0.0%) | 116.5μs (-16.49%) | 91.11μs (-34.7%) |
| gmean wakeup latency (surfaceflinger) | 124.0μs (0.0%) | 151.9μs (22.47%) | 87.65μs (-29.33%) |
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
[ rjw: Comment edits and white space adjustments ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Add a no_poll flag to teo_find_shallower_state() that will let the
function optionally not consider polling states.
This allows the caller to guard against the function inadvertently
resulting in TEO putting the CPU in a polling state when that
behaviour is undesirable.
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Add trace points as are implemented in KVM host halt polling.
This helps tune guest halt polling params.
Signed-off-by: Eiichi Tsukata <eiichi.tsukata@nutanix.com>
Acked-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The word `these' in a comment is repeated, so drop one.
Signed-off-by: Jason Wang <wangborong@cdjrlc.com>
[ rjw: Changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Rename two local variables in teo_select() so that their names better
reflect their purpose.
No functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
There are three mistakes in the loop in teo_select() that is looking
for an alternative candidate idle state. First, it should walk all
of the idle states shallower than the current candidate one,
including all of the disabled ones, but it terminates after the first
enabled idle state. Second, it should not terminate its last step
if idle state 0 is disabled (which is related to the first issue).
Finally, it may return the current alternative candidate idle state
prematurely if the time span criterion is not met by the idle state
under consideration at the moment.
To address the issues mentioned above, make the loop in question walk
all of the idle states shallower than the current candidate idle state
all the way down to idle state 0 and rearrange the checks in it.
Fixes: 77577558f2 ("cpuidle: teo: Rework most recent idle duration values treatment")
Reported-by: Doug Smythies <dsmythies@telus.net>
Tested-by: Doug Smythies <dsmythies@telus.net>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
- Make intel_pstate support hybrid processors using abstract
performance units in the HWP interface (Rafael Wysocki).
- Add Icelake servers and Cometlake support in no-HWP mode to
intel_pstate (Giovanni Gherdovich).
- Make cpufreq_online() error path be consistent with the CPU
device removal path in cpufreq (Rafael Wysocki).
- Clean up 3 cpufreq drivers and the statistics code (Hailong Liu,
Randy Dunlap, Shaokun Zhang).
- Make intel_idle use special idle state parameters for C6 when
package C-states are disabled (Chen Yu).
- Rework the TEO (timer events oriented) cpuidle governor to address
some theoretical shortcomings in it (Rafael Wysocki).
- Drop unneeded semicolon from the TEO governor (Wan Jiabing).
- Modify the runtime PM framework to accept unassigned suspend
and resume callback pointers (Ulf Hansson).
- Improve pm_runtime_get_sync() documentation (Krzysztof Kozlowski).
- Improve device performance states support in the generic power
domains (genpd) framework (Ulf Hansson).
- Fix some documentation issues in genpd (Yang Yingliang).
- Make the operating performance points (OPP) framework use the
required-opps DT property in use cases that are not related to
genpd (Hsin-Yi Wang).
- Make lazy_link_required_opp_table() use list_del_init instead of
list_del/INIT_LIST_HEAD (Yang Yingliang).
- Simplify wake IRQs handling in the core system-wide sleep support
code and clean up some coding style inconsistencies in it (Tian
Tao, Zhen Lei).
- Add cooling support to the tegra30 devfreq driver and improve its
DT bindings (Dmitry Osipenko).
- Fix some assorted issues in the devfreq core and drivers (Chanwoo
Choi, Dong Aisheng, YueHaibing).
-----BEGIN PGP SIGNATURE-----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=mgLK
-----END PGP SIGNATURE-----
Merge tag 'pm-5.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management updates from Rafael Wysocki:
"These add hybrid processors support to the intel_pstate driver and
make it work with more processor models when HWP is disabled, make the
intel_idle driver use special C6 idle state paremeters when package
C-states are disabled, add cooling support to the tegra30 devfreq
driver, rework the TEO (timer events oriented) cpuidle governor,
extend the OPP (operating performance points) framework to use the
required-opps DT property in more cases, fix some issues and clean up
a number of assorted pieces of code.
Specifics:
- Make intel_pstate support hybrid processors using abstract
performance units in the HWP interface (Rafael Wysocki).
- Add Icelake servers and Cometlake support in no-HWP mode to
intel_pstate (Giovanni Gherdovich).
- Make cpufreq_online() error path be consistent with the CPU device
removal path in cpufreq (Rafael Wysocki).
- Clean up 3 cpufreq drivers and the statistics code (Hailong Liu,
Randy Dunlap, Shaokun Zhang).
- Make intel_idle use special idle state parameters for C6 when
package C-states are disabled (Chen Yu).
- Rework the TEO (timer events oriented) cpuidle governor to address
some theoretical shortcomings in it (Rafael Wysocki).
- Drop unneeded semicolon from the TEO governor (Wan Jiabing).
- Modify the runtime PM framework to accept unassigned suspend and
resume callback pointers (Ulf Hansson).
- Improve pm_runtime_get_sync() documentation (Krzysztof Kozlowski).
- Improve device performance states support in the generic power
domains (genpd) framework (Ulf Hansson).
- Fix some documentation issues in genpd (Yang Yingliang).
- Make the operating performance points (OPP) framework use the
required-opps DT property in use cases that are not related to
genpd (Hsin-Yi Wang).
- Make lazy_link_required_opp_table() use list_del_init instead of
list_del/INIT_LIST_HEAD (Yang Yingliang).
- Simplify wake IRQs handling in the core system-wide sleep support
code and clean up some coding style inconsistencies in it (Tian
Tao, Zhen Lei).
- Add cooling support to the tegra30 devfreq driver and improve its
DT bindings (Dmitry Osipenko).
- Fix some assorted issues in the devfreq core and drivers (Chanwoo
Choi, Dong Aisheng, YueHaibing)"
* tag 'pm-5.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (39 commits)
PM / devfreq: passive: Fix get_target_freq when not using required-opp
cpufreq: Make cpufreq_online() call driver->offline() on errors
opp: Allow required-opps to be used for non genpd use cases
cpuidle: teo: remove unneeded semicolon in teo_select()
dt-bindings: devfreq: tegra30-actmon: Add cooling-cells
dt-bindings: devfreq: tegra30-actmon: Convert to schema
PM / devfreq: userspace: Use DEVICE_ATTR_RW macro
PM: runtime: Clarify documentation when callbacks are unassigned
PM: runtime: Allow unassigned ->runtime_suspend|resume callbacks
PM: runtime: Improve path in rpm_idle() when no callback
PM: hibernate: remove leading spaces before tabs
PM: sleep: remove trailing spaces and tabs
PM: domains: Drop/restore performance state votes for devices at runtime PM
PM: domains: Return early if perf state is already set for the device
PM: domains: Split code in dev_pm_genpd_set_performance_state()
cpuidle: teo: Use kerneldoc documentation in admin-guide
cpuidle: teo: Rework most recent idle duration values treatment
cpuidle: teo: Change the main idle state selection logic
cpuidle: teo: Cosmetic modification of teo_select()
cpuidle: teo: Cosmetic modifications of teo_update()
...
Fix following coccicheck warning:
drivers/cpuidle/governors/teo.c:315:10-11: Unneeded semicolon
Signed-off-by: Wan Jiabing <wanjiabing@vivo.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
There are two descriptions of the TEO (Timer Events Oriented) cpuidle
governor in the kernel source tree, one in the C file containing its
code and one in cpuidle.rst which is part of admin-guide.
Instead of trying to keep them both in sync and in order to reduce
text duplication, include the governor description from the C file
directly into cpuidle.rst.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The TEO (Timer Events Oriented) cpuidle governor uses several most
recent idle duration values for a given CPU to refine the idle state
selection in case the previous long-term trends have not been
followed recently and a new trend appears to be forming. That is
done by computing the average of the most recent idle duration
values falling below the time till the next timer event ("sleep
length"), provided that they are the majority of the most recent
idle duration values taken into account, and using it as the new
expected idle duration value.
However, idle state selection based on that value may not be optimal,
because the average does not really indicate which of the idle states
with target residencies less than or equal to it is likely to be the
best fit.
Thus, instead of computing the average, make the governor carry out
computations based on the distribution of the most recent idle
duration values among the bins corresponding to different idle
states. Namely, if the majority of the most recent idle duration
values taken into consideration are less than the current sleep
length (which means that the CPU is likely to wake up early), find
the idle state closest to the "candidate" one "matching" the sleep
length whose target residency is less than or equal to the majority
of the most recent idle duration values that have fallen below the
current sleep length (which means that it is likely to be "shallow
enough" this time).
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Two aspects of the current main idle state selection logic in the
TEO (Timer Events Oriented) cpuidle governor are quite questionable.
First of all, the "hits" and "misses" metrics used by it are only
updated for a given idle state if the time till the next timer event
("sleep length") is between the target residency of that state and
the target residency of the next one. Consequently, they are likely
to become stale if the sleep length tends to fall outside that
interval which increases the likelihood of subomtimal idle state
selection.
Second, the decision on whether or not to select the idle state
"matching" the sleep length is based on the metrics collected for
that state alone, whereas in principle the metrics collected for
the other idle states should be taken into consideration when that
decision is made. For example, if the measured idle duration is less
than the target residency of the idle state "matching" the sleep
length, then it is also less than the target residency of any deeper
idle state and that should be taken into account when considering
whether or not to select any of those states, but currently it is
not.
In order to address the above shortcomings, modify the main idle
state selection logic in the TEO governor to take the metrics
collected for all of the idle states into account when deciding
whether or not to select the one "matching" the sleep length.
Moreover, drop the "misses" metric that becomes redundant after the
above change and rename the "early_hits" metric to "intercepts" so
that its role is better reflected by its name (the idea being that
if a CPU wakes up earlier than indicated by the sleep length, then
it must be a result of a non-timer interrupt that "intercepts" the
CPU).
Also rename the states[] array in struct struct teo_cpu to
state_bins[] to avoid confusing it with the states[] array in
struct cpuidle_driver and update the documentation to match the
new code (and make it more comprehensive while at it).
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Initialize local variables in teo_select() where they are declared.
No functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Rename a local variable in teo_update() so that its purpose is better
reflected by its name and use one more local variable in the loop
over the CPU idle states in that function to make the code somewhat
easier to read.
No functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Make the menu governor check the tick_nohz_get_next_hrtimer()
return value so as to avoid dealing with negative "sleep length"
values and make it use that value directly when the tick is stopped.
While at it, rename local variable delta_next in menu_select() to
delta_tick which better reflects its purpose.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Modify the TEO governor to take possible negative return values of
tick_nohz_get_next_hrtimer() into account by changing the data type
of some variables used by it to s64 which allows it to carry out
computations without potentially problematic data type conversions
into u64.
Also change the computations in teo_select() so that the negative
values themselves are handled in a natural way to avoid adding extra
negative value checks to that function.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
If the time till the next timer event is shorter than the target
residency of the first idle state (state 0), the TEO governor does
not update its metrics for any idle states, but arguably it should
record a "hit" for idle state 0 in that case, so modify it to do
that.
Accordingly, also make it record an "early hit" for idle state 0 if
the measured idle duration is less than its target residency, which
allows one branch more to be dropped from teo_update().
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Fix a simple bug in rotating array index.
Fixes: b26bf6ab71 ("cpuidle: New timer events oriented governor for tickless systems")
Signed-off-by: Ikjoon Jang <ikjn@chromium.org>
Cc: 5.1+ <stable@vger.kernel.org> # 5.1+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
There are three places in teo_select() where a given amount of time
is compared with TICK_NSEC if tick_nohz_tick_stopped() returns true,
which is a bit of duplicated code.
Avoid that code duplication by defining a helper function to do the
check and using it in all of the places in question.
No intentional functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
If the current state with the maximum "early hits" metric in
teo_select() is also the one "matching" the expected idle duration,
it will be used as the candidate one for selection even if its
"misses" metric is greater than its "hits" metric, which is not
correct.
In that case, the candidate state should be shallower than the
current one and its "early hits" metric should be the maximum
among the idle states shallower than the current one.
To make that happen, modify teo_select() to save the index of
the state whose "early hits" metric is the maximum for the
range of states below the current one and go back to that state
if it turns out that the current one should be rejected.
Fixes: 159e48560f ("cpuidle: teo: Fix "early hits" handling for disabled idle states")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
One purpose of the computations in teo_update() is to determine
whether or not the (saved) time till the next timer event and the
measured idle duration fall into the same "bin", so avoid using
values that include the cpuidle overhead to obtain the latter.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Currently, the cpuidle subsystem uses microseconds as the unit of
time which (among other things) causes the idle loop to incur some
integer division overhead for no clear benefit.
In order to allow cpuidle to measure time in nanoseconds, add two
new fields, exit_latency_ns and target_residency_ns, to represent the
exit latency and target residency of an idle state in nanoseconds,
respectively, to struct cpuidle_state and initialize them with the
help of the corresponding values in microseconds provided by drivers.
Additionally, change cpuidle_governor_latency_req() to return the
idle state exit latency constraint in nanoseconds.
Also meeasure idle state residency (last_residency_ns in struct
cpuidle_device and time_ns in struct cpuidle_driver) in nanoseconds
and update the cpuidle core and governors accordingly.
However, the menu governor still computes typical intervals in
microseconds to avoid integer overflows.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Doug Smythies <dsmythies@telus.net>
Tested-by: Doug Smythies <dsmythies@telus.net>
There are two reasons why CPU idle states may be disabled: either
because the driver has disabled them or because they have been
disabled by user space via sysfs.
In the former case, the state's "disabled" flag is set once during
the initialization of the driver and it is never cleared later (it
is read-only effectively). In the latter case, the "disable" field
of the given state's cpuidle_state_usage struct is set and it may be
changed via sysfs. Thus checking whether or not an idle state has
been disabled involves reading these two flags every time.
In order to avoid the additional check of the state's "disabled" flag
(which is effectively read-only anyway), use the value of it at the
init time to set a (new) flag in the "disable" field of that state's
cpuidle_state_usage structure and use the sysfs interface to
manipulate another (new) flag in it. This way the state is disabled
whenever the "disable" field of its cpuidle_state_usage structure is
nonzero, whatever the reason, and it is the only place to look into
to check whether or not the state has been disabled.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
The TEO governor uses idle duration "bins" defined in accordance with
the CPU idle states table provided by the driver, so that each "bin"
covers the idle duration range between the target residency of the
idle state corresponding to it and the target residency of the closest
deeper idle state. The governor collects statistics for each bin
regardless of whether or not the idle state corresponding to it is
currently enabled.
In particular, the "early hits" metric measures the likelihood of a
situation in which the idle duration measured after wakeup falls into
to given bin, but the time till the next timer (sleep length) falls
into a bin corresponding to one of the deeper idle states. It is
used when the "hits" and "misses" metrics indicate that the state
"matching" the sleep length should not be selected, so that the state
with the maximum "early hits" value is selected instead of it.
If the idle state corresponding to the given bin is disabled, it
cannot be selected and if it turns out to be the one that should be
selected, a shallower idle state needs to be used instead of it.
Nevertheless, the metrics collected for the bin corresponding to it
are still valid and need to be taken into account as though that
state had not been disabled.
As far as the "early hits" metric is concerned, teo_select() tries to
take disabled states into account, but the state index corresponding
to the maximum "early hits" value computed by it may be incorrect.
Namely, it always uses the index of the previous maximum "early hits"
state then, but there may be enabled idle states closer to the
disabled one in question. In particular, if the current candidate
state (whose index is the idx value) is closer to the disabled one
and the "early hits" value of the disabled state is greater than the
current maximum, the index of the current candidate state (idx)
should replace the "maximum early hits state" index.
Modify the code to handle that case correctly.
Fixes: b26bf6ab71 ("cpuidle: New timer events oriented governor for tickless systems")
Reported-by: Doug Smythies <dsmythies@telus.net>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: 5.1+ <stable@vger.kernel.org> # 5.1+
The TEO governor uses idle duration "bins" defined in accordance with
the CPU idle states table provided by the driver, so that each "bin"
covers the idle duration range between the target residency of the
idle state corresponding to it and the target residency of the closest
deeper idle state. The governor collects statistics for each bin
regardless of whether or not the idle state corresponding to it is
currently enabled.
In particular, the "hits" and "misses" metrics measure the likelihood
of a situation in which both the time till the next timer (sleep
length) and the idle duration measured after wakeup fall into the
given bin. Namely, if the "hits" value is greater than the "misses"
one, that situation is more likely than the one in which the sleep
length falls into the given bin, but the idle duration measured after
wakeup falls into a bin corresponding to one of the shallower idle
states.
If the idle state corresponding to the given bin is disabled, it
cannot be selected and if it turns out to be the one that should be
selected, a shallower idle state needs to be used instead of it.
Nevertheless, the metrics collected for the bin corresponding to it
are still valid and need to be taken into account as though that
state had not been disabled.
For this reason, make teo_select() always use the "hits" and "misses"
values of the idle duration range that the sleep length falls into
even if the specific idle state corresponding to it is disabled and
if the "hits" values is greater than the "misses" one, select the
closest enabled shallower idle state in that case.
Fixes: b26bf6ab71 ("cpuidle: New timer events oriented governor for tickless systems")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: 5.1+ <stable@vger.kernel.org> # 5.1+
Rename a local variable in teo_select() in preparation for subsequent
code modifications, no intentional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: 5.1+ <stable@vger.kernel.org> # 5.1+
Prevent disabled CPU idle state with target residencies beyond the
anticipated idle duration from being taken into account by the TEO
governor.
Fixes: b26bf6ab71 ("cpuidle: New timer events oriented governor for tickless systems")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: 5.1+ <stable@vger.kernel.org> # 5.1+
Right now, guest current governors have the following ratings:
* ladder -> 10
* teo -> 19
* menu -> 20
* haltpoll -> 21
* ladder + nohz=off -> 25
haltpoll governor got introduced and it is now the default governor given
its highest rating -- with ladder+nohz being the exception -- regardless of
idle driver in the guest. An example of an undesirable case is x86 KVM
guests with MWAIT which have intel_idle registered first, and consequently
will have haltpoll be used as governor which would get limited to a poll
state and state 1 and the other states wouldn't get used.
To keep the previous defaults we decrease rating of governor to 9 (below
current lowest rating) and thus rely on @governor switch on
cpuidle_register_driver() to tie in haltpoll idle driver and governor
together.
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Notice that setting measured_us to UINT_MAX in teo_update() earlier
doesn't change the behavior of the following code, so do that and
eliminate a redundant check used for setting measured_us to UINT_MAX.
This change is not expected to alter functionality.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The TEO goveror prevents the scheduler tick from being stopped (unless
stopped already) if there is a PM QoS latency constraint for the given
CPU and the target residency of the deepest idle state matching that
constraint is below the tick boundary.
However, that is problematic if CPUs with PM QoS latency constraints
are idle for long times, because it effectively causes the tick to
run on them all the time which is wasteful. [It is also confusing
and questionable if they are full dynticks CPUs.]
To address that issue, modify the TEO governor to carry out the
entire search for the most suitable idle state (from the target
residency perspective) even if a latency constraint is present,
to allow it to determine the expected idle duration in all cases.
Also, when using the last several measured idle duration values
to refine the idle state selection, make it compare those values
with the current expected idle duration value (instead of
comparing them with the target residency of the idle state
selected so far) which should prevent the tick from being
retained when it makes sense to stop it sometimes (especially
in the presence of PM QoS latency constraints).
Fixes: b26bf6ab71 ("cpuidle: New timer events oriented governor for tickless systems")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
After commit 554c8aa8ec ("sched: idle: Select idle state before
stopping the tick") the menu governor prevents the scheduler tick from
being stopped (unless stopped already) if there is a PM QoS latency
constraint for the given CPU and the target residency of the deepest
idle state matching that constraint is below the tick boundary.
However, that is problematic if CPUs with PM QoS latency constraints
are idle for long times, because it effectively causes the tick to
run on them all the time which is wasteful. [It is also confusing
and questionable if they are full dynticks CPUs.]
To address that issue, make the menu governor allow the tick to be
stopped only if the idle duration predicted by it is beyond the tick
boundary, except when the shallowest idle state is selected upfront
and it is not a "polling" one.
Fixes: 554c8aa8ec ("sched: idle: Select idle state before stopping the tick")
Link: https://lore.kernel.org/lkml/79b247b3-e056-610e-9a07-e685dfdaa6c9@gmail.com/
Reported-by: Thomas Lindroth <thomas.lindroth@gmail.com>
Tested-by: Thomas Lindroth <thomas.lindroth@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The cpuidle_haltpoll governor, in conjunction with the haltpoll cpuidle
driver, allows guest vcpus to poll for a specified amount of time before
halting.
This provides the following benefits to host side polling:
1) The POLL flag is set while polling is performed, which allows
a remote vCPU to avoid sending an IPI (and the associated
cost of handling the IPI) when performing a wakeup.
2) The VM-exit cost can be avoided.
The downside of guest side polling is that polling is performed
even with other runnable tasks in the host.
Results comparing halt_poll_ns and server/client application
where a small packet is ping-ponged:
host --> 31.33
halt_poll_ns=300000 / no guest busy spin --> 33.40 (93.8%)
halt_poll_ns=0 / guest_halt_poll_ns=300000 --> 32.73 (95.7%)
For the SAP HANA benchmarks (where idle_spin is a parameter
of the previous version of the patch, results should be the
same):
hpns == halt_poll_ns
idle_spin=0/ idle_spin=800/ idle_spin=0/
hpns=200000 hpns=0 hpns=800000
DeleteC06T03 (100 thread) 1.76 1.71 (-3%) 1.78 (+1%)
InsertC16T02 (100 thread) 2.14 2.07 (-3%) 2.18 (+1.8%)
DeleteC00T01 (1 thread) 1.34 1.28 (-4.5%) 1.29 (-3.7%)
UpdateC00T03 (1 thread) 4.72 4.18 (-12%) 4.53 (-5%)
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Since this field is shared by all governors, move it to
cpuidle device structure.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Based on 1 normalized pattern(s):
this code is licenced under the gpl version 2 as described in the
copying file that acompanies the linux kernel
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-only
has been chosen to replace the boilerplate/reference in 1 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Alexios Zavras <alexios.zavras@intel.com>
Reviewed-by: Allison Randal <allison@lohutok.net>
Reviewed-by: Steve Winslow <swinslow@gmail.com>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190528171439.466585205@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add SPDX license identifiers to all Make/Kconfig files which:
- Have no license information of any form
These files fall under the project license, GPL v2 only. The resulting SPDX
license identifier is:
GPL-2.0-only
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The variance computation in get_typical_interval() may overflow if
the square of the value of diff exceeds the maximum for the int64_t
data type value which basically is the case when it is of the order
of UINT_MAX.
However, data points so far in the future don't matter for idle
state selection anyway, so change the initial threshold value in
get_typical_interval() to INT_MAX which will cause more "outlying"
data points to be discarded without affecting the selection result.
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The venerable menu governor does some things that are quite
questionable in my view.
First, it includes timer wakeups in the pattern detection data and
mixes them up with wakeups from other sources which in some cases
causes it to expect what essentially would be a timer wakeup in a
time frame in which no timer wakeups are possible (because it knows
the time until the next timer event and that is later than the
expected wakeup time).
Second, it uses the extra exit latency limit based on the predicted
idle duration and depending on the number of tasks waiting on I/O,
even though those tasks may run on a different CPU when they are
woken up. Moreover, the time ranges used by it for the sleep length
correction factors depend on whether or not there are tasks waiting
on I/O, which again doesn't imply anything in particular, and they
are not correlated to the list of available idle states in any way
whatever.
Also, the pattern detection code in menu may end up considering
values that are too large to matter at all, in which cases running
it is a waste of time.
A major rework of the menu governor would be required to address
these issues and the performance of at least some workloads (tuned
specifically to the current behavior of the menu governor) is likely
to suffer from that. It is thus better to introduce an entirely new
governor without them and let everybody use the governor that works
better with their actual workloads.
The new governor introduced here, the timer events oriented (TEO)
governor, uses the same basic strategy as menu: it always tries to
find the deepest idle state that can be used in the given conditions.
However, it applies a different approach to that problem.
First, it doesn't use "correction factors" for the time till the
closest timer, but instead it tries to correlate the measured idle
duration values with the available idle states and use that
information to pick up the idle state that is most likely to "match"
the upcoming CPU idle interval.
Second, it doesn't take the number of "I/O waiters" into account at
all and the pattern detection code in it avoids taking timer wakeups
into account. It also only uses idle duration values less than the
current time till the closest timer (with the tick excluded) for that
purpose.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
- Fix build regression in the intel_pstate driver that doesn't
build without CONFIG_ACPI after recent changes (Dominik Brodowski).
- One of the heuristics in the menu cpuidle governor is based on a
function returning 0 most of the time, so drop it and clean up
the scheduler code related to it (Daniel Lezcano).
- Prevent the arm_big_little cpufreq driver from being used on ARM64
which is not suitable for it and drop the arm_big_little_dt driver
that is not used any more (Sudeep Holla).
- Prevent the hung task watchdog from triggering during resume from
system-wide sleep states by disabling it before freezing tasks and
enabling it again after they have been thawed (Vitaly Kuznetsov).
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2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=/bnR
-----END PGP SIGNATURE-----
Merge tag 'pm-4.20-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull more power management updates from Rafael Wysocki:
"These remove a questionable heuristic from the menu cpuidle governor,
fix a recent build regression in the intel_pstate driver, clean up ARM
big-Little support in cpufreq and fix up hung task watchdog's
interaction with system-wide power management transitions.
Specifics:
- Fix build regression in the intel_pstate driver that doesn't build
without CONFIG_ACPI after recent changes (Dominik Brodowski).
- One of the heuristics in the menu cpuidle governor is based on a
function returning 0 most of the time, so drop it and clean up the
scheduler code related to it (Daniel Lezcano).
- Prevent the arm_big_little cpufreq driver from being used on ARM64
which is not suitable for it and drop the arm_big_little_dt driver
that is not used any more (Sudeep Holla).
- Prevent the hung task watchdog from triggering during resume from
system-wide sleep states by disabling it before freezing tasks and
enabling it again after they have been thawed (Vitaly Kuznetsov)"
* tag 'pm-4.20-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
kernel: hung_task.c: disable on suspend
cpufreq: remove unused arm_big_little_dt driver
cpufreq: drop ARM_BIG_LITTLE_CPUFREQ support for ARM64
cpufreq: intel_pstate: Fix compilation for !CONFIG_ACPI
cpuidle: menu: Remove get_loadavg() from the performance multiplier
sched: Factor out nr_iowait and nr_iowait_cpu
There are several definitions of those functions/macros in places that
mess with fixed-point load averages. Provide an official version.
[akpm@linux-foundation.org: fix missed conversion in block/blk-iolatency.c]
Link: http://lkml.kernel.org/r/20180828172258.3185-5-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Tested-by: Daniel Drake <drake@endlessm.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The function get_loadavg() returns almost always zero. To be more
precise, statistically speaking for a total of 1023379 times passing
in the function, the load is equal to zero 1020728 times, greater than
100, 610 times, the remaining is between 0 and 5.
In 2011, the get_loadavg() was removed from the Android tree because
of the above [1]. At this time, the load was:
unsigned long this_cpu_load(void)
{
struct rq *this = this_rq();
return this->cpu_load[0];
}
In 2014, the code was changed by commit 372ba8cb46 (cpuidle: menu: Lookup CPU
runqueues less) and the load is:
void get_iowait_load(unsigned long *nr_waiters, unsigned long *load)
{
struct rq *rq = this_rq();
*nr_waiters = atomic_read(&rq->nr_iowait);
*load = rq->load.weight;
}
with the same result.
Both measurements show using the load in this code path does no matter
anymore. Removing it.
[1] 4dedd9f124
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
If the minimum interval taken into account in the average computation
loop in get_typical_interval() is less than the expected idle
duration determined so far, the resultant average cannot be greater
than that value as well and the entire return result of the function
is going to be discarded anyway going forward.
In that case, it is a waste of time to carry out the remaining
computations in get_typical_interval(), so avoid that by returning
early if the minimum interval is not below the expected idle duration.
No intentional changes of behavior.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Since the correction factor cannot be greater than RESOLUTION * DECAY,
the result of the predicted_us computation in menu_select() cannot be
greater than data->next_timer_us, so it is not necessary to compare
the "typical interval" value coming from get_typical_interval() with
data->next_timer_us separately.
It is sufficient to copmare predicted_us with the return value of
get_typical_interval() directly, so do that and drop the now
redundant expected_interval variable.
No intentional changes of behavior.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
After some recent menu governor changes, the promotion of the
"polling" state to a physical one is mostly controlled by the
latency limit (resulting from the "interactivity" factor) and
not by the time to the closest timer event, so it should be
sufficient to check the exit latency of that state for this
purpose (of course, its target residency still needs to be
within the next timer event range for energy-efficiency).
Also, the physical state the "polling" one is promoted to need not
be the next one in principle (in case the next state is disabled,
for example).
For these reasons, simplify the checks made to decide whether or
not to promote the "polling" state to a physical one and update
the target idle duration when it is promoted in case the residency
of the new state turns out to be above the tick boundary (in which
case there is no reason to stop the tick).
Tested-by: Doug Smythies <dsmythies@telus.net>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
It is better to always update data->bucket before returning from
menu_select() to avoid updating the correction factor for a stale
bucket, so combine the latency_req == 0 special check with the more
general check below.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
If the next timer event (not including the tick) is closer than the
target residency of the second state or the PM QoS latency constraint
is below its exit latency, state[0] will be used regardless of any
other factors, so skip the computations in menu_select() then and
return 0 straight away from it.
Still, do that after the bucket has been determined to avoid
updating the correction factor for a stale bucket.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
It is not necessary to update data->last_state_idx in menu_select()
as it only is used in menu_update() which only runs when
data->needs_update is set and that is set only when updating
data->last_state_idx in menu_reflect().
Accordingly, drop the update of data->last_state_idx from
menu_select() and get rid of the (now redundant) "out" label
from it.
No intentional behavior changes.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Rearrange the code in menu_select() so that the loop over idle states
always starts from 0 and get rid of the first_idx variable.
While at it, add two empty lines to separate conditional statements
from one another.
No intentional behavior changes.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Since menu_select() can only set first_idx to 1 if the exit latency
of the second state is not greater than the latency limit, it should
first determine that limit. Thus first_idx should be computed after
the "interactivity" factor has been taken into account.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewedy-by: Daniel Lezcano <daniel.lezcano@linaro.org>
If the CPU exits the "polling" state due to the time limit in the
loop in poll_idle(), this is not a real wakeup and it just means
that the "polling" state selection was not adequate. The governor
mispredicted short idle duration, but had a more suitable state been
selected, the CPU might have spent more time in it. In fact, there
is no reason to expect that there would have been a wakeup event
earlier than the next timer in that case.
Handling such cases as regular wakeups in menu_update() may cause the
menu governor to make suboptimal decisions going forward, but ignoring
them altogether would not be correct either, because every time
menu_select() is invoked, it makes a separate new attempt to predict
the idle duration taking distinct time to the closest timer event as
input and the outcomes of all those attempts should be recorded.
For this reason, make menu_update() always assume that if the
"polling" state was exited due to the time limit, the next proper
wakeup event for the CPU would be the next timer event (not
including the tick).
Fixes: a37b969a61 "cpuidle: poll_state: Add time limit to poll_idle()"
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The predicted_us field in struct menu_device is only accessed in
menu_select(), so replace it with a local variable in that function.
With that, stop using expected_interval instead of predicted_us to
store the new predicted idle duration value if it is set to the
selected state's target residency which is quite confusing.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
