wonderful-mobile/exynos-linux-stable
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
exynos-linux-stable/kernel/sched/cpufreq_schedutil.c

1471 lines
38 KiB
C
Raw Permalink Normal View History

cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
/*
* CPUFreq governor based on scheduler-provided CPU utilization data.
*
* Copyright (C) 2016, Intel Corporation
* Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
cpufreq: schedutil: Improve prints messages with pr_fmt Prefix print messages with KBUILD_MODNAME, i.e 'cpufreq_schedutil: '. This helps to keep similar formatting for all the print messages particular to a file and identify those easily in kernel logs. Its already done this way for rest of the governors. Along with that, remove the (now) redundant bits from a print message. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-05-18 17:55:28 +05:30
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
#include <linux/cpufreq.h>
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
#include <linux/kthread.h>
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
#include <linux/slab.h>
sched: schedutil: Add slack_timer and pm_qos handler 1. the slack_timer works to prevent the fequency from being held unnecessarily high when the cpu sleeps long in a high utilization. 2. The Pm_qos handler tells the governor if the frequency that was held high by qos could fall. Change-Id: I5f2d5be73b8d8294a35b2810cc9721264ec55964 Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com>
2017-11-02 20:47:24 +09:00
#include <linux/cpu_pm.h>
#include <linux/of.h>
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
#include <linux/ems.h>
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
#include <trace/events/power.h>
#include "sched.h"
ANDROID: sched/cpufreq: make schedutil use WALT signal If WALT is available and enabled, make schedutil governor use its utilization signal. Change-Id: I92bc37989447a76616e9bcc4e9e8616774fb9925 Signed-off-by: Juri Lelli <juri.lelli@arm.com> [we need to use boosted_cpu_util for schedutil, so make it not static] Signed-off-by: Chris Redpath <chris.redpath@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-12-14 16:10:10 +00:00
#include "tune.h"
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
#include "ems/ems.h"
ANDROID: sched/cpufreq: make schedutil use WALT signal If WALT is available and enabled, make schedutil governor use its utilization signal. Change-Id: I92bc37989447a76616e9bcc4e9e8616774fb9925 Signed-off-by: Juri Lelli <juri.lelli@arm.com> [we need to use boosted_cpu_util for schedutil, so make it not static] Signed-off-by: Chris Redpath <chris.redpath@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-12-14 16:10:10 +00:00
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
#include <linux/kair.h>
/**
* 2nd argument of kair_obj_creator() experimentally decided by KAIR client
* itself, which represents how much variant the random variable registered to
* the KAIR instance can behave at most, in terms of referencing d2u_decl_cmtpdf
* table(maximum index of d2u_decl_cmtpdf table).
**/
#define UTILAVG_KAIR_VARIANCE 16
DECLARE_KAIRISTICS(cpufreq, 32, 25, 24, 25);
#endif
ANDROID: sched/rt: Add schedtune accounting to rt task enqueue/dequeue rt tasks are currently not eligible for schedtune boosting. Make it so by adding enqueue/dequeue hooks. For rt tasks, schedtune only acts as a frequency boosting framework, it has no impact on placement decisions and the prefer_idle attribute is not used. Also prepare schedutil use of boosted util for rt task boosting With this change, schedtune accounting will include rt class tasks, however boosting currently only applies to the utilization provided by fair class tasks. Sum up the tracked CPU utilization applying boost to the aggregate util instead - this includes RT task util in the boosting if any tasks are runnable. Scenario 1, considering one CPU: 1x rt task running, util 250, boost 0 1x cfs task runnable, util 250, boost 50 previous util=250+(50pct_boosted_250) = 887 new util=50_pct_boosted_500 = 762 Scenario 2, considering one CPU: 1x rt task running, util 250, boost 50 1x cfs task runnable, util 250, boost 0 previous util=250+250 = 500 new util=50_pct_boosted_500 = 762 Scenario 3, considering one CPU: 1x rt task running, util 250, boost 50 1x cfs task runnable, util 250, boost 50 previous util=250+(50pct_boosted_250) = 887 new util=50_pct_boosted_500 = 762 Scenario 4: 1x rt task running, util 250, boost 50 previous util=250 = 250 new util=50_pct_boosted_250 = 637 Change-Id: Ie287cbd0692468525095b5024db9faac8b2f4878 Signed-off-by: Chris Redpath <chris.redpath@arm.com>
2018-07-09 16:54:02 +01:00
unsigned long boosted_cpu_util(int cpu, unsigned long other_util);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
/* Stub out fast switch routines present on mainline to reduce the backport
* overhead. */
#define cpufreq_driver_fast_switch(x, y) 0
#define cpufreq_enable_fast_switch(x)
#define cpufreq_disable_fast_switch(x)
#define LATENCY_MULTIPLIER (1000)
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
#define SUGOV_KTHREAD_PRIORITY 50
UPSTREAM: cpufreq: schedutil: move slow path from workqueue to SCHED_FIFO task If slow path frequency changes are conducted in a SCHED_OTHER context then they may be delayed for some amount of time, including indefinitely, when real time or deadline activity is taking place. Move the slow path to a real time kernel thread. In the future the thread should be made SCHED_DEADLINE. The RT priority is arbitrarily set to 50 for now. Hackbench results on ARM Exynos, dual core A15 platform for 10 iterations: $ hackbench -s 100 -l 100 -g 10 -f 20 Before After --------------------------------- 1.808 1.603 1.847 1.251 2.229 1.590 1.952 1.600 1.947 1.257 1.925 1.627 2.694 1.620 1.258 1.621 1.919 1.632 1.250 1.240 Average: 1.8829 1.5041 Based on initial work by Steve Muckle. Change-Id: I8bb111435b0871e961a6f26db1151c91624817b8 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry-picked from commit 02a7b1ee3baa15a98b541d8cfd156bbe1a091c20) [fixed cherry-pick issue] Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-15 13:53:22 +05:30
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
struct sugov_tunables {
struct gov_attr_set attr_set;
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
unsigned int up_rate_limit_us;
unsigned int down_rate_limit_us;
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
bool fb_legacy;
#endif
FROMLIST: sched: Make iowait_boost optional in schedutil We should apply the iowait boost only if cpufreq policy has iowait boost enabled. Also make it a schedutil configuration from sysfs so it can be turned on/off if needed (by default initialize it to the policy value). For systems that don't need/want it enabled, such as those on arm64 based mobile devices that are battery operated, it saves energy when the cpufreq driver policy doesn't have it enabled (details below): Here are some results for energy measurements collected running a YouTube video for 30 seconds: Before: 8.042533 mWh After: 7.948377 mWh Energy savings is ~1.2% Bug: 38010527 Link: https://lkml.org/lkml/2017/5/19/42 Change-Id: If124076ad0c16ade369253840dedfbf870aff927 Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-05-18 22:46:10 -07:00
bool iowait_boost_enable;
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
bool exp_util;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
};
struct sugov_policy {
struct cpufreq_policy *policy;
struct sugov_tunables *tunables;
struct list_head tunables_hook;
raw_spinlock_t update_lock; /* For shared policies */
u64 last_freq_update_time;
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
s64 min_rate_limit_ns;
s64 up_rate_delay_ns;
s64 down_rate_delay_ns;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
unsigned int next_freq;
cpufreq: schedutil: move cached_raw_freq to struct sugov_policy commit 6c4f0fa643cb9e775dcc976e3db00d649468ff1d upstream. cached_raw_freq applies to the entire cpufreq policy and not individual CPUs. Apart from wasting per-cpu memory, it is actually wrong to keep it in struct sugov_cpu as we may end up comparing next_freq with a stale cached_raw_freq of a random CPU. Move cached_raw_freq to struct sugov_policy. Fixes: 5cbea46984d6 (cpufreq: schedutil: map raw required frequency to driver frequency) Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Amit Pundir <amit.pundir@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-03-02 14:03:20 +05:30
unsigned int cached_raw_freq;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
/* The next fields are only needed if fast switch cannot be used. */
struct irq_work irq_work;
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
struct kthread_work work;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
struct mutex work_lock;
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
struct kthread_worker worker;
struct task_struct *thread;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
bool work_in_progress;
bool need_freq_update;
cpufreq: schedutil: Don't skip freq update when limits change To avoid reducing the frequency of a CPU prematurely, we skip reducing the frequency if the CPU had been busy recently. This should not be done when the limits of the policy are changed, for example due to thermal throttling. We should always get the frequency within the new limits as soon as possible. Trying to fix this by using only one flag, i.e. need_freq_update, can lead to a race condition where the flag gets cleared without forcing us to change the frequency at least once. And so this patch introduces another flag to avoid that race condition. Fixes: ecd288429126 ("cpufreq: schedutil: Don't set next_freq to UINT_MAX") Cc: v4.18+ <stable@vger.kernel.org> # v4.18+ Reported-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-08-07 12:36:01 +05:30
bool limits_changed;
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
bool be_stochastic;
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
};
struct sugov_cpu {
struct update_util_data update_util;
struct sugov_policy *sg_policy;
UPSTREAM: cpufreq: schedutil: Make iowait boost more energy efficient Currently the iowait_boost feature in schedutil makes the frequency go to max on iowait wakeups. This feature was added to handle a case that Peter described where the throughput of operations involving continuous I/O requests [1] is reduced due to running at a lower frequency, however the lower throughput itself causes utilization to be low and hence causing frequency to be low hence its "stuck". Instead of going to max, its also possible to achieve the same effect by ramping up to max if there are repeated in_iowait wakeups happening. This patch is an attempt to do that. We start from a lower frequency (policy->min) and double the boost for every consecutive iowait update until we reach the maximum iowait boost frequency (iowait_boost_max). I ran a synthetic test (continuous O_DIRECT writes in a loop) on an x86 machine with intel_pstate in passive mode using schedutil. In this test the iowait_boost value ramped from 800MHz to 4GHz in 60ms. The patch achieves the desired improved throughput as the existing behavior. [1] https://patchwork.kernel.org/patch/9735885/ Change-Id: I4a018434a50f4ca29ec15b03465f6dc212e54423 Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: Len Brown <lenb@kernel.org> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Suggested-by: Peter Zijlstra <peterz@infradead.org> Suggested-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-08-28 09:56:27 -07:00
bool iowait_boost_pending;
UPSTREAM: cpufreq: schedutil: Use unsigned int for iowait boost Make iowait_boost and iowait_boost_max as unsigned int since its unit is kHz and this is consistent with struct cpufreq_policy. Also change the local variables in sugov_iowait_boost to match this. Change-Id: I6c67ed94c57c4bdb24bada4b97045593fcb95d2e Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: Len Brown <lenb@kernel.org> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-07-23 08:54:26 -07:00
unsigned int iowait_boost;
unsigned int iowait_boost_max;
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
u64 last_update;
cpufreq: schedutil: map raw required frequency to driver frequency The slow-path frequency transition path is relatively expensive as it requires waking up a thread to do work. Should support be added for remote CPU cpufreq updates that is also expensive since it requires an IPI. These activities should be avoided if they are not necessary. To that end, calculate the actual driver-supported frequency required by the new utilization value in schedutil by using the recently added cpufreq_driver_resolve_freq API. If it is the same as the previously requested driver frequency then there is no need to continue with the update assuming the cpu frequency limits have not changed. This will have additional benefits should the semantics of the rate limit be changed to apply solely to frequency transitions rather than to frequency calculations in schedutil. The last raw required frequency is cached. This allows the driver frequency lookup to be skipped in the event that the new raw required frequency matches the last one, assuming a frequency update has not been forced due to limits changing (indicated by a next_freq value of UINT_MAX, see sugov_should_update_freq). Signed-off-by: Steve Muckle <smuckle@linaro.org> Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-07-13 13:25:26 -07:00
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
/**
* KAIR instance which should be referenced in percpu manner,
* and data accordingly to handle the target job intensity.
**/
struct kair_class *util_vessel;
unsigned long cached_util;
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
/* The fields below are only needed when sharing a policy. */
unsigned long util;
unsigned long max;
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
unsigned int flags;
UPSTREAM: cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely The way the schedutil governor uses the PELT metric causes it to underestimate the CPU utilization in some cases. That can be easily demonstrated by running kernel compilation on a Sandy Bridge Intel processor, running turbostat in parallel with it and looking at the values written to the MSR_IA32_PERF_CTL register. Namely, the expected result would be that when all CPUs were 100% busy, all of them would be requested to run in the maximum P-state, but observation shows that this clearly isn't the case. The CPUs run in the maximum P-state for a while and then are requested to run slower and go back to the maximum P-state after a while again. That causes the actual frequency of the processor to visibly oscillate below the sustainable maximum in a jittery fashion which clearly is not desirable. That has been attributed to CPU utilization metric updates on task migration that cause the total utilization value for the CPU to be reduced by the utilization of the migrated task. If that happens, the schedutil governor may see a CPU utilization reduction and will attempt to reduce the CPU frequency accordingly right away. That may be premature, though, for example if the system is generally busy and there are other runnable tasks waiting to be run on that CPU already. This is unlikely to be an issue on systems where cpufreq policies are shared between multiple CPUs, because in those cases the policy utilization is computed as the maximum of the CPU utilization values over the whole policy and if that turns out to be low, reducing the frequency for the policy most likely is a good idea anyway. On systems with one CPU per policy, however, it may affect performance adversely and even lead to increased energy consumption in some cases. On those systems it may be addressed by taking another utilization metric into consideration, like whether or not the CPU whose frequency is about to be reduced has been idle recently, because if that's not the case, the CPU is likely to be busy in the near future and its frequency should not be reduced. To that end, use the counter of idle calls in the timekeeping code. Namely, make the schedutil governor look at that counter for the current CPU every time before its frequency is about to be reduced. If the counter has not changed since the previous iteration of the governor computations for that CPU, the CPU has been busy for all that time and its frequency should not be decreased, so if the new frequency would be lower than the one set previously, the governor will skip the frequency update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes <joelaf@google.com> (cherry picked from commit b7eaf1aab9f8bd2e49fceed77ebc66c1b5800718) (simple CPUFREQ_RT_DL vs CPUFREQ_DL usage conflicts) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: I531ec02c052944ee07a904dc2a25c59948ee762b Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:24:58 +01:00
/* The field below is for single-CPU policies only. */
#ifdef CONFIG_NO_HZ_COMMON
unsigned long saved_idle_calls;
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
};
static DEFINE_PER_CPU(struct sugov_cpu, sugov_cpu);
BACKPORT: cpufreq: schedutil: Cache tunables on governor exit Currently when all the related CPUs from a policy go offline or the governor is switched, cpufreq framework calls sugov_exit() that frees the governor tunables. When any of the related CPUs comes back online or governor is switched back to schedutil sugov_init() gets called which allocates a fresh set of tunables that are set to default values. This can cause the userspace settings to those tunables to be lost across governor switches or when an entire cluster is hotplugged out. To prevent this, save the tunable values on governor exit. Restore these values to the newly allocated tunables on governor init. Change-Id: I671d4d0e1a4e63e948bfddb0005367df33c0c249 Signed-off-by: Rohit Gupta <rohgup@codeaurora.org> [Caching and restoring different tunables.] Signed-off-by: joshuous <joshuous@gmail.com> Signed-off-by: joshuous <joshuous@gmail.com> (cherry picked from commit a6447dc2b4f53b87e968248935d7ef51d62df73f) (cherry picked from commit d4fbc6341db19a040685e9d295cc5c967d5c3bc4)
2018-05-29 17:23:04 +02:00
static DEFINE_PER_CPU(struct sugov_tunables *, cached_tunables);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sched: schedutil: Add slack_timer and pm_qos handler 1. the slack_timer works to prevent the fequency from being held unnecessarily high when the cpu sleeps long in a high utilization. 2. The Pm_qos handler tells the governor if the frequency that was held high by qos could fall. Change-Id: I5f2d5be73b8d8294a35b2810cc9721264ec55964 Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com>
2017-11-02 20:47:24 +09:00
/******************* exynos specific function *******************/
#define DEFAULT_EXPIRED_TIME 70
struct sugov_exynos {
/* for slack timer */
unsigned long min;
int enabled;
bool started;
int expired_time;
struct timer_list timer;
/* pm_qos_class */
int qos_min_class;
};
static DEFINE_PER_CPU(struct sugov_exynos, sugov_exynos);
static void sugov_stop_slack(int cpu);
static void sugov_start_slack(int cpu);
static void sugov_update_min(struct cpufreq_policy *policy);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
/************************ Governor internals ***********************/
static bool sugov_should_update_freq(struct sugov_policy *sg_policy, u64 time)
{
s64 delta_ns;
cpufreq: schedutil: Don't skip freq update when limits change To avoid reducing the frequency of a CPU prematurely, we skip reducing the frequency if the CPU had been busy recently. This should not be done when the limits of the policy are changed, for example due to thermal throttling. We should always get the frequency within the new limits as soon as possible. Trying to fix this by using only one flag, i.e. need_freq_update, can lead to a race condition where the flag gets cleared without forcing us to change the frequency at least once. And so this patch introduces another flag to avoid that race condition. Fixes: ecd288429126 ("cpufreq: schedutil: Don't set next_freq to UINT_MAX") Cc: v4.18+ <stable@vger.kernel.org> # v4.18+ Reported-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-08-07 12:36:01 +05:30
if (unlikely(sg_policy->limits_changed)) {
sg_policy->limits_changed = false;
sg_policy->need_freq_update = true;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
return true;
cpufreq: schedutil: Don't skip freq update when limits change To avoid reducing the frequency of a CPU prematurely, we skip reducing the frequency if the CPU had been busy recently. This should not be done when the limits of the policy are changed, for example due to thermal throttling. We should always get the frequency within the new limits as soon as possible. Trying to fix this by using only one flag, i.e. need_freq_update, can lead to a race condition where the flag gets cleared without forcing us to change the frequency at least once. And so this patch introduces another flag to avoid that race condition. Fixes: ecd288429126 ("cpufreq: schedutil: Don't set next_freq to UINT_MAX") Cc: v4.18+ <stable@vger.kernel.org> # v4.18+ Reported-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-08-07 12:36:01 +05:30
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
/* No need to recalculate next freq for min_rate_limit_us
* at least. However we might still decide to further rate
* limit once frequency change direction is decided, according
* to the separate rate limits.
*/
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
delta_ns = time - sg_policy->last_freq_update_time;
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
return delta_ns >= sg_policy->min_rate_limit_ns;
}
static bool sugov_up_down_rate_limit(struct sugov_policy *sg_policy, u64 time,
unsigned int next_freq)
{
s64 delta_ns;
delta_ns = time - sg_policy->last_freq_update_time;
if (next_freq > sg_policy->next_freq &&
delta_ns < sg_policy->up_rate_delay_ns)
return true;
if (next_freq < sg_policy->next_freq &&
delta_ns < sg_policy->down_rate_delay_ns)
return true;
return false;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
static void sugov_update_commit(struct sugov_policy *sg_policy, u64 time,
unsigned int next_freq)
{
struct cpufreq_policy *policy = sg_policy->policy;
BACKPORT: schedutil: Reset cached freq if it is not in sync with next_freq 'cached_raw_freq' is used to get the next frequency quickly but should always be in sync with sg_policy->next_freq. There are cases where it is not and in such cases it should be reset to avoid switching to incorrect frequencies. Consider this case for example: - policy->cur is 1.2 GHz (Max) - New request comes for 780 MHz and we store that in cached_raw_freq. - Based on 780 MHz, we calculate the effective frequency as 800 MHz. - We then decide not to update the frequency as sugov_up_down_rate_limit() return true. - Here cached_raw_freq is 780 MHz and sg_policy->next_freq is 1.2 GHz. - Now if the utilization doesn't change in next request, then the next target frequency will still be 780 MHz and it will match with cached_raw_freq and so we will directly return 1.2 GHz instead of 800 MHz. BACKPORT of upstream commit 07458f6a5171 ("cpufreq: schedutil: Reset cached_raw_freq when not in sync with next_freq"). This also updates sugov_update_commit() for handling up/down tunables, which aren't present in mainline. Change-Id: Ie86465231e7cb265e5b4c26f59d6faf8d9630b0a Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
2017-11-08 19:47:36 +05:30
if (sugov_up_down_rate_limit(sg_policy, time, next_freq)) {
/* Reset cached freq as next_freq isn't changed */
sg_policy->cached_raw_freq = 0;
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
return;
BACKPORT: schedutil: Reset cached freq if it is not in sync with next_freq 'cached_raw_freq' is used to get the next frequency quickly but should always be in sync with sg_policy->next_freq. There are cases where it is not and in such cases it should be reset to avoid switching to incorrect frequencies. Consider this case for example: - policy->cur is 1.2 GHz (Max) - New request comes for 780 MHz and we store that in cached_raw_freq. - Based on 780 MHz, we calculate the effective frequency as 800 MHz. - We then decide not to update the frequency as sugov_up_down_rate_limit() return true. - Here cached_raw_freq is 780 MHz and sg_policy->next_freq is 1.2 GHz. - Now if the utilization doesn't change in next request, then the next target frequency will still be 780 MHz and it will match with cached_raw_freq and so we will directly return 1.2 GHz instead of 800 MHz. BACKPORT of upstream commit 07458f6a5171 ("cpufreq: schedutil: Reset cached_raw_freq when not in sync with next_freq"). This also updates sugov_update_commit() for handling up/down tunables, which aren't present in mainline. Change-Id: Ie86465231e7cb265e5b4c26f59d6faf8d9630b0a Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
2017-11-08 19:47:36 +05:30
}
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
sched: fair: get rid of frt_boost_threshold * And remove other sched changes / Sync with N770F sched
2023-03-21 20:07:23 +04:00
if (policy->cur == next_freq)
UPSTREAM: cpufreq: schedutil: Trace frequency only if it has changed sugov_update_commit() calls trace_cpu_frequency() to record the current CPU frequency if it has not changed in the fast switch case to prevent utilities from getting confused (they may report that the CPU is idle if the frequency has not been recorded for too long, for example). However, that may cause the tracepoint to be triggered quite often for no real reason (if the frequency doesn't change, we will not modify the last update time stamp and governor computations may run again shortly when that happens), so don't do that (arguably, it is done to work around a utilities bug anyway). That allows code duplication in sugov_update_commit() to be reduced somewhat too. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> (cherry picked from commit 38d4ea229d25d30be6bf41bcd6cd663a587866ca) (conflicts with sugov_up_down_rate_limit resolved) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ia019dda29b8c1c4cf3553da75c88d066eb5674e9 Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:27:07 +01:00
return;
sg_policy->next_freq = next_freq;
sg_policy->last_freq_update_time = time;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
if (policy->fast_switch_enabled) {
next_freq = cpufreq_driver_fast_switch(policy, next_freq);
cpufreq: Return 0 from ->fast_switch() on errors CPUFREQ_ENTRY_INVALID is a special symbol which is used to specify that an entry in the cpufreq table is invalid. But using it outside of the scope of the cpufreq table looks a bit incorrect. We can represent an invalid frequency by writing it as 0 instead if we need. Note that it is already done that way for the return value of the ->get() callback. Lets do the same for ->fast_switch() and not use CPUFREQ_ENTRY_INVALID outside of the scope of cpufreq table. Also update the comment over cpufreq_driver_fast_switch() to clearly mention what this returns. None of the drivers return CPUFREQ_ENTRY_INVALID as of now from ->fast_switch() callback and so we don't need to update any of those. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2017-08-09 10:21:46 +05:30
if (!next_freq)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
return;
policy->cur = next_freq;
trace_cpu_frequency(next_freq, smp_processor_id());
schedutil: Allow cpufreq requests to be made even when kthread kicked Currently there is a chance of a schedutil cpufreq update request to be dropped if there is a pending update request. This pending request can be delayed if there is a scheduling delay of the irq_work and the wake up of the schedutil governor kthread. A very bad scenario is when a schedutil request was already just made, such as to reduce the CPU frequency, then a newer request to increase CPU frequency (even sched deadline urgent frequency increase requests) can be dropped, even though the rate limits suggest that its Ok to process a request. This is because of the way the work_in_progress flag is used. This patch improves the situation by allowing new requests to happen even though the old one is still being processed. Note that in this approach, if an irq_work was already issued, we just update next_freq and don't bother to queue another request so there's no extra work being done to make this happen. Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Juri Lelli <juri.lelli@redhat.com> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2018-05-22 15:55:53 -07:00
} else if (!sg_policy->work_in_progress) {
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sg_policy->work_in_progress = true;
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
irq_work_queue(&sg_policy->irq_work);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
}
[COMMON] sched: support freqvar tuner Change-Id: Ifd2fb13a6ab44f3137aca816c8537a85af0e7850 Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2018-01-19 16:42:10 +09:00
#ifdef CONFIG_FREQVAR_TUNE
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
unsigned long freqvar_boost_vector(int cpu, unsigned long util);
[COMMON] sched: support freqvar tuner Change-Id: Ifd2fb13a6ab44f3137aca816c8537a85af0e7850 Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2018-01-19 16:42:10 +09:00
#else
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
static inline unsigned long freqvar_boost_vector(int cpu, unsigned long util)
[COMMON] sched: support freqvar tuner Change-Id: Ifd2fb13a6ab44f3137aca816c8537a85af0e7850 Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2018-01-19 16:42:10 +09:00
{
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
return util;
[COMMON] sched: support freqvar tuner Change-Id: Ifd2fb13a6ab44f3137aca816c8537a85af0e7850 Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2018-01-19 16:42:10 +09:00
}
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
/**
* get_next_freq - Compute a new frequency for a given cpufreq policy.
UPSTREAM: cpufreq: schedutil: Pass sg_policy to get_next_freq() get_next_freq() uses sg_cpu only to get sg_policy, which the callers of get_next_freq() already have. Pass sg_policy instead of sg_cpu to get_next_freq(), to make it more efficient. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 655cb1ebff4b7918fc560502c3297af2d3c7d114) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ia210058da32930a6cdb18258aa679cd1a44a747e Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-03-02 14:03:21 +05:30
* @sg_policy: schedutil policy object to compute the new frequency for.
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
* @util: Current CPU utilization.
* @max: CPU capacity.
*
* If the utilization is frequency-invariant, choose the new frequency to be
* proportional to it, that is
*
* next_freq = C * max_freq * util / max
*
* Otherwise, approximate the would-be frequency-invariant utilization by
* util_raw * (curr_freq / max_freq) which leads to
*
* next_freq = C * curr_freq * util_raw / max
*
* Take C = 1.25 for the frequency tipping point at (util / max) = 0.8.
cpufreq: schedutil: map raw required frequency to driver frequency The slow-path frequency transition path is relatively expensive as it requires waking up a thread to do work. Should support be added for remote CPU cpufreq updates that is also expensive since it requires an IPI. These activities should be avoided if they are not necessary. To that end, calculate the actual driver-supported frequency required by the new utilization value in schedutil by using the recently added cpufreq_driver_resolve_freq API. If it is the same as the previously requested driver frequency then there is no need to continue with the update assuming the cpu frequency limits have not changed. This will have additional benefits should the semantics of the rate limit be changed to apply solely to frequency transitions rather than to frequency calculations in schedutil. The last raw required frequency is cached. This allows the driver frequency lookup to be skipped in the event that the new raw required frequency matches the last one, assuming a frequency update has not been forced due to limits changing (indicated by a next_freq value of UINT_MAX, see sugov_should_update_freq). Signed-off-by: Steve Muckle <smuckle@linaro.org> Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-07-13 13:25:26 -07:00
*
* The lowest driver-supported frequency which is equal or greater than the raw
* next_freq (as calculated above) is returned, subject to policy min/max and
* cpufreq driver limitations.
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
*/
UPSTREAM: cpufreq: schedutil: Pass sg_policy to get_next_freq() get_next_freq() uses sg_cpu only to get sg_policy, which the callers of get_next_freq() already have. Pass sg_policy instead of sg_cpu to get_next_freq(), to make it more efficient. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 655cb1ebff4b7918fc560502c3297af2d3c7d114) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ia210058da32930a6cdb18258aa679cd1a44a747e Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-03-02 14:03:21 +05:30
static unsigned int get_next_freq(struct sugov_policy *sg_policy,
unsigned long util, unsigned long max)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
cpufreq: schedutil: map raw required frequency to driver frequency The slow-path frequency transition path is relatively expensive as it requires waking up a thread to do work. Should support be added for remote CPU cpufreq updates that is also expensive since it requires an IPI. These activities should be avoided if they are not necessary. To that end, calculate the actual driver-supported frequency required by the new utilization value in schedutil by using the recently added cpufreq_driver_resolve_freq API. If it is the same as the previously requested driver frequency then there is no need to continue with the update assuming the cpu frequency limits have not changed. This will have additional benefits should the semantics of the rate limit be changed to apply solely to frequency transitions rather than to frequency calculations in schedutil. The last raw required frequency is cached. This allows the driver frequency lookup to be skipped in the event that the new raw required frequency matches the last one, assuming a frequency update has not been forced due to limits changing (indicated by a next_freq value of UINT_MAX, see sugov_should_update_freq). Signed-off-by: Steve Muckle <smuckle@linaro.org> Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-07-13 13:25:26 -07:00
struct cpufreq_policy *policy = sg_policy->policy;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
unsigned int freq = arch_scale_freq_invariant() ?
import G965FXXU7DTAA OSRC *First release for Android (Q). Signed-off-by: FAROVITUS <farovitus@gmail.com>
2020-02-04 13:44:48 +02:00
policy->max : policy->cur;
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
struct sugov_cpu *sg_cpu;
struct kair_class *vessel;
unsigned int delta_max, delta_min;
int util_delta;
unsigned int legacy_freq;
#ifdef KAIR_CLUSTER_TRAVERSING
unsigned int each;
unsigned int sigma_cpu = policy->cpu;
randomness most_rand = 0;
#endif
int cur_rand = KAIR_DIVERGING;
RV_DECLARE(rv);
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
if (sg_policy->tunables->exp_util)
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
freq = (freq + (freq >> 2)) * int_sqrt(util * 100 / max) / 10;
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
else
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
freq = (freq + (freq >> 2)) * util / max;
cpufreq: schedutil: map raw required frequency to driver frequency The slow-path frequency transition path is relatively expensive as it requires waking up a thread to do work. Should support be added for remote CPU cpufreq updates that is also expensive since it requires an IPI. These activities should be avoided if they are not necessary. To that end, calculate the actual driver-supported frequency required by the new utilization value in schedutil by using the recently added cpufreq_driver_resolve_freq API. If it is the same as the previously requested driver frequency then there is no need to continue with the update assuming the cpu frequency limits have not changed. This will have additional benefits should the semantics of the rate limit be changed to apply solely to frequency transitions rather than to frequency calculations in schedutil. The last raw required frequency is cached. This allows the driver frequency lookup to be skipped in the event that the new raw required frequency matches the last one, assuming a frequency update has not been forced due to limits changing (indicated by a next_freq value of UINT_MAX, see sugov_should_update_freq). Signed-off-by: Steve Muckle <smuckle@linaro.org> Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-07-13 13:25:26 -07:00
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
legacy_freq = freq;
if (sg_policy->tunables->fb_legacy)
goto skip_betting;
#ifndef KAIR_CLUSTER_TRAVERSING
sg_cpu = &per_cpu(sugov_cpu, policy->cpu);
vessel = sg_cpu->util_vessel;
if (!vessel)
goto skip_betting;
cur_rand = vessel->job_inferer(vessel);
if (cur_rand == KAIR_DIVERGING)
goto skip_betting;
#else
for_each_cpu(each, policy->cpus) {
sg_cpu = &per_cpu(sugov_cpu, each);
vessel = sg_cpu->util_vessel;
if (vessel) {
cur_rand = vessel->job_inferer(vessel);
if (cur_rand == KAIR_DIVERGING)
goto skip_betting;
else {
if (cur_rand > (int)most_rand) {
most_rand = (randomness)cur_rand;
sigma_cpu = each;
}
}
} else
goto skip_betting;
}
sg_cpu = &per_cpu(sugov_cpu, sigma_cpu);
vessel = sg_cpu->util_vessel;
#endif
util_delta = sg_cpu->util - sg_cpu->cached_util;
delta_max = sg_cpu->max - sg_cpu->cached_util;
delta_min = sg_cpu->cached_util;
RV_SET(rv, util_delta, delta_max, delta_min);
freq = vessel->cap_bettor(vessel, &rv, freq);
skip_betting:
trace_sugov_kair_freq(policy->cpu, util, max, cur_rand, legacy_freq, freq);
#endif
UPSTREAM: cpufreq: schedutil: Don't set next_freq to UINT_MAX The schedutil driver sets sg_policy->next_freq to UINT_MAX on certain occasions to discard the cached value of next freq: - In sugov_start(), when the schedutil governor is started for a group of CPUs. - And whenever we need to force a freq update before rate-limit duration, which happens when: - there is an update in cpufreq policy limits. - Or when the utilization of DL scheduling class increases. In return, get_next_freq() doesn't return a cached next_freq value but recalculates the next frequency instead. But having special meaning for a particular value of frequency makes the code less readable and error prone. We recently fixed a bug where the UINT_MAX value was considered as valid frequency in sugov_update_single(). All we need is a flag which can be used to discard the value of sg_policy->next_freq and we already have need_freq_update for that. Lets reuse it instead of setting next_freq to UINT_MAX. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: joshuous <joshuous@gmail.com> Conflicts: kernel/sched/cpufreq_schedutil.c
2018-05-09 16:05:24 +05:30
if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update)
cpufreq: schedutil: map raw required frequency to driver frequency The slow-path frequency transition path is relatively expensive as it requires waking up a thread to do work. Should support be added for remote CPU cpufreq updates that is also expensive since it requires an IPI. These activities should be avoided if they are not necessary. To that end, calculate the actual driver-supported frequency required by the new utilization value in schedutil by using the recently added cpufreq_driver_resolve_freq API. If it is the same as the previously requested driver frequency then there is no need to continue with the update assuming the cpu frequency limits have not changed. This will have additional benefits should the semantics of the rate limit be changed to apply solely to frequency transitions rather than to frequency calculations in schedutil. The last raw required frequency is cached. This allows the driver frequency lookup to be skipped in the event that the new raw required frequency matches the last one, assuming a frequency update has not been forced due to limits changing (indicated by a next_freq value of UINT_MAX, see sugov_should_update_freq). Signed-off-by: Steve Muckle <smuckle@linaro.org> Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-07-13 13:25:26 -07:00
return sg_policy->next_freq;
UPSTREAM: cpufreq: schedutil: Don't set next_freq to UINT_MAX The schedutil driver sets sg_policy->next_freq to UINT_MAX on certain occasions to discard the cached value of next freq: - In sugov_start(), when the schedutil governor is started for a group of CPUs. - And whenever we need to force a freq update before rate-limit duration, which happens when: - there is an update in cpufreq policy limits. - Or when the utilization of DL scheduling class increases. In return, get_next_freq() doesn't return a cached next_freq value but recalculates the next frequency instead. But having special meaning for a particular value of frequency makes the code less readable and error prone. We recently fixed a bug where the UINT_MAX value was considered as valid frequency in sugov_update_single(). All we need is a flag which can be used to discard the value of sg_policy->next_freq and we already have need_freq_update for that. Lets reuse it instead of setting next_freq to UINT_MAX. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: joshuous <joshuous@gmail.com> Conflicts: kernel/sched/cpufreq_schedutil.c
2018-05-09 16:05:24 +05:30
sg_policy->need_freq_update = false;
cpufreq: schedutil: move cached_raw_freq to struct sugov_policy commit 6c4f0fa643cb9e775dcc976e3db00d649468ff1d upstream. cached_raw_freq applies to the entire cpufreq policy and not individual CPUs. Apart from wasting per-cpu memory, it is actually wrong to keep it in struct sugov_cpu as we may end up comparing next_freq with a stale cached_raw_freq of a random CPU. Move cached_raw_freq to struct sugov_policy. Fixes: 5cbea46984d6 (cpufreq: schedutil: map raw required frequency to driver frequency) Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Amit Pundir <amit.pundir@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-03-02 14:03:20 +05:30
sg_policy->cached_raw_freq = freq;
cpufreq: schedutil: map raw required frequency to driver frequency The slow-path frequency transition path is relatively expensive as it requires waking up a thread to do work. Should support be added for remote CPU cpufreq updates that is also expensive since it requires an IPI. These activities should be avoided if they are not necessary. To that end, calculate the actual driver-supported frequency required by the new utilization value in schedutil by using the recently added cpufreq_driver_resolve_freq API. If it is the same as the previously requested driver frequency then there is no need to continue with the update assuming the cpu frequency limits have not changed. This will have additional benefits should the semantics of the rate limit be changed to apply solely to frequency transitions rather than to frequency calculations in schedutil. The last raw required frequency is cached. This allows the driver frequency lookup to be skipped in the event that the new raw required frequency matches the last one, assuming a frequency update has not been forced due to limits changing (indicated by a next_freq value of UINT_MAX, see sugov_should_update_freq). Signed-off-by: Steve Muckle <smuckle@linaro.org> Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-07-13 13:25:26 -07:00
return cpufreq_driver_resolve_freq(policy, freq);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
ANDROID: cpufreq/schedutil: use boosted_cpu_util for PELT to match WALT When using WALT we always used boosted cpu util for OPP selection. This is the primary purpose for boosted cpu util, but we hadn't changed the PELT utilization check to do the same thing. Fix that here. Change-Id: Id5ffb26eac23b25fe754255221f6d21b8cededfd Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com> Signed-off-by: Chris Redpath <chris.redpath@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-03-24 17:37:28 +00:00
static inline bool use_pelt(void)
{
#ifdef CONFIG_SCHED_WALT
return (!sysctl_sched_use_walt_cpu_util || walt_disabled);
#else
return true;
#endif
}
sched: fair: get rid of frt_boost_threshold * And remove other sched changes / Sync with N770F sched
2023-03-21 20:07:23 +04:00
unsigned int sched_rt_remove_ratio_for_freq = 0;
import G965FXXU7DTAA OSRC *First release for Android (Q). Signed-off-by: FAROVITUS <farovitus@gmail.com>
2020-02-04 13:44:48 +02:00
ANDROID: sched: cpufreq: use rt_avg as estimate of required RT CPU capacity A policy of going to fmax on any RT activity will be detrimental for power on many platforms. Often RT accounts for only a small amount of CPU activity so sending the CPU frequency to fmax is overkill. Worse still, some platforms may not be able to even complete the CPU frequency change before the RT activity has already completed. Cpufreq governors have not treated RT activity this way in the past so it is not part of the expected semantics of the RT scheduling class. The DL class offers guarantees about task completion and could be used for this purpose. Modify the schedutil algorithm to instead use rt_avg as an estimate of RT utilization of the CPU. Based on previous work by Vincent Guittot <vincent.guittot@linaro.org>. Change-Id: I1ed605a3e2512a94d34217a8e57c3fd97cca60be Signed-off-by: Steve Muckle <smuckle@linaro.org> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-08-25 15:59:17 -07:00
static void sugov_get_util(unsigned long *util, unsigned long *max, u64 time)
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
{
ANDROID: sched: cpufreq: use rt_avg as estimate of required RT CPU capacity A policy of going to fmax on any RT activity will be detrimental for power on many platforms. Often RT accounts for only a small amount of CPU activity so sending the CPU frequency to fmax is overkill. Worse still, some platforms may not be able to even complete the CPU frequency change before the RT activity has already completed. Cpufreq governors have not treated RT activity this way in the past so it is not part of the expected semantics of the RT scheduling class. The DL class offers guarantees about task completion and could be used for this purpose. Modify the schedutil algorithm to instead use rt_avg as an estimate of RT utilization of the CPU. Based on previous work by Vincent Guittot <vincent.guittot@linaro.org>. Change-Id: I1ed605a3e2512a94d34217a8e57c3fd97cca60be Signed-off-by: Steve Muckle <smuckle@linaro.org> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-08-25 15:59:17 -07:00
int cpu = smp_processor_id();
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
struct rq *rq = cpu_rq(cpu);
unsigned long max_cap, rt;
s64 delta;
ANDROID: sched: cpufreq: use rt_avg as estimate of required RT CPU capacity A policy of going to fmax on any RT activity will be detrimental for power on many platforms. Often RT accounts for only a small amount of CPU activity so sending the CPU frequency to fmax is overkill. Worse still, some platforms may not be able to even complete the CPU frequency change before the RT activity has already completed. Cpufreq governors have not treated RT activity this way in the past so it is not part of the expected semantics of the RT scheduling class. The DL class offers guarantees about task completion and could be used for this purpose. Modify the schedutil algorithm to instead use rt_avg as an estimate of RT utilization of the CPU. Based on previous work by Vincent Guittot <vincent.guittot@linaro.org>. Change-Id: I1ed605a3e2512a94d34217a8e57c3fd97cca60be Signed-off-by: Steve Muckle <smuckle@linaro.org> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-08-25 15:59:17 -07:00
max_cap = arch_scale_cpu_capacity(NULL, cpu);
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
sched_avg_update(rq);
delta = time - rq->age_stamp;
if (unlikely(delta < 0))
delta = 0;
rt = div64_u64(rq->rt_avg, sched_avg_period() + delta);
rt = (rt * max_cap) >> SCHED_CAPACITY_SHIFT;
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
#ifdef CONFIG_SCHED_EMS
cpufreq_schedutil: account for rt utilizations too
2020-09-15 18:26:14 +03:00
*util = ml_boosted_cpu_util(cpu) + rt;
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
#else
ANDROID: sched/rt: Add schedtune accounting to rt task enqueue/dequeue rt tasks are currently not eligible for schedtune boosting. Make it so by adding enqueue/dequeue hooks. For rt tasks, schedtune only acts as a frequency boosting framework, it has no impact on placement decisions and the prefer_idle attribute is not used. Also prepare schedutil use of boosted util for rt task boosting With this change, schedtune accounting will include rt class tasks, however boosting currently only applies to the utilization provided by fair class tasks. Sum up the tracked CPU utilization applying boost to the aggregate util instead - this includes RT task util in the boosting if any tasks are runnable. Scenario 1, considering one CPU: 1x rt task running, util 250, boost 0 1x cfs task runnable, util 250, boost 50 previous util=250+(50pct_boosted_250) = 887 new util=50_pct_boosted_500 = 762 Scenario 2, considering one CPU: 1x rt task running, util 250, boost 50 1x cfs task runnable, util 250, boost 0 previous util=250+250 = 500 new util=50_pct_boosted_500 = 762 Scenario 3, considering one CPU: 1x rt task running, util 250, boost 50 1x cfs task runnable, util 250, boost 50 previous util=250+(50pct_boosted_250) = 887 new util=50_pct_boosted_500 = 762 Scenario 4: 1x rt task running, util 250, boost 50 previous util=250 = 250 new util=50_pct_boosted_250 = 637 Change-Id: Ie287cbd0692468525095b5024db9faac8b2f4878 Signed-off-by: Chris Redpath <chris.redpath@arm.com>
2018-07-09 16:54:02 +01:00
*util = boosted_cpu_util(cpu, rt);
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
#endif
ANDROID: cpufreq/schedutil: use boosted_cpu_util for PELT to match WALT When using WALT we always used boosted cpu util for OPP selection. This is the primary purpose for boosted cpu util, but we hadn't changed the PELT utilization check to do the same thing. Fix that here. Change-Id: Id5ffb26eac23b25fe754255221f6d21b8cededfd Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com> Signed-off-by: Chris Redpath <chris.redpath@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-03-24 17:37:28 +00:00
if (likely(use_pelt()))
cpufreq: schedutil: clamp util to CPU maximum capacity The code is to get the CPU util by accumulate different scheduling classes and when the total util value is larger than CPU capacity then it clamps util to CPU maximum capacity. So we can get correct util value when use PELT signal but if with WALT signal it misses to clamp util value. On the other hand, WALT doesn't accumulate different class utilization but it needs to applying boost margin for WALT signal the CPU util value is possible to be larger than CPU capacity; so this patch is to always clamp util to CPU maximum capacity. Change-Id: I05481ddbf20246bb9be15b6bd21b6ec039015ea8 Signed-off-by: Leo Yan <leo.yan@linaro.org>
2017-09-15 08:25:32 +08:00
*util = *util + rt;
ANDROID: cpufreq/schedutil: use boosted_cpu_util for PELT to match WALT When using WALT we always used boosted cpu util for OPP selection. This is the primary purpose for boosted cpu util, but we hadn't changed the PELT utilization check to do the same thing. Fix that here. Change-Id: Id5ffb26eac23b25fe754255221f6d21b8cededfd Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com> Signed-off-by: Chris Redpath <chris.redpath@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-03-24 17:37:28 +00:00
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
*util = freqvar_boost_vector(cpu, *util);
cpufreq: schedutil: clamp util to CPU maximum capacity The code is to get the CPU util by accumulate different scheduling classes and when the total util value is larger than CPU capacity then it clamps util to CPU maximum capacity. So we can get correct util value when use PELT signal but if with WALT signal it misses to clamp util value. On the other hand, WALT doesn't accumulate different class utilization but it needs to applying boost margin for WALT signal the CPU util value is possible to be larger than CPU capacity; so this patch is to always clamp util to CPU maximum capacity. Change-Id: I05481ddbf20246bb9be15b6bd21b6ec039015ea8 Signed-off-by: Leo Yan <leo.yan@linaro.org>
2017-09-15 08:25:32 +08:00
*util = min(*util, max_cap);
ANDROID: sched: cpufreq: use rt_avg as estimate of required RT CPU capacity A policy of going to fmax on any RT activity will be detrimental for power on many platforms. Often RT accounts for only a small amount of CPU activity so sending the CPU frequency to fmax is overkill. Worse still, some platforms may not be able to even complete the CPU frequency change before the RT activity has already completed. Cpufreq governors have not treated RT activity this way in the past so it is not part of the expected semantics of the RT scheduling class. The DL class offers guarantees about task completion and could be used for this purpose. Modify the schedutil algorithm to instead use rt_avg as an estimate of RT utilization of the CPU. Based on previous work by Vincent Guittot <vincent.guittot@linaro.org>. Change-Id: I1ed605a3e2512a94d34217a8e57c3fd97cca60be Signed-off-by: Steve Muckle <smuckle@linaro.org> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-08-25 15:59:17 -07:00
*max = max_cap;
sched: ems: Update EMS to latest beyond2lte source Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-07-28 00:51:52 +07:00
#ifdef CONFIG_SCHED_EMS
part_cpu_active_ratio(util, max, cpu);
#endif
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
}
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
static inline void sugov_util_collapse(struct sugov_cpu *sg_cpu)
{
struct kair_class *vessel = sg_cpu->util_vessel;
int util_delta = min(sg_cpu->max, sg_cpu->util) - sg_cpu->cached_util;
unsigned int delta_max = sg_cpu->max - sg_cpu->cached_util;
unsigned int delta_min = sg_cpu->cached_util;
RV_DECLARE(job);
if (vessel) {
RV_SET(job, util_delta, delta_max, delta_min);
vessel->job_learner(vessel, &job);
}
}
#endif
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
static void sugov_set_iowait_boost(struct sugov_cpu *sg_cpu, u64 time,
unsigned int flags)
{
FROMLIST: sched: Make iowait_boost optional in schedutil We should apply the iowait boost only if cpufreq policy has iowait boost enabled. Also make it a schedutil configuration from sysfs so it can be turned on/off if needed (by default initialize it to the policy value). For systems that don't need/want it enabled, such as those on arm64 based mobile devices that are battery operated, it saves energy when the cpufreq driver policy doesn't have it enabled (details below): Here are some results for energy measurements collected running a YouTube video for 30 seconds: Before: 8.042533 mWh After: 7.948377 mWh Energy savings is ~1.2% Bug: 38010527 Link: https://lkml.org/lkml/2017/5/19/42 Change-Id: If124076ad0c16ade369253840dedfbf870aff927 Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-05-18 22:46:10 -07:00
struct sugov_policy *sg_policy = sg_cpu->sg_policy;
if (!sg_policy->tunables->iowait_boost_enable)
return;
cpufreq: schedutil: Fix iowait boost reset A more energy efficient update of the IO wait boosting mechanism has been introduced in: commit a5a0809bc58e ("cpufreq: schedutil: Make iowait boost more energy efficient") where the boost value is expected to be: - doubled at each successive wakeup from IO staring from the minimum frequency supported by a CPU - reset when a CPU is not updated for more then one tick by either disabling the IO wait boost or resetting its value to the minimum frequency if this new update requires an IO boost. This approach is supposed to "ignore" boosting for sporadic wakeups from IO, while still getting the frequency boosted to the maximum to benefit long sequence of wakeup from IO operations. However, these assumptions are not always satisfied. For example, when an IO boosted CPU enters idle for more the one tick and then wakes up after an IO wait, since in sugov_set_iowait_boost() we first check the IOWAIT flag, we keep doubling the iowait boost instead of restarting from the minimum frequency value. This misbehavior could happen mainly on non-shared frequency domains, thus defeating the energy efficiency optimization, but it can also happen on shared frequency domain systems. Let fix this issue in sugov_set_iowait_boost() by: - first check the IO wait boost reset conditions to eventually reset the boost value - then applying the correct IO boost value if required by the caller Fixes: a5a0809bc58e (cpufreq: schedutil: Make iowait boost more energy efficient) Reported-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com> Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2018-05-22 12:07:53 +01:00
if (sg_cpu->iowait_boost) {
s64 delta_ns = time - sg_cpu->last_update;
/* Clear iowait_boost if the CPU apprears to have been idle. */
if (delta_ns > TICK_NSEC) {
sg_cpu->iowait_boost = 0;
sg_cpu->iowait_boost_pending = false;
}
}
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
if (flags & SCHED_CPUFREQ_IOWAIT) {
UPSTREAM: cpufreq: schedutil: Make iowait boost more energy efficient Currently the iowait_boost feature in schedutil makes the frequency go to max on iowait wakeups. This feature was added to handle a case that Peter described where the throughput of operations involving continuous I/O requests [1] is reduced due to running at a lower frequency, however the lower throughput itself causes utilization to be low and hence causing frequency to be low hence its "stuck". Instead of going to max, its also possible to achieve the same effect by ramping up to max if there are repeated in_iowait wakeups happening. This patch is an attempt to do that. We start from a lower frequency (policy->min) and double the boost for every consecutive iowait update until we reach the maximum iowait boost frequency (iowait_boost_max). I ran a synthetic test (continuous O_DIRECT writes in a loop) on an x86 machine with intel_pstate in passive mode using schedutil. In this test the iowait_boost value ramped from 800MHz to 4GHz in 60ms. The patch achieves the desired improved throughput as the existing behavior. [1] https://patchwork.kernel.org/patch/9735885/ Change-Id: I4a018434a50f4ca29ec15b03465f6dc212e54423 Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: Len Brown <lenb@kernel.org> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Suggested-by: Peter Zijlstra <peterz@infradead.org> Suggested-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-08-28 09:56:27 -07:00
if (sg_cpu->iowait_boost_pending)
return;
sg_cpu->iowait_boost_pending = true;
if (sg_cpu->iowait_boost) {
sg_cpu->iowait_boost <<= 1;
if (sg_cpu->iowait_boost > sg_cpu->iowait_boost_max)
sg_cpu->iowait_boost = sg_cpu->iowait_boost_max;
} else {
sg_cpu->iowait_boost = sg_cpu->sg_policy->policy->min;
}
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
}
}
static void sugov_iowait_boost(struct sugov_cpu *sg_cpu, unsigned long *util,
unsigned long *max)
{
UPSTREAM: cpufreq: schedutil: Use unsigned int for iowait boost Make iowait_boost and iowait_boost_max as unsigned int since its unit is kHz and this is consistent with struct cpufreq_policy. Also change the local variables in sugov_iowait_boost to match this. Change-Id: I6c67ed94c57c4bdb24bada4b97045593fcb95d2e Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: Len Brown <lenb@kernel.org> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-07-23 08:54:26 -07:00
unsigned int boost_util, boost_max;
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
UPSTREAM: cpufreq: schedutil: Make iowait boost more energy efficient Currently the iowait_boost feature in schedutil makes the frequency go to max on iowait wakeups. This feature was added to handle a case that Peter described where the throughput of operations involving continuous I/O requests [1] is reduced due to running at a lower frequency, however the lower throughput itself causes utilization to be low and hence causing frequency to be low hence its "stuck". Instead of going to max, its also possible to achieve the same effect by ramping up to max if there are repeated in_iowait wakeups happening. This patch is an attempt to do that. We start from a lower frequency (policy->min) and double the boost for every consecutive iowait update until we reach the maximum iowait boost frequency (iowait_boost_max). I ran a synthetic test (continuous O_DIRECT writes in a loop) on an x86 machine with intel_pstate in passive mode using schedutil. In this test the iowait_boost value ramped from 800MHz to 4GHz in 60ms. The patch achieves the desired improved throughput as the existing behavior. [1] https://patchwork.kernel.org/patch/9735885/ Change-Id: I4a018434a50f4ca29ec15b03465f6dc212e54423 Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: Len Brown <lenb@kernel.org> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Suggested-by: Peter Zijlstra <peterz@infradead.org> Suggested-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-08-28 09:56:27 -07:00
if (!sg_cpu->iowait_boost)
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
return;
UPSTREAM: cpufreq: schedutil: Make iowait boost more energy efficient Currently the iowait_boost feature in schedutil makes the frequency go to max on iowait wakeups. This feature was added to handle a case that Peter described where the throughput of operations involving continuous I/O requests [1] is reduced due to running at a lower frequency, however the lower throughput itself causes utilization to be low and hence causing frequency to be low hence its "stuck". Instead of going to max, its also possible to achieve the same effect by ramping up to max if there are repeated in_iowait wakeups happening. This patch is an attempt to do that. We start from a lower frequency (policy->min) and double the boost for every consecutive iowait update until we reach the maximum iowait boost frequency (iowait_boost_max). I ran a synthetic test (continuous O_DIRECT writes in a loop) on an x86 machine with intel_pstate in passive mode using schedutil. In this test the iowait_boost value ramped from 800MHz to 4GHz in 60ms. The patch achieves the desired improved throughput as the existing behavior. [1] https://patchwork.kernel.org/patch/9735885/ Change-Id: I4a018434a50f4ca29ec15b03465f6dc212e54423 Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: Len Brown <lenb@kernel.org> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Suggested-by: Peter Zijlstra <peterz@infradead.org> Suggested-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-08-28 09:56:27 -07:00
if (sg_cpu->iowait_boost_pending) {
sg_cpu->iowait_boost_pending = false;
} else {
sg_cpu->iowait_boost >>= 1;
if (sg_cpu->iowait_boost < sg_cpu->sg_policy->policy->min) {
sg_cpu->iowait_boost = 0;
return;
}
}
boost_util = sg_cpu->iowait_boost;
boost_max = sg_cpu->iowait_boost_max;
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
if (*util * boost_max < *max * boost_util) {
*util = boost_util;
*max = boost_max;
}
}
UPSTREAM: cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely The way the schedutil governor uses the PELT metric causes it to underestimate the CPU utilization in some cases. That can be easily demonstrated by running kernel compilation on a Sandy Bridge Intel processor, running turbostat in parallel with it and looking at the values written to the MSR_IA32_PERF_CTL register. Namely, the expected result would be that when all CPUs were 100% busy, all of them would be requested to run in the maximum P-state, but observation shows that this clearly isn't the case. The CPUs run in the maximum P-state for a while and then are requested to run slower and go back to the maximum P-state after a while again. That causes the actual frequency of the processor to visibly oscillate below the sustainable maximum in a jittery fashion which clearly is not desirable. That has been attributed to CPU utilization metric updates on task migration that cause the total utilization value for the CPU to be reduced by the utilization of the migrated task. If that happens, the schedutil governor may see a CPU utilization reduction and will attempt to reduce the CPU frequency accordingly right away. That may be premature, though, for example if the system is generally busy and there are other runnable tasks waiting to be run on that CPU already. This is unlikely to be an issue on systems where cpufreq policies are shared between multiple CPUs, because in those cases the policy utilization is computed as the maximum of the CPU utilization values over the whole policy and if that turns out to be low, reducing the frequency for the policy most likely is a good idea anyway. On systems with one CPU per policy, however, it may affect performance adversely and even lead to increased energy consumption in some cases. On those systems it may be addressed by taking another utilization metric into consideration, like whether or not the CPU whose frequency is about to be reduced has been idle recently, because if that's not the case, the CPU is likely to be busy in the near future and its frequency should not be reduced. To that end, use the counter of idle calls in the timekeeping code. Namely, make the schedutil governor look at that counter for the current CPU every time before its frequency is about to be reduced. If the counter has not changed since the previous iteration of the governor computations for that CPU, the CPU has been busy for all that time and its frequency should not be decreased, so if the new frequency would be lower than the one set previously, the governor will skip the frequency update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes <joelaf@google.com> (cherry picked from commit b7eaf1aab9f8bd2e49fceed77ebc66c1b5800718) (simple CPUFREQ_RT_DL vs CPUFREQ_DL usage conflicts) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: I531ec02c052944ee07a904dc2a25c59948ee762b Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:24:58 +01:00
#ifdef CONFIG_NO_HZ_COMMON
static bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu)
{
unsigned long idle_calls = tick_nohz_get_idle_calls();
bool ret = idle_calls == sg_cpu->saved_idle_calls;
sg_cpu->saved_idle_calls = idle_calls;
return ret;
}
#else
static inline bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu) { return false; }
#endif /* CONFIG_NO_HZ_COMMON */
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
static void sugov_update_single(struct update_util_data *hook, u64 time,
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
unsigned int flags)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
struct sugov_cpu *sg_cpu = container_of(hook, struct sugov_cpu, update_util);
struct sugov_policy *sg_policy = sg_cpu->sg_policy;
struct cpufreq_policy *policy = sg_policy->policy;
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
unsigned long util, max;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
unsigned int next_f;
UPSTREAM: cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely The way the schedutil governor uses the PELT metric causes it to underestimate the CPU utilization in some cases. That can be easily demonstrated by running kernel compilation on a Sandy Bridge Intel processor, running turbostat in parallel with it and looking at the values written to the MSR_IA32_PERF_CTL register. Namely, the expected result would be that when all CPUs were 100% busy, all of them would be requested to run in the maximum P-state, but observation shows that this clearly isn't the case. The CPUs run in the maximum P-state for a while and then are requested to run slower and go back to the maximum P-state after a while again. That causes the actual frequency of the processor to visibly oscillate below the sustainable maximum in a jittery fashion which clearly is not desirable. That has been attributed to CPU utilization metric updates on task migration that cause the total utilization value for the CPU to be reduced by the utilization of the migrated task. If that happens, the schedutil governor may see a CPU utilization reduction and will attempt to reduce the CPU frequency accordingly right away. That may be premature, though, for example if the system is generally busy and there are other runnable tasks waiting to be run on that CPU already. This is unlikely to be an issue on systems where cpufreq policies are shared between multiple CPUs, because in those cases the policy utilization is computed as the maximum of the CPU utilization values over the whole policy and if that turns out to be low, reducing the frequency for the policy most likely is a good idea anyway. On systems with one CPU per policy, however, it may affect performance adversely and even lead to increased energy consumption in some cases. On those systems it may be addressed by taking another utilization metric into consideration, like whether or not the CPU whose frequency is about to be reduced has been idle recently, because if that's not the case, the CPU is likely to be busy in the near future and its frequency should not be reduced. To that end, use the counter of idle calls in the timekeeping code. Namely, make the schedutil governor look at that counter for the current CPU every time before its frequency is about to be reduced. If the counter has not changed since the previous iteration of the governor computations for that CPU, the CPU has been busy for all that time and its frequency should not be decreased, so if the new frequency would be lower than the one set previously, the governor will skip the frequency update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes <joelaf@google.com> (cherry picked from commit b7eaf1aab9f8bd2e49fceed77ebc66c1b5800718) (simple CPUFREQ_RT_DL vs CPUFREQ_DL usage conflicts) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: I531ec02c052944ee07a904dc2a25c59948ee762b Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:24:58 +01:00
bool busy;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
sugov_set_iowait_boost(sg_cpu, time, flags);
sg_cpu->last_update = time;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
if (!sugov_should_update_freq(sg_policy, time))
return;
cpufreq: schedutil: Don't skip freq update when limits change To avoid reducing the frequency of a CPU prematurely, we skip reducing the frequency if the CPU had been busy recently. This should not be done when the limits of the policy are changed, for example due to thermal throttling. We should always get the frequency within the new limits as soon as possible. Trying to fix this by using only one flag, i.e. need_freq_update, can lead to a race condition where the flag gets cleared without forcing us to change the frequency at least once. And so this patch introduces another flag to avoid that race condition. Fixes: ecd288429126 ("cpufreq: schedutil: Don't set next_freq to UINT_MAX") Cc: v4.18+ <stable@vger.kernel.org> # v4.18+ Reported-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-08-07 12:36:01 +05:30
/* Limits may have changed, don't skip frequency update */
busy = !sg_policy->need_freq_update && sugov_cpu_is_busy(sg_cpu);
UPSTREAM: cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely The way the schedutil governor uses the PELT metric causes it to underestimate the CPU utilization in some cases. That can be easily demonstrated by running kernel compilation on a Sandy Bridge Intel processor, running turbostat in parallel with it and looking at the values written to the MSR_IA32_PERF_CTL register. Namely, the expected result would be that when all CPUs were 100% busy, all of them would be requested to run in the maximum P-state, but observation shows that this clearly isn't the case. The CPUs run in the maximum P-state for a while and then are requested to run slower and go back to the maximum P-state after a while again. That causes the actual frequency of the processor to visibly oscillate below the sustainable maximum in a jittery fashion which clearly is not desirable. That has been attributed to CPU utilization metric updates on task migration that cause the total utilization value for the CPU to be reduced by the utilization of the migrated task. If that happens, the schedutil governor may see a CPU utilization reduction and will attempt to reduce the CPU frequency accordingly right away. That may be premature, though, for example if the system is generally busy and there are other runnable tasks waiting to be run on that CPU already. This is unlikely to be an issue on systems where cpufreq policies are shared between multiple CPUs, because in those cases the policy utilization is computed as the maximum of the CPU utilization values over the whole policy and if that turns out to be low, reducing the frequency for the policy most likely is a good idea anyway. On systems with one CPU per policy, however, it may affect performance adversely and even lead to increased energy consumption in some cases. On those systems it may be addressed by taking another utilization metric into consideration, like whether or not the CPU whose frequency is about to be reduced has been idle recently, because if that's not the case, the CPU is likely to be busy in the near future and its frequency should not be reduced. To that end, use the counter of idle calls in the timekeeping code. Namely, make the schedutil governor look at that counter for the current CPU every time before its frequency is about to be reduced. If the counter has not changed since the previous iteration of the governor computations for that CPU, the CPU has been busy for all that time and its frequency should not be decreased, so if the new frequency would be lower than the one set previously, the governor will skip the frequency update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes <joelaf@google.com> (cherry picked from commit b7eaf1aab9f8bd2e49fceed77ebc66c1b5800718) (simple CPUFREQ_RT_DL vs CPUFREQ_DL usage conflicts) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: I531ec02c052944ee07a904dc2a25c59948ee762b Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:24:58 +01:00
ANDROID: sched: cpufreq: use rt_avg as estimate of required RT CPU capacity A policy of going to fmax on any RT activity will be detrimental for power on many platforms. Often RT accounts for only a small amount of CPU activity so sending the CPU frequency to fmax is overkill. Worse still, some platforms may not be able to even complete the CPU frequency change before the RT activity has already completed. Cpufreq governors have not treated RT activity this way in the past so it is not part of the expected semantics of the RT scheduling class. The DL class offers guarantees about task completion and could be used for this purpose. Modify the schedutil algorithm to instead use rt_avg as an estimate of RT utilization of the CPU. Based on previous work by Vincent Guittot <vincent.guittot@linaro.org>. Change-Id: I1ed605a3e2512a94d34217a8e57c3fd97cca60be Signed-off-by: Steve Muckle <smuckle@linaro.org> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-08-25 15:59:17 -07:00
if (flags & SCHED_CPUFREQ_DL) {
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
next_f = policy->cpuinfo.max_freq;
} else {
ANDROID: sched: cpufreq: use rt_avg as estimate of required RT CPU capacity A policy of going to fmax on any RT activity will be detrimental for power on many platforms. Often RT accounts for only a small amount of CPU activity so sending the CPU frequency to fmax is overkill. Worse still, some platforms may not be able to even complete the CPU frequency change before the RT activity has already completed. Cpufreq governors have not treated RT activity this way in the past so it is not part of the expected semantics of the RT scheduling class. The DL class offers guarantees about task completion and could be used for this purpose. Modify the schedutil algorithm to instead use rt_avg as an estimate of RT utilization of the CPU. Based on previous work by Vincent Guittot <vincent.guittot@linaro.org>. Change-Id: I1ed605a3e2512a94d34217a8e57c3fd97cca60be Signed-off-by: Steve Muckle <smuckle@linaro.org> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-08-25 15:59:17 -07:00
sugov_get_util(&util, &max, time);
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
sugov_iowait_boost(sg_cpu, &util, &max);
UPSTREAM: cpufreq: schedutil: Pass sg_policy to get_next_freq() get_next_freq() uses sg_cpu only to get sg_policy, which the callers of get_next_freq() already have. Pass sg_policy instead of sg_cpu to get_next_freq(), to make it more efficient. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 655cb1ebff4b7918fc560502c3297af2d3c7d114) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ia210058da32930a6cdb18258aa679cd1a44a747e Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-03-02 14:03:21 +05:30
next_f = get_next_freq(sg_policy, util, max);
UPSTREAM: cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely The way the schedutil governor uses the PELT metric causes it to underestimate the CPU utilization in some cases. That can be easily demonstrated by running kernel compilation on a Sandy Bridge Intel processor, running turbostat in parallel with it and looking at the values written to the MSR_IA32_PERF_CTL register. Namely, the expected result would be that when all CPUs were 100% busy, all of them would be requested to run in the maximum P-state, but observation shows that this clearly isn't the case. The CPUs run in the maximum P-state for a while and then are requested to run slower and go back to the maximum P-state after a while again. That causes the actual frequency of the processor to visibly oscillate below the sustainable maximum in a jittery fashion which clearly is not desirable. That has been attributed to CPU utilization metric updates on task migration that cause the total utilization value for the CPU to be reduced by the utilization of the migrated task. If that happens, the schedutil governor may see a CPU utilization reduction and will attempt to reduce the CPU frequency accordingly right away. That may be premature, though, for example if the system is generally busy and there are other runnable tasks waiting to be run on that CPU already. This is unlikely to be an issue on systems where cpufreq policies are shared between multiple CPUs, because in those cases the policy utilization is computed as the maximum of the CPU utilization values over the whole policy and if that turns out to be low, reducing the frequency for the policy most likely is a good idea anyway. On systems with one CPU per policy, however, it may affect performance adversely and even lead to increased energy consumption in some cases. On those systems it may be addressed by taking another utilization metric into consideration, like whether or not the CPU whose frequency is about to be reduced has been idle recently, because if that's not the case, the CPU is likely to be busy in the near future and its frequency should not be reduced. To that end, use the counter of idle calls in the timekeeping code. Namely, make the schedutil governor look at that counter for the current CPU every time before its frequency is about to be reduced. If the counter has not changed since the previous iteration of the governor computations for that CPU, the CPU has been busy for all that time and its frequency should not be decreased, so if the new frequency would be lower than the one set previously, the governor will skip the frequency update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes <joelaf@google.com> (cherry picked from commit b7eaf1aab9f8bd2e49fceed77ebc66c1b5800718) (simple CPUFREQ_RT_DL vs CPUFREQ_DL usage conflicts) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: I531ec02c052944ee07a904dc2a25c59948ee762b Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:24:58 +01:00
/*
* Do not reduce the frequency if the CPU has not been idle
* recently, as the reduction is likely to be premature then.
*/
cpufreq: schedutil: Avoid using invalid next_freq. commit 97739501f207efe33145b918817f305b822987f8 upstream. If the next_freq field of struct sugov_policy is set to UINT_MAX, it shouldn't be used for updating the CPU frequency (this is a special "invalid" value), but after commit b7eaf1aab9f8 (cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely) it may be passed as the new frequency to sugov_update_commit() in sugov_update_single(). Fix that by adding an extra check for the special UINT_MAX value of next_freq to sugov_update_single(). Change-Id: I516963d8b0fda677e9d82290c0f3d4dc3cb6e477 Fixes: b7eaf1aab9f8 (cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely) Reported-by: Viresh Kumar <viresh.kumar@linaro.org> Cc: 4.12+ <stable@vger.kernel.org> # 4.12+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-05-09 11:44:56 +02:00
if (busy && next_f < sg_policy->next_freq &&
sg_policy->next_freq != UINT_MAX) {
UPSTREAM: cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely The way the schedutil governor uses the PELT metric causes it to underestimate the CPU utilization in some cases. That can be easily demonstrated by running kernel compilation on a Sandy Bridge Intel processor, running turbostat in parallel with it and looking at the values written to the MSR_IA32_PERF_CTL register. Namely, the expected result would be that when all CPUs were 100% busy, all of them would be requested to run in the maximum P-state, but observation shows that this clearly isn't the case. The CPUs run in the maximum P-state for a while and then are requested to run slower and go back to the maximum P-state after a while again. That causes the actual frequency of the processor to visibly oscillate below the sustainable maximum in a jittery fashion which clearly is not desirable. That has been attributed to CPU utilization metric updates on task migration that cause the total utilization value for the CPU to be reduced by the utilization of the migrated task. If that happens, the schedutil governor may see a CPU utilization reduction and will attempt to reduce the CPU frequency accordingly right away. That may be premature, though, for example if the system is generally busy and there are other runnable tasks waiting to be run on that CPU already. This is unlikely to be an issue on systems where cpufreq policies are shared between multiple CPUs, because in those cases the policy utilization is computed as the maximum of the CPU utilization values over the whole policy and if that turns out to be low, reducing the frequency for the policy most likely is a good idea anyway. On systems with one CPU per policy, however, it may affect performance adversely and even lead to increased energy consumption in some cases. On those systems it may be addressed by taking another utilization metric into consideration, like whether or not the CPU whose frequency is about to be reduced has been idle recently, because if that's not the case, the CPU is likely to be busy in the near future and its frequency should not be reduced. To that end, use the counter of idle calls in the timekeeping code. Namely, make the schedutil governor look at that counter for the current CPU every time before its frequency is about to be reduced. If the counter has not changed since the previous iteration of the governor computations for that CPU, the CPU has been busy for all that time and its frequency should not be decreased, so if the new frequency would be lower than the one set previously, the governor will skip the frequency update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes <joelaf@google.com> (cherry picked from commit b7eaf1aab9f8bd2e49fceed77ebc66c1b5800718) (simple CPUFREQ_RT_DL vs CPUFREQ_DL usage conflicts) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: I531ec02c052944ee07a904dc2a25c59948ee762b Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:24:58 +01:00
next_f = sg_policy->next_freq;
BACKPORT: schedutil: Reset cached freq if it is not in sync with next_freq 'cached_raw_freq' is used to get the next frequency quickly but should always be in sync with sg_policy->next_freq. There are cases where it is not and in such cases it should be reset to avoid switching to incorrect frequencies. Consider this case for example: - policy->cur is 1.2 GHz (Max) - New request comes for 780 MHz and we store that in cached_raw_freq. - Based on 780 MHz, we calculate the effective frequency as 800 MHz. - We then decide not to update the frequency as sugov_up_down_rate_limit() return true. - Here cached_raw_freq is 780 MHz and sg_policy->next_freq is 1.2 GHz. - Now if the utilization doesn't change in next request, then the next target frequency will still be 780 MHz and it will match with cached_raw_freq and so we will directly return 1.2 GHz instead of 800 MHz. BACKPORT of upstream commit 07458f6a5171 ("cpufreq: schedutil: Reset cached_raw_freq when not in sync with next_freq"). This also updates sugov_update_commit() for handling up/down tunables, which aren't present in mainline. Change-Id: Ie86465231e7cb265e5b4c26f59d6faf8d9630b0a Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
2017-11-08 19:47:36 +05:30
/* Reset cached freq as next_freq has changed */
sg_policy->cached_raw_freq = 0;
}
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sugov_update_commit(sg_policy, time, next_f);
}
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
static unsigned int sugov_next_freq_shared(struct sugov_cpu *sg_cpu, u64 time)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
cpufreq: schedutil: map raw required frequency to driver frequency The slow-path frequency transition path is relatively expensive as it requires waking up a thread to do work. Should support be added for remote CPU cpufreq updates that is also expensive since it requires an IPI. These activities should be avoided if they are not necessary. To that end, calculate the actual driver-supported frequency required by the new utilization value in schedutil by using the recently added cpufreq_driver_resolve_freq API. If it is the same as the previously requested driver frequency then there is no need to continue with the update assuming the cpu frequency limits have not changed. This will have additional benefits should the semantics of the rate limit be changed to apply solely to frequency transitions rather than to frequency calculations in schedutil. The last raw required frequency is cached. This allows the driver frequency lookup to be skipped in the event that the new raw required frequency matches the last one, assuming a frequency update has not been forced due to limits changing (indicated by a next_freq value of UINT_MAX, see sugov_should_update_freq). Signed-off-by: Steve Muckle <smuckle@linaro.org> Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-07-13 13:25:26 -07:00
struct sugov_policy *sg_policy = sg_cpu->sg_policy;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
struct cpufreq_policy *policy = sg_policy->policy;
UPSTREAM: cpufreq: schedutil: Refactor sugov_next_freq_shared() The loop in sugov_next_freq_shared() contains an if block to skip the loop for the current CPU. This turns out to be an unnecessary conditional in the scheduler's hot-path for every CPU in the policy. It would be better to drop the conditional and make the loop treat all the CPUs in the same way. That would eliminate the need of calling sugov_iowait_boost() at the top of the routine. To keep the code optimized to return early if the current CPU has RT/DL flags set, move the flags check to sugov_update_shared() instead in order to avoid the function call entirely. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit cba1dfb57b94c234728b689d9b00d4267fa1a879) (modified for SCHED_CPUFREQ_DL vs SCHED_CPUFREQ_RT) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ie046fdc8eda46821356750edd0fb6f7d077af363 Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:22:59 +01:00
unsigned long util = 0, max = 1;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
unsigned int j;
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
for_each_cpu(j, policy->cpus) {
UPSTREAM: cpufreq: schedutil: Refactor sugov_next_freq_shared() The loop in sugov_next_freq_shared() contains an if block to skip the loop for the current CPU. This turns out to be an unnecessary conditional in the scheduler's hot-path for every CPU in the policy. It would be better to drop the conditional and make the loop treat all the CPUs in the same way. That would eliminate the need of calling sugov_iowait_boost() at the top of the routine. To keep the code optimized to return early if the current CPU has RT/DL flags set, move the flags check to sugov_update_shared() instead in order to avoid the function call entirely. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit cba1dfb57b94c234728b689d9b00d4267fa1a879) (modified for SCHED_CPUFREQ_DL vs SCHED_CPUFREQ_RT) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ie046fdc8eda46821356750edd0fb6f7d077af363 Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:22:59 +01:00
struct sugov_cpu *j_sg_cpu = &per_cpu(sugov_cpu, j);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
unsigned long j_util, j_max;
s64 delta_ns;
/*
* If the CPU utilization was last updated before the previous
* frequency update and the time elapsed between the last update
* of the CPU utilization and the last frequency update is long
* enough, don't take the CPU into account as it probably is
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
* idle now (and clear iowait_boost for it).
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
*/
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
delta_ns = time - j_sg_cpu->last_update;
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
if (delta_ns > TICK_NSEC) {
j_sg_cpu->iowait_boost = 0;
UPSTREAM: cpufreq: schedutil: Make iowait boost more energy efficient Currently the iowait_boost feature in schedutil makes the frequency go to max on iowait wakeups. This feature was added to handle a case that Peter described where the throughput of operations involving continuous I/O requests [1] is reduced due to running at a lower frequency, however the lower throughput itself causes utilization to be low and hence causing frequency to be low hence its "stuck". Instead of going to max, its also possible to achieve the same effect by ramping up to max if there are repeated in_iowait wakeups happening. This patch is an attempt to do that. We start from a lower frequency (policy->min) and double the boost for every consecutive iowait update until we reach the maximum iowait boost frequency (iowait_boost_max). I ran a synthetic test (continuous O_DIRECT writes in a loop) on an x86 machine with intel_pstate in passive mode using schedutil. In this test the iowait_boost value ramped from 800MHz to 4GHz in 60ms. The patch achieves the desired improved throughput as the existing behavior. [1] https://patchwork.kernel.org/patch/9735885/ Change-Id: I4a018434a50f4ca29ec15b03465f6dc212e54423 Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: Len Brown <lenb@kernel.org> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Suggested-by: Peter Zijlstra <peterz@infradead.org> Suggested-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-08-28 09:56:27 -07:00
j_sg_cpu->iowait_boost_pending = false;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
continue;
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
}
ANDROID: sched: cpufreq: use rt_avg as estimate of required RT CPU capacity A policy of going to fmax on any RT activity will be detrimental for power on many platforms. Often RT accounts for only a small amount of CPU activity so sending the CPU frequency to fmax is overkill. Worse still, some platforms may not be able to even complete the CPU frequency change before the RT activity has already completed. Cpufreq governors have not treated RT activity this way in the past so it is not part of the expected semantics of the RT scheduling class. The DL class offers guarantees about task completion and could be used for this purpose. Modify the schedutil algorithm to instead use rt_avg as an estimate of RT utilization of the CPU. Based on previous work by Vincent Guittot <vincent.guittot@linaro.org>. Change-Id: I1ed605a3e2512a94d34217a8e57c3fd97cca60be Signed-off-by: Steve Muckle <smuckle@linaro.org> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-08-25 15:59:17 -07:00
if (j_sg_cpu->flags & SCHED_CPUFREQ_DL)
UPSTREAM: cpufreq: schedutil: Refactor sugov_next_freq_shared() The loop in sugov_next_freq_shared() contains an if block to skip the loop for the current CPU. This turns out to be an unnecessary conditional in the scheduler's hot-path for every CPU in the policy. It would be better to drop the conditional and make the loop treat all the CPUs in the same way. That would eliminate the need of calling sugov_iowait_boost() at the top of the routine. To keep the code optimized to return early if the current CPU has RT/DL flags set, move the flags check to sugov_update_shared() instead in order to avoid the function call entirely. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit cba1dfb57b94c234728b689d9b00d4267fa1a879) (modified for SCHED_CPUFREQ_DL vs SCHED_CPUFREQ_RT) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ie046fdc8eda46821356750edd0fb6f7d077af363 Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:22:59 +01:00
return policy->cpuinfo.max_freq;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
j_util = j_sg_cpu->util;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
j_max = j_sg_cpu->max;
cpufreq: schedutil: Use >= when aggregating CPU loads in a policy When the normal utilization value for all the CPUs in a policy is 0, the current aggregation scheme of using a > check will result in the aggregated utilization being 0 and the max value being 1. This is not a problem for upstream code. However, since we also use other statistics provided by WALT to update the aggregated utilization value across all CPUs in a policy, we can end up with a non-zero aggregated utilization while the max remains as 1. Then when get_next_freq() tries to compute the frequency using: max-frequency * 1.25 * (util / max) it ends up with a frequency that is greater than max-frequency. So the policy frequency jumps to max-frequency. By changing the aggregation check to >=, we make sure that the max value is updated with something reported by the scheduler for a CPU in that policy. With the correct max, we can continue using the WALT specific statistics without spurious jumps to max-frequency. Change-Id: I14996cd796191192ea112f949dc42450782105f7 Signed-off-by: Saravana Kannan <skannan@codeaurora.org>
2017-10-12 15:56:12 -07:00
if (j_util * max >= j_max * util) {
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
util = j_util;
max = j_max;
}
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
sugov_iowait_boost(j_sg_cpu, &util, &max);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
UPSTREAM: cpufreq: schedutil: Pass sg_policy to get_next_freq() get_next_freq() uses sg_cpu only to get sg_policy, which the callers of get_next_freq() already have. Pass sg_policy instead of sg_cpu to get_next_freq(), to make it more efficient. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 655cb1ebff4b7918fc560502c3297af2d3c7d114) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ia210058da32930a6cdb18258aa679cd1a44a747e Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-03-02 14:03:21 +05:30
return get_next_freq(sg_policy, util, max);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
static void sugov_update_shared(struct update_util_data *hook, u64 time,
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
unsigned int flags)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
struct sugov_cpu *sg_cpu = container_of(hook, struct sugov_cpu, update_util);
struct sugov_policy *sg_policy = sg_cpu->sg_policy;
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
unsigned long util, max;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
unsigned int next_f;
ANDROID: sched: cpufreq: use rt_avg as estimate of required RT CPU capacity A policy of going to fmax on any RT activity will be detrimental for power on many platforms. Often RT accounts for only a small amount of CPU activity so sending the CPU frequency to fmax is overkill. Worse still, some platforms may not be able to even complete the CPU frequency change before the RT activity has already completed. Cpufreq governors have not treated RT activity this way in the past so it is not part of the expected semantics of the RT scheduling class. The DL class offers guarantees about task completion and could be used for this purpose. Modify the schedutil algorithm to instead use rt_avg as an estimate of RT utilization of the CPU. Based on previous work by Vincent Guittot <vincent.guittot@linaro.org>. Change-Id: I1ed605a3e2512a94d34217a8e57c3fd97cca60be Signed-off-by: Steve Muckle <smuckle@linaro.org> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-08-25 15:59:17 -07:00
sugov_get_util(&util, &max, time);
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
raw_spin_lock(&sg_policy->update_lock);
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
sg_cpu->cached_util = min(max, sg_cpu->max ?
mult_frac(sg_cpu->util, max, sg_cpu->max) : sg_cpu->util);
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sg_cpu->util = util;
sg_cpu->max = max;
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
sg_cpu->flags = flags;
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
sugov_util_collapse(sg_cpu);
#endif
cpufreq: schedutil: Add iowait boosting Modify the schedutil cpufreq governor to boost the CPU frequency if the SCHED_CPUFREQ_IOWAIT flag is passed to it via cpufreq_update_util(). If that happens, the frequency is set to the maximum during the first update after receiving the SCHED_CPUFREQ_IOWAIT flag and then the boost is reduced by half during each following update. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Looks-good-to: Steve Muckle <smuckle@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-09-10 00:00:31 +02:00
sugov_set_iowait_boost(sg_cpu, time, flags);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sg_cpu->last_update = time;
if (sugov_should_update_freq(sg_policy, time)) {
UPSTREAM: cpufreq: schedutil: Refactor sugov_next_freq_shared() The loop in sugov_next_freq_shared() contains an if block to skip the loop for the current CPU. This turns out to be an unnecessary conditional in the scheduler's hot-path for every CPU in the policy. It would be better to drop the conditional and make the loop treat all the CPUs in the same way. That would eliminate the need of calling sugov_iowait_boost() at the top of the routine. To keep the code optimized to return early if the current CPU has RT/DL flags set, move the flags check to sugov_update_shared() instead in order to avoid the function call entirely. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit cba1dfb57b94c234728b689d9b00d4267fa1a879) (modified for SCHED_CPUFREQ_DL vs SCHED_CPUFREQ_RT) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ie046fdc8eda46821356750edd0fb6f7d077af363 Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:22:59 +01:00
if (flags & SCHED_CPUFREQ_DL)
next_f = sg_policy->policy->cpuinfo.max_freq;
else
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
next_f = sugov_next_freq_shared(sg_cpu, time);
UPSTREAM: cpufreq: schedutil: Refactor sugov_next_freq_shared() The loop in sugov_next_freq_shared() contains an if block to skip the loop for the current CPU. This turns out to be an unnecessary conditional in the scheduler's hot-path for every CPU in the policy. It would be better to drop the conditional and make the loop treat all the CPUs in the same way. That would eliminate the need of calling sugov_iowait_boost() at the top of the routine. To keep the code optimized to return early if the current CPU has RT/DL flags set, move the flags check to sugov_update_shared() instead in order to avoid the function call entirely. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit cba1dfb57b94c234728b689d9b00d4267fa1a879) (modified for SCHED_CPUFREQ_DL vs SCHED_CPUFREQ_RT) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ie046fdc8eda46821356750edd0fb6f7d077af363 Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:22:59 +01:00
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sugov_update_commit(sg_policy, time, next_f);
}
raw_spin_unlock(&sg_policy->update_lock);
}
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
static void sugov_work(struct kthread_work *work)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
struct sugov_policy *sg_policy = container_of(work, struct sugov_policy, work);
schedutil: Allow cpufreq requests to be made even when kthread kicked Currently there is a chance of a schedutil cpufreq update request to be dropped if there is a pending update request. This pending request can be delayed if there is a scheduling delay of the irq_work and the wake up of the schedutil governor kthread. A very bad scenario is when a schedutil request was already just made, such as to reduce the CPU frequency, then a newer request to increase CPU frequency (even sched deadline urgent frequency increase requests) can be dropped, even though the rate limits suggest that its Ok to process a request. This is because of the way the work_in_progress flag is used. This patch improves the situation by allowing new requests to happen even though the old one is still being processed. Note that in this approach, if an irq_work was already issued, we just update next_freq and don't bother to queue another request so there's no extra work being done to make this happen. Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Juri Lelli <juri.lelli@redhat.com> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2018-05-22 15:55:53 -07:00
unsigned int freq;
unsigned long flags;
/*
* Hold sg_policy->update_lock shortly to handle the case where:
* incase sg_policy->next_freq is read here, and then updated by
* sugov_update_shared just before work_in_progress is set to false
* here, we may miss queueing the new update.
*
* Note: If a work was queued after the update_lock is released,
* sugov_work will just be called again by kthread_work code; and the
* request will be proceed before the sugov thread sleeps.
*/
raw_spin_lock_irqsave(&sg_policy->update_lock, flags);
freq = sg_policy->next_freq;
sg_policy->work_in_progress = false;
raw_spin_unlock_irqrestore(&sg_policy->update_lock, flags);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
mutex_lock(&sg_policy->work_lock);
schedutil: Allow cpufreq requests to be made even when kthread kicked Currently there is a chance of a schedutil cpufreq update request to be dropped if there is a pending update request. This pending request can be delayed if there is a scheduling delay of the irq_work and the wake up of the schedutil governor kthread. A very bad scenario is when a schedutil request was already just made, such as to reduce the CPU frequency, then a newer request to increase CPU frequency (even sched deadline urgent frequency increase requests) can be dropped, even though the rate limits suggest that its Ok to process a request. This is because of the way the work_in_progress flag is used. This patch improves the situation by allowing new requests to happen even though the old one is still being processed. Note that in this approach, if an irq_work was already issued, we just update next_freq and don't bother to queue another request so there's no extra work being done to make this happen. Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Juri Lelli <juri.lelli@redhat.com> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2018-05-22 15:55:53 -07:00
__cpufreq_driver_target(sg_policy->policy, freq, CPUFREQ_RELATION_L);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
mutex_unlock(&sg_policy->work_lock);
}
static void sugov_irq_work(struct irq_work *irq_work)
{
struct sugov_policy *sg_policy;
sg_policy = container_of(irq_work, struct sugov_policy, irq_work);
UPSTREAM: cpufreq: schedutil: move slow path from workqueue to SCHED_FIFO task If slow path frequency changes are conducted in a SCHED_OTHER context then they may be delayed for some amount of time, including indefinitely, when real time or deadline activity is taking place. Move the slow path to a real time kernel thread. In the future the thread should be made SCHED_DEADLINE. The RT priority is arbitrarily set to 50 for now. Hackbench results on ARM Exynos, dual core A15 platform for 10 iterations: $ hackbench -s 100 -l 100 -g 10 -f 20 Before After --------------------------------- 1.808 1.603 1.847 1.251 2.229 1.590 1.952 1.600 1.947 1.257 1.925 1.627 2.694 1.620 1.258 1.621 1.919 1.632 1.250 1.240 Average: 1.8829 1.5041 Based on initial work by Steve Muckle. Change-Id: I8bb111435b0871e961a6f26db1151c91624817b8 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry-picked from commit 02a7b1ee3baa15a98b541d8cfd156bbe1a091c20) [fixed cherry-pick issue] Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-15 13:53:22 +05:30
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
/*
* For RT and deadline tasks, the schedutil governor shoots the
* frequency to maximum. Special care must be taken to ensure that this
* kthread doesn't result in the same behavior.
*
* This is (mostly) guaranteed by the work_in_progress flag. The flag is
* updated only at the end of the sugov_work() function and before that
* the schedutil governor rejects all other frequency scaling requests.
*
* There is a very rare case though, where the RT thread yields right
* after the work_in_progress flag is cleared. The effects of that are
* neglected for now.
*/
kthread_queue_work(&sg_policy->worker, &sg_policy->work);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
[COMMON] sched: support freqvar tuner Change-Id: Ifd2fb13a6ab44f3137aca816c8537a85af0e7850 Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2018-01-19 16:42:10 +09:00
/************************ Governor externals ***********************/
static void update_min_rate_limit_us(struct sugov_policy *sg_policy);
void sugov_update_rate_limit_us(struct cpufreq_policy *policy,
int up_rate_limit, int down_rate_limit)
{
struct sugov_policy *sg_policy;
struct sugov_tunables *tunables;
sg_policy = policy->governor_data;
if (!sg_policy)
return;
tunables = sg_policy->tunables;
if (!tunables)
return;
tunables->up_rate_limit_us = (unsigned int)up_rate_limit;
tunables->down_rate_limit_us = (unsigned int)down_rate_limit;
sg_policy->up_rate_delay_ns = up_rate_limit * NSEC_PER_USEC;
sg_policy->down_rate_delay_ns = down_rate_limit * NSEC_PER_USEC;
update_min_rate_limit_us(sg_policy);
}
int sugov_sysfs_add_attr(struct cpufreq_policy *policy, const struct attribute *attr)
{
struct sugov_policy *sg_policy;
struct sugov_tunables *tunables;
sg_policy = policy->governor_data;
if (!sg_policy)
return -ENODEV;
tunables = sg_policy->tunables;
if (!tunables)
return -ENODEV;
return sysfs_create_file(&tunables->attr_set.kobj, attr);
}
struct cpufreq_policy *sugov_get_attr_policy(struct gov_attr_set *attr_set)
{
struct sugov_policy *sg_policy = list_first_entry(&attr_set->policy_list,
typeof(*sg_policy), tunables_hook);
return sg_policy->policy;
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
/************************** sysfs interface ************************/
static struct sugov_tunables *global_tunables;
static DEFINE_MUTEX(global_tunables_lock);
static inline struct sugov_tunables *to_sugov_tunables(struct gov_attr_set *attr_set)
{
return container_of(attr_set, struct sugov_tunables, attr_set);
}
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
static DEFINE_MUTEX(min_rate_lock);
static void update_min_rate_limit_us(struct sugov_policy *sg_policy)
{
mutex_lock(&min_rate_lock);
sg_policy->min_rate_limit_ns = min(sg_policy->up_rate_delay_ns,
sg_policy->down_rate_delay_ns);
mutex_unlock(&min_rate_lock);
}
static ssize_t up_rate_limit_us_show(struct gov_attr_set *attr_set, char *buf)
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
return sprintf(buf, "%u\n", tunables->up_rate_limit_us);
}
static ssize_t down_rate_limit_us_show(struct gov_attr_set *attr_set, char *buf)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
return sprintf(buf, "%u\n", tunables->down_rate_limit_us);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
static ssize_t up_rate_limit_us_store(struct gov_attr_set *attr_set,
const char *buf, size_t count)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
struct sugov_policy *sg_policy;
unsigned int rate_limit_us;
if (kstrtouint(buf, 10, &rate_limit_us))
return -EINVAL;
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
tunables->up_rate_limit_us = rate_limit_us;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
list_for_each_entry(sg_policy, &attr_set->policy_list, tunables_hook) {
sg_policy->up_rate_delay_ns = rate_limit_us * NSEC_PER_USEC;
update_min_rate_limit_us(sg_policy);
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
return count;
}
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
static ssize_t down_rate_limit_us_store(struct gov_attr_set *attr_set,
const char *buf, size_t count)
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
struct sugov_policy *sg_policy;
unsigned int rate_limit_us;
if (kstrtouint(buf, 10, &rate_limit_us))
return -EINVAL;
tunables->down_rate_limit_us = rate_limit_us;
list_for_each_entry(sg_policy, &attr_set->policy_list, tunables_hook) {
sg_policy->down_rate_delay_ns = rate_limit_us * NSEC_PER_USEC;
update_min_rate_limit_us(sg_policy);
}
return count;
}
FROMLIST: sched: Make iowait_boost optional in schedutil We should apply the iowait boost only if cpufreq policy has iowait boost enabled. Also make it a schedutil configuration from sysfs so it can be turned on/off if needed (by default initialize it to the policy value). For systems that don't need/want it enabled, such as those on arm64 based mobile devices that are battery operated, it saves energy when the cpufreq driver policy doesn't have it enabled (details below): Here are some results for energy measurements collected running a YouTube video for 30 seconds: Before: 8.042533 mWh After: 7.948377 mWh Energy savings is ~1.2% Bug: 38010527 Link: https://lkml.org/lkml/2017/5/19/42 Change-Id: If124076ad0c16ade369253840dedfbf870aff927 Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-05-18 22:46:10 -07:00
static ssize_t iowait_boost_enable_show(struct gov_attr_set *attr_set,
char *buf)
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
return sprintf(buf, "%u\n", tunables->iowait_boost_enable);
}
static ssize_t iowait_boost_enable_store(struct gov_attr_set *attr_set,
const char *buf, size_t count)
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
bool enable;
if (kstrtobool(buf, &enable))
return -EINVAL;
tunables->iowait_boost_enable = enable;
return count;
}
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
static ssize_t exp_util_show(struct gov_attr_set *attr_set, char *buf)
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
return scnprintf(buf, PAGE_SIZE, "%u\n", tunables->exp_util);
}
static ssize_t exp_util_store(struct gov_attr_set *attr_set, const char *buf,
size_t count)
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
if (kstrtobool(buf, &tunables->exp_util))
return -EINVAL;
return count;
}
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
static ssize_t fb_legacy_show(struct gov_attr_set *attr_set, char *buf)
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
return scnprintf(buf, PAGE_SIZE, "%u\n", tunables->fb_legacy);
}
static ssize_t fb_legacy_store(struct gov_attr_set *attr_set, const char *buf,
size_t count)
{
struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
if (kstrtobool(buf, &tunables->fb_legacy))
return -EINVAL;
return count;
}
#endif
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
static struct governor_attr up_rate_limit_us = __ATTR_RW(up_rate_limit_us);
static struct governor_attr down_rate_limit_us = __ATTR_RW(down_rate_limit_us);
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
static struct governor_attr fb_legacy = __ATTR_RW(fb_legacy);
#endif
FROMLIST: sched: Make iowait_boost optional in schedutil We should apply the iowait boost only if cpufreq policy has iowait boost enabled. Also make it a schedutil configuration from sysfs so it can be turned on/off if needed (by default initialize it to the policy value). For systems that don't need/want it enabled, such as those on arm64 based mobile devices that are battery operated, it saves energy when the cpufreq driver policy doesn't have it enabled (details below): Here are some results for energy measurements collected running a YouTube video for 30 seconds: Before: 8.042533 mWh After: 7.948377 mWh Energy savings is ~1.2% Bug: 38010527 Link: https://lkml.org/lkml/2017/5/19/42 Change-Id: If124076ad0c16ade369253840dedfbf870aff927 Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-05-18 22:46:10 -07:00
static struct governor_attr iowait_boost_enable = __ATTR_RW(iowait_boost_enable);
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
static struct governor_attr exp_util = __ATTR_RW(exp_util);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
static struct attribute *sugov_attributes[] = {
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
&up_rate_limit_us.attr,
&down_rate_limit_us.attr,
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
&fb_legacy.attr,
#endif
FROMLIST: sched: Make iowait_boost optional in schedutil We should apply the iowait boost only if cpufreq policy has iowait boost enabled. Also make it a schedutil configuration from sysfs so it can be turned on/off if needed (by default initialize it to the policy value). For systems that don't need/want it enabled, such as those on arm64 based mobile devices that are battery operated, it saves energy when the cpufreq driver policy doesn't have it enabled (details below): Here are some results for energy measurements collected running a YouTube video for 30 seconds: Before: 8.042533 mWh After: 7.948377 mWh Energy savings is ~1.2% Bug: 38010527 Link: https://lkml.org/lkml/2017/5/19/42 Change-Id: If124076ad0c16ade369253840dedfbf870aff927 Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-05-18 22:46:10 -07:00
&iowait_boost_enable.attr,
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
&exp_util.attr,
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
NULL
};
static struct kobj_type sugov_tunables_ktype = {
.default_attrs = sugov_attributes,
.sysfs_ops = &governor_sysfs_ops,
};
/********************** cpufreq governor interface *********************/
static struct cpufreq_governor schedutil_gov;
static struct sugov_policy *sugov_policy_alloc(struct cpufreq_policy *policy)
{
struct sugov_policy *sg_policy;
sg_policy = kzalloc(sizeof(*sg_policy), GFP_KERNEL);
if (!sg_policy)
return NULL;
sg_policy->policy = policy;
raw_spin_lock_init(&sg_policy->update_lock);
return sg_policy;
}
static void sugov_policy_free(struct sugov_policy *sg_policy)
{
import G965FXXU7DTAA OSRC *First release for Android (Q). Signed-off-by: FAROVITUS <farovitus@gmail.com>
2020-02-04 13:44:48 +02:00
kfree(sg_policy);
UPSTREAM: cpufreq: schedutil: move slow path from workqueue to SCHED_FIFO task If slow path frequency changes are conducted in a SCHED_OTHER context then they may be delayed for some amount of time, including indefinitely, when real time or deadline activity is taking place. Move the slow path to a real time kernel thread. In the future the thread should be made SCHED_DEADLINE. The RT priority is arbitrarily set to 50 for now. Hackbench results on ARM Exynos, dual core A15 platform for 10 iterations: $ hackbench -s 100 -l 100 -g 10 -f 20 Before After --------------------------------- 1.808 1.603 1.847 1.251 2.229 1.590 1.952 1.600 1.947 1.257 1.925 1.627 2.694 1.620 1.258 1.621 1.919 1.632 1.250 1.240 Average: 1.8829 1.5041 Based on initial work by Steve Muckle. Change-Id: I8bb111435b0871e961a6f26db1151c91624817b8 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry-picked from commit 02a7b1ee3baa15a98b541d8cfd156bbe1a091c20) [fixed cherry-pick issue] Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-15 13:53:22 +05:30
}
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
static int sugov_kthread_create(struct sugov_policy *sg_policy)
{
struct task_struct *thread;
struct sched_param param = { .sched_priority = MAX_USER_RT_PRIO / 2 };
struct cpufreq_policy *policy = sg_policy->policy;
int ret;
/* kthread only required for slow path */
if (policy->fast_switch_enabled)
return 0;
kthread_init_work(&sg_policy->work, sugov_work);
kthread_init_worker(&sg_policy->worker);
thread = kthread_create(kthread_worker_fn, &sg_policy->worker,
"sugov:%d",
cpumask_first(policy->related_cpus));
if (IS_ERR(thread)) {
pr_err("failed to create sugov thread: %ld\n", PTR_ERR(thread));
return PTR_ERR(thread);
}
ret = sched_setscheduler_nocheck(thread, SCHED_FIFO, &param);
if (ret) {
kthread_stop(thread);
pr_warn("%s: failed to set SCHED_FIFO\n", __func__);
return ret;
}
sg_policy->thread = thread;
kthread_bind_mask(thread, policy->related_cpus);
init_irq_work(&sg_policy->irq_work, sugov_irq_work);
mutex_init(&sg_policy->work_lock);
wake_up_process(thread);
return 0;
}
static void sugov_kthread_stop(struct sugov_policy *sg_policy)
{
/* kthread only required for slow path */
if (sg_policy->policy->fast_switch_enabled)
return;
kthread_flush_worker(&sg_policy->worker);
kthread_stop(sg_policy->thread);
mutex_destroy(&sg_policy->work_lock);
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
static struct sugov_tunables *sugov_tunables_alloc(struct sugov_policy *sg_policy)
{
struct sugov_tunables *tunables;
tunables = kzalloc(sizeof(*tunables), GFP_KERNEL);
if (tunables) {
gov_attr_set_init(&tunables->attr_set, &sg_policy->tunables_hook);
if (!have_governor_per_policy())
global_tunables = tunables;
}
return tunables;
}
BACKPORT: cpufreq: schedutil: Cache tunables on governor exit Currently when all the related CPUs from a policy go offline or the governor is switched, cpufreq framework calls sugov_exit() that frees the governor tunables. When any of the related CPUs comes back online or governor is switched back to schedutil sugov_init() gets called which allocates a fresh set of tunables that are set to default values. This can cause the userspace settings to those tunables to be lost across governor switches or when an entire cluster is hotplugged out. To prevent this, save the tunable values on governor exit. Restore these values to the newly allocated tunables on governor init. Change-Id: I671d4d0e1a4e63e948bfddb0005367df33c0c249 Signed-off-by: Rohit Gupta <rohgup@codeaurora.org> [Caching and restoring different tunables.] Signed-off-by: joshuous <joshuous@gmail.com> Signed-off-by: joshuous <joshuous@gmail.com> (cherry picked from commit a6447dc2b4f53b87e968248935d7ef51d62df73f) (cherry picked from commit d4fbc6341db19a040685e9d295cc5c967d5c3bc4)
2018-05-29 17:23:04 +02:00
static void sugov_tunables_save(struct cpufreq_policy *policy,
struct sugov_tunables *tunables)
{
int cpu;
struct sugov_tunables *cached = per_cpu(cached_tunables, policy->cpu);
if (!have_governor_per_policy())
return;
if (!cached) {
cached = kzalloc(sizeof(*tunables), GFP_KERNEL);
if (!cached) {
pr_warn("Couldn't allocate tunables for caching\n");
return;
}
for_each_cpu(cpu, policy->related_cpus)
per_cpu(cached_tunables, cpu) = cached;
}
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
cached->exp_util = tunables->exp_util;
BACKPORT: cpufreq: schedutil: Cache tunables on governor exit Currently when all the related CPUs from a policy go offline or the governor is switched, cpufreq framework calls sugov_exit() that frees the governor tunables. When any of the related CPUs comes back online or governor is switched back to schedutil sugov_init() gets called which allocates a fresh set of tunables that are set to default values. This can cause the userspace settings to those tunables to be lost across governor switches or when an entire cluster is hotplugged out. To prevent this, save the tunable values on governor exit. Restore these values to the newly allocated tunables on governor init. Change-Id: I671d4d0e1a4e63e948bfddb0005367df33c0c249 Signed-off-by: Rohit Gupta <rohgup@codeaurora.org> [Caching and restoring different tunables.] Signed-off-by: joshuous <joshuous@gmail.com> Signed-off-by: joshuous <joshuous@gmail.com> (cherry picked from commit a6447dc2b4f53b87e968248935d7ef51d62df73f) (cherry picked from commit d4fbc6341db19a040685e9d295cc5c967d5c3bc4)
2018-05-29 17:23:04 +02:00
cached->up_rate_limit_us = tunables->up_rate_limit_us;
cached->down_rate_limit_us = tunables->down_rate_limit_us;
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
static void sugov_tunables_free(struct sugov_tunables *tunables)
{
if (!have_governor_per_policy())
global_tunables = NULL;
kfree(tunables);
}
BACKPORT: cpufreq: schedutil: Cache tunables on governor exit Currently when all the related CPUs from a policy go offline or the governor is switched, cpufreq framework calls sugov_exit() that frees the governor tunables. When any of the related CPUs comes back online or governor is switched back to schedutil sugov_init() gets called which allocates a fresh set of tunables that are set to default values. This can cause the userspace settings to those tunables to be lost across governor switches or when an entire cluster is hotplugged out. To prevent this, save the tunable values on governor exit. Restore these values to the newly allocated tunables on governor init. Change-Id: I671d4d0e1a4e63e948bfddb0005367df33c0c249 Signed-off-by: Rohit Gupta <rohgup@codeaurora.org> [Caching and restoring different tunables.] Signed-off-by: joshuous <joshuous@gmail.com> Signed-off-by: joshuous <joshuous@gmail.com> (cherry picked from commit a6447dc2b4f53b87e968248935d7ef51d62df73f) (cherry picked from commit d4fbc6341db19a040685e9d295cc5c967d5c3bc4)
2018-05-29 17:23:04 +02:00
static void sugov_tunables_restore(struct cpufreq_policy *policy)
{
struct sugov_policy *sg_policy = policy->governor_data;
struct sugov_tunables *tunables = sg_policy->tunables;
struct sugov_tunables *cached = per_cpu(cached_tunables, policy->cpu);
if (!cached)
return;
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
tunables->exp_util = cached->exp_util;
BACKPORT: cpufreq: schedutil: Cache tunables on governor exit Currently when all the related CPUs from a policy go offline or the governor is switched, cpufreq framework calls sugov_exit() that frees the governor tunables. When any of the related CPUs comes back online or governor is switched back to schedutil sugov_init() gets called which allocates a fresh set of tunables that are set to default values. This can cause the userspace settings to those tunables to be lost across governor switches or when an entire cluster is hotplugged out. To prevent this, save the tunable values on governor exit. Restore these values to the newly allocated tunables on governor init. Change-Id: I671d4d0e1a4e63e948bfddb0005367df33c0c249 Signed-off-by: Rohit Gupta <rohgup@codeaurora.org> [Caching and restoring different tunables.] Signed-off-by: joshuous <joshuous@gmail.com> Signed-off-by: joshuous <joshuous@gmail.com> (cherry picked from commit a6447dc2b4f53b87e968248935d7ef51d62df73f) (cherry picked from commit d4fbc6341db19a040685e9d295cc5c967d5c3bc4)
2018-05-29 17:23:04 +02:00
tunables->up_rate_limit_us = cached->up_rate_limit_us;
tunables->down_rate_limit_us = cached->down_rate_limit_us;
sg_policy->up_rate_delay_ns =
tunables->up_rate_limit_us * NSEC_PER_USEC;
sg_policy->down_rate_delay_ns =
tunables->down_rate_limit_us * NSEC_PER_USEC;
sg_policy->min_rate_limit_ns = min(sg_policy->up_rate_delay_ns,
sg_policy->down_rate_delay_ns);
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
static int sugov_init(struct cpufreq_policy *policy)
{
struct sugov_policy *sg_policy;
struct sugov_tunables *tunables;
int ret = 0;
/* State should be equivalent to EXIT */
if (policy->governor_data)
return -EBUSY;
UPSTREAM: cpufreq: schedutil: enable fast switch earlier The fast_switch_enabled flag will be used by both sugov_policy_alloc() and sugov_policy_free() with a later patch. Prepare for that by moving the calls to enable and disable it to the beginning of sugov_init() and end of sugov_exit(). Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 4a71ce4348bb61740d411822357061f8bf870f4c) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ia174f423ca02d59360657ac2e77a5098ce5cf99c Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-07-20 16:32:35 +01:00
cpufreq_enable_fast_switch(policy);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sg_policy = sugov_policy_alloc(policy);
UPSTREAM: cpufreq: schedutil: enable fast switch earlier The fast_switch_enabled flag will be used by both sugov_policy_alloc() and sugov_policy_free() with a later patch. Prepare for that by moving the calls to enable and disable it to the beginning of sugov_init() and end of sugov_exit(). Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 4a71ce4348bb61740d411822357061f8bf870f4c) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ia174f423ca02d59360657ac2e77a5098ce5cf99c Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-07-20 16:32:35 +01:00
if (!sg_policy) {
ret = -ENOMEM;
goto disable_fast_switch;
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
ret = sugov_kthread_create(sg_policy);
if (ret)
goto free_sg_policy;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
mutex_lock(&global_tunables_lock);
if (global_tunables) {
if (WARN_ON(have_governor_per_policy())) {
ret = -EINVAL;
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
goto stop_kthread;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
policy->governor_data = sg_policy;
sg_policy->tunables = global_tunables;
gov_attr_set_get(&global_tunables->attr_set, &sg_policy->tunables_hook);
goto out;
}
tunables = sugov_tunables_alloc(sg_policy);
if (!tunables) {
ret = -ENOMEM;
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
goto stop_kthread;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
if (policy->up_transition_delay_us && policy->down_transition_delay_us) {
tunables->up_rate_limit_us = policy->up_transition_delay_us;
tunables->down_rate_limit_us = policy->down_transition_delay_us;
} else {
unsigned int lat;
tunables->up_rate_limit_us = LATENCY_MULTIPLIER;
tunables->down_rate_limit_us = LATENCY_MULTIPLIER;
lat = policy->cpuinfo.transition_latency / NSEC_PER_USEC;
if (lat) {
tunables->up_rate_limit_us *= lat;
tunables->down_rate_limit_us *= lat;
}
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
}
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
tunables->fb_legacy = false;
sg_policy->be_stochastic = false;
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
FROMLIST: sched: Make iowait_boost optional in schedutil We should apply the iowait boost only if cpufreq policy has iowait boost enabled. Also make it a schedutil configuration from sysfs so it can be turned on/off if needed (by default initialize it to the policy value). For systems that don't need/want it enabled, such as those on arm64 based mobile devices that are battery operated, it saves energy when the cpufreq driver policy doesn't have it enabled (details below): Here are some results for energy measurements collected running a YouTube video for 30 seconds: Before: 8.042533 mWh After: 7.948377 mWh Energy savings is ~1.2% Bug: 38010527 Link: https://lkml.org/lkml/2017/5/19/42 Change-Id: If124076ad0c16ade369253840dedfbf870aff927 Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-05-18 22:46:10 -07:00
tunables->iowait_boost_enable = policy->iowait_boost_enable;
cpufreq: schedutil: Add tunable for exponential freq selection Signed-off-by: Tyler Nijmeh <tylernij@gmail.com>
2020-03-02 20:34:04 -08:00
tunables->exp_util = false;
FROMLIST: sched: Make iowait_boost optional in schedutil We should apply the iowait boost only if cpufreq policy has iowait boost enabled. Also make it a schedutil configuration from sysfs so it can be turned on/off if needed (by default initialize it to the policy value). For systems that don't need/want it enabled, such as those on arm64 based mobile devices that are battery operated, it saves energy when the cpufreq driver policy doesn't have it enabled (details below): Here are some results for energy measurements collected running a YouTube video for 30 seconds: Before: 8.042533 mWh After: 7.948377 mWh Energy savings is ~1.2% Bug: 38010527 Link: https://lkml.org/lkml/2017/5/19/42 Change-Id: If124076ad0c16ade369253840dedfbf870aff927 Signed-off-by: Joel Fernandes <joelaf@google.com>
2017-05-18 22:46:10 -07:00
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
policy->governor_data = sg_policy;
sg_policy->tunables = tunables;
BACKPORT: cpufreq: schedutil: Cache tunables on governor exit Currently when all the related CPUs from a policy go offline or the governor is switched, cpufreq framework calls sugov_exit() that frees the governor tunables. When any of the related CPUs comes back online or governor is switched back to schedutil sugov_init() gets called which allocates a fresh set of tunables that are set to default values. This can cause the userspace settings to those tunables to be lost across governor switches or when an entire cluster is hotplugged out. To prevent this, save the tunable values on governor exit. Restore these values to the newly allocated tunables on governor init. Change-Id: I671d4d0e1a4e63e948bfddb0005367df33c0c249 Signed-off-by: Rohit Gupta <rohgup@codeaurora.org> [Caching and restoring different tunables.] Signed-off-by: joshuous <joshuous@gmail.com> Signed-off-by: joshuous <joshuous@gmail.com> (cherry picked from commit a6447dc2b4f53b87e968248935d7ef51d62df73f) (cherry picked from commit d4fbc6341db19a040685e9d295cc5c967d5c3bc4)
2018-05-29 17:23:04 +02:00
sugov_tunables_restore(policy);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
ret = kobject_init_and_add(&tunables->attr_set.kobj, &sugov_tunables_ktype,
get_governor_parent_kobj(policy), "%s",
schedutil_gov.name);
if (ret)
goto fail;
UPSTREAM: cpufreq: schedutil: Avoid indented labels Switch to the more common practice of writing labels. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 8e2ddb03643eb9d0bc4926946d7ce0d308eef0a5) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ida75c99cf3dff5cae24d3866454c83bcdb3385b9 Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:04:04 +01:00
out:
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
mutex_unlock(&global_tunables_lock);
return 0;
UPSTREAM: cpufreq: schedutil: Avoid indented labels Switch to the more common practice of writing labels. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 8e2ddb03643eb9d0bc4926946d7ce0d308eef0a5) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ida75c99cf3dff5cae24d3866454c83bcdb3385b9 Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-05-25 15:04:04 +01:00
fail:
sched/cpufreq: Fix kobject memleak [ Upstream commit 9a4f26cc98d81b67ecc23b890c28e2df324e29f3 ] Currently the error return path from kobject_init_and_add() is not followed by a call to kobject_put() - which means we are leaking the kobject. Fix it by adding a call to kobject_put() in the error path of kobject_init_and_add(). Signed-off-by: Tobin C. Harding <tobin@kernel.org> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tobin C. Harding <tobin@kernel.org> Cc: Vincent Guittot <vincent.guittot@linaro.org> Cc: Viresh Kumar <viresh.kumar@linaro.org> Link: http://lkml.kernel.org/r/20190430001144.24890-1-tobin@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
2019-04-30 10:11:44 +10:00
kobject_put(&tunables->attr_set.kobj);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
policy->governor_data = NULL;
sugov_tunables_free(tunables);
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
stop_kthread:
sugov_kthread_stop(sg_policy);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
mutex_unlock(&global_tunables_lock);
schedutil: Fix error path mutex unlock This patch prevents the 'global_tunables_lock' mutex from being unlocked before being locked. This mutex is not locked if the sugov_kthread_create() function fails. Change-Id: Ie1547781395ea9e39d89e02f66075a13fd0e79a7 Signed-off-by: Jules Maselbas <jules.maselbas@arm.com> Signed-off-by: mydongistiny <jaysonedson@gmail.com> Signed-off-by: twisted <36546624+TwistedPrime@users.noreply.github.com>
2018-03-29 15:43:01 +01:00
free_sg_policy:
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sugov_policy_free(sg_policy);
UPSTREAM: cpufreq: schedutil: enable fast switch earlier The fast_switch_enabled flag will be used by both sugov_policy_alloc() and sugov_policy_free() with a later patch. Prepare for that by moving the calls to enable and disable it to the beginning of sugov_init() and end of sugov_exit(). Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> (cherry picked from commit 4a71ce4348bb61740d411822357061f8bf870f4c) Signed-off-by: Chris Redpath <chris.redpath@arm.com> Change-Id: Ia174f423ca02d59360657ac2e77a5098ce5cf99c Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2017-07-20 16:32:35 +01:00
disable_fast_switch:
cpufreq_disable_fast_switch(policy);
cpufreq: schedutil: Improve prints messages with pr_fmt Prefix print messages with KBUILD_MODNAME, i.e 'cpufreq_schedutil: '. This helps to keep similar formatting for all the print messages particular to a file and identify those easily in kernel logs. Its already done this way for rest of the governors. Along with that, remove the (now) redundant bits from a print message. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-05-18 17:55:28 +05:30
pr_err("initialization failed (error %d)\n", ret);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
return ret;
}
cpufreq: governor: Get rid of governor events The design of the cpufreq governor API is not very straightforward, as struct cpufreq_governor provides only one callback to be invoked from different code paths for different purposes. The purpose it is invoked for is determined by its second "event" argument, causing it to act as a "callback multiplexer" of sorts. Unfortunately, that leads to extra complexity in governors, some of which implement the ->governor() callback as a switch statement that simply checks the event argument and invokes a separate function to handle that specific event. That extra complexity can be eliminated by replacing the all-purpose ->governor() callback with a family of callbacks to carry out specific governor operations: initialization and exit, start and stop and policy limits updates. That also turns out to reduce the code size too, so do it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-06-02 23:24:15 +02:00
static void sugov_exit(struct cpufreq_policy *policy)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
struct sugov_policy *sg_policy = policy->governor_data;
struct sugov_tunables *tunables = sg_policy->tunables;
unsigned int count;
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, policy->cpu);
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
mutex_lock(&global_tunables_lock);
count = gov_attr_set_put(&tunables->attr_set, &sg_policy->tunables_hook);
policy->governor_data = NULL;
BACKPORT: cpufreq: schedutil: Cache tunables on governor exit Currently when all the related CPUs from a policy go offline or the governor is switched, cpufreq framework calls sugov_exit() that frees the governor tunables. When any of the related CPUs comes back online or governor is switched back to schedutil sugov_init() gets called which allocates a fresh set of tunables that are set to default values. This can cause the userspace settings to those tunables to be lost across governor switches or when an entire cluster is hotplugged out. To prevent this, save the tunable values on governor exit. Restore these values to the newly allocated tunables on governor init. Change-Id: I671d4d0e1a4e63e948bfddb0005367df33c0c249 Signed-off-by: Rohit Gupta <rohgup@codeaurora.org> [Caching and restoring different tunables.] Signed-off-by: joshuous <joshuous@gmail.com> Signed-off-by: joshuous <joshuous@gmail.com> (cherry picked from commit a6447dc2b4f53b87e968248935d7ef51d62df73f) (cherry picked from commit d4fbc6341db19a040685e9d295cc5c967d5c3bc4)
2018-05-29 17:23:04 +02:00
if (!count) {
sugov_tunables_save(policy, tunables);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sugov_tunables_free(tunables);
BACKPORT: cpufreq: schedutil: Cache tunables on governor exit Currently when all the related CPUs from a policy go offline or the governor is switched, cpufreq framework calls sugov_exit() that frees the governor tunables. When any of the related CPUs comes back online or governor is switched back to schedutil sugov_init() gets called which allocates a fresh set of tunables that are set to default values. This can cause the userspace settings to those tunables to be lost across governor switches or when an entire cluster is hotplugged out. To prevent this, save the tunable values on governor exit. Restore these values to the newly allocated tunables on governor init. Change-Id: I671d4d0e1a4e63e948bfddb0005367df33c0c249 Signed-off-by: Rohit Gupta <rohgup@codeaurora.org> [Caching and restoring different tunables.] Signed-off-by: joshuous <joshuous@gmail.com> Signed-off-by: joshuous <joshuous@gmail.com> (cherry picked from commit a6447dc2b4f53b87e968248935d7ef51d62df73f) (cherry picked from commit d4fbc6341db19a040685e9d295cc5c967d5c3bc4)
2018-05-29 17:23:04 +02:00
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
if (sg_cpu->util_vessel) {
sg_cpu->util_vessel->finalizer(sg_cpu->util_vessel);
kair_obj_destructor(sg_cpu->util_vessel);
sg_cpu->util_vessel = NULL;
}
sg_policy->be_stochastic = false;
#endif
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
sugov_kthread_stop(sg_policy);
sugov_policy_free(sg_policy);
cpufreq: prevent to access NULL sg_policy Change-Id: If6a532db0dc3a7fe7b3146fd48c03550d5e16fbd Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2017-08-01 19:36:30 +09:00
mutex_unlock(&global_tunables_lock);
import G965FXXU7DTAA OSRC *First release for Android (Q). Signed-off-by: FAROVITUS <farovitus@gmail.com>
2020-02-04 13:44:48 +02:00
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
cpufreq_disable_fast_switch(policy);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
static int sugov_start(struct cpufreq_policy *policy)
{
struct sugov_policy *sg_policy = policy->governor_data;
unsigned int cpu;
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
char alias[KAIR_ALIAS_LEN];
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
ANDROID: cpufreq: schedutil: add up/down frequency transition rate limits The rate-limit tunable in the schedutil governor applies to transitions to both lower and higher frequencies. On several platforms it is not the ideal tunable though, as it is difficult to get best power/performance figures using the same limit in both directions. It is common on mobile platforms with demanding user interfaces to want to increase frequency rapidly for example but decrease slowly. One of the example can be a case where we have short busy periods followed by similar or longer idle periods. If we keep the rate-limit high enough, we will not go to higher frequencies soon enough. On the other hand, if we keep it too low, we will have too many frequency transitions, as we will always reduce the frequency after the busy period. It would be very useful if we can set low rate-limit while increasing the frequency (so that we can respond to the short busy periods quickly) and high rate-limit while decreasing frequency (so that we don't reduce the frequency immediately after the short busy period and that may avoid frequency transitions before the next busy period). Implement separate up/down transition rate limits. Note that the governor avoids frequency recalculations for a period equal to minimum of up and down rate-limit. A global mutex is also defined to protect updates to min_rate_limit_us via two separate sysfs files. Note that this wouldn't change behavior of the schedutil governor for the platforms which wish to keep same values for both up and down rate limits. This is tested with the rt-app [1] on ARM Exynos, dual A15 processor platform. Testcase: Run a SCHED_OTHER thread on CPU0 which will emulate work-load for X ms of busy period out of the total period of Y ms, i.e. Y - X ms of idle period. The values of X/Y taken were: 20/40, 20/50, 20/70, i.e idle periods of 20, 30 and 50 ms respectively. These were tested against values of up/down rate limits as: 10/10 ms and 10/40 ms. For every test we noticed a performance increase of 5-10% with the schedutil governor, which was very much expected. [Viresh]: Simplified user interface and introduced min_rate_limit_us + mutex, rewrote commit log and included test results. [1] https://github.com/scheduler-tools/rt-app/ Change-Id: I18720a83855b196b8e21dcdc8deae79131635b84 Signed-off-by: Steve Muckle <smuckle.linux@gmail.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> (applied from https://marc.info/?l=linux-kernel&m=147936011103832&w=2) [trivial adaptations] Signed-off-by: Juri Lelli <juri.lelli@arm.com> Signed-off-by: Quentin Perret <quentin.perret@arm.com>
2016-11-17 10:48:45 +05:30
sg_policy->up_rate_delay_ns =
sg_policy->tunables->up_rate_limit_us * NSEC_PER_USEC;
sg_policy->down_rate_delay_ns =
sg_policy->tunables->down_rate_limit_us * NSEC_PER_USEC;
update_min_rate_limit_us(sg_policy);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sg_policy->last_freq_update_time = 0;
UPSTREAM: cpufreq: schedutil: Don't set next_freq to UINT_MAX The schedutil driver sets sg_policy->next_freq to UINT_MAX on certain occasions to discard the cached value of next freq: - In sugov_start(), when the schedutil governor is started for a group of CPUs. - And whenever we need to force a freq update before rate-limit duration, which happens when: - there is an update in cpufreq policy limits. - Or when the utilization of DL scheduling class increases. In return, get_next_freq() doesn't return a cached next_freq value but recalculates the next frequency instead. But having special meaning for a particular value of frequency makes the code less readable and error prone. We recently fixed a bug where the UINT_MAX value was considered as valid frequency in sugov_update_single(). All we need is a flag which can be used to discard the value of sg_policy->next_freq and we already have need_freq_update for that. Lets reuse it instead of setting next_freq to UINT_MAX. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: joshuous <joshuous@gmail.com> Conflicts: kernel/sched/cpufreq_schedutil.c
2018-05-09 16:05:24 +05:30
sg_policy->next_freq = 0;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sg_policy->work_in_progress = false;
sg_policy->need_freq_update = false;
cpufreq: schedutil: move cached_raw_freq to struct sugov_policy commit 6c4f0fa643cb9e775dcc976e3db00d649468ff1d upstream. cached_raw_freq applies to the entire cpufreq policy and not individual CPUs. Apart from wasting per-cpu memory, it is actually wrong to keep it in struct sugov_cpu as we may end up comparing next_freq with a stale cached_raw_freq of a random CPU. Move cached_raw_freq to struct sugov_policy. Fixes: 5cbea46984d6 (cpufreq: schedutil: map raw required frequency to driver frequency) Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Amit Pundir <amit.pundir@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-03-02 14:03:20 +05:30
sg_policy->cached_raw_freq = 0;
cpufreq: schedutil: Don't skip freq update when limits change To avoid reducing the frequency of a CPU prematurely, we skip reducing the frequency if the CPU had been busy recently. This should not be done when the limits of the policy are changed, for example due to thermal throttling. We should always get the frequency within the new limits as soon as possible. Trying to fix this by using only one flag, i.e. need_freq_update, can lead to a race condition where the flag gets cleared without forcing us to change the frequency at least once. And so this patch introduces another flag to avoid that race condition. Fixes: ecd288429126 ("cpufreq: schedutil: Don't set next_freq to UINT_MAX") Cc: v4.18+ <stable@vger.kernel.org> # v4.18+ Reported-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-08-07 12:36:01 +05:30
sg_policy->limits_changed = false;
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
for_each_cpu(cpu, policy->cpus) {
struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
if (cpu != policy->cpu) {
memset(sg_cpu, 0, sizeof(*sg_cpu));
goto skip_subcpus;
}
if (!sg_policy->be_stochastic) {
memset(alias, 0, KAIR_ALIAS_LEN);
sprintf(alias, "govern%d", cpu);
memset(sg_cpu, 0, sizeof(*sg_cpu));
sg_cpu->util_vessel =
kair_obj_creator(alias,
UTILAVG_KAIR_VARIANCE,
policy->cpuinfo.max_freq,
policy->cpuinfo.min_freq,
&kairistic_cpufreq);
if (sg_cpu->util_vessel->initializer(sg_cpu->util_vessel) < 0) {
sg_cpu->util_vessel->finalizer(sg_cpu->util_vessel);
kair_obj_destructor(sg_cpu->util_vessel);
sg_cpu->util_vessel = NULL;
}
} else {
struct kair_class *vptr = sg_cpu->util_vessel;
memset(sg_cpu, 0, sizeof(*sg_cpu));
sg_cpu->util_vessel = vptr;
}
skip_subcpus:
#else
cpufreq: schedutil: Fix per-CPU structure initialization in sugov_start() commit 4296f23ed49a15d36949458adcc66ff993dee2a8 upstream. sugov_start() only initializes struct sugov_cpu per-CPU structures for shared policies, but it should do that for single-CPU policies too. That in particular makes the IO-wait boost mechanism work in the cases when cpufreq policies correspond to individual CPUs. Fixes: 21ca6d2c52f8 (cpufreq: schedutil: Add iowait boosting) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Amit Pundir <amit.pundir@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-03-19 14:30:02 +01:00
memset(sg_cpu, 0, sizeof(*sg_cpu));
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#endif
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
sg_cpu->sg_policy = sg_policy;
sched: su_gov: remove DL initial flag when starting governor In shared policy, DL initial flag could set max freq util flag removed by updating utilization so, remove this flag. Change-Id: Ib04def4e8d6d42cf3d3b65b71b048f10d1285a5c Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com> Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2018-01-19 16:52:59 +09:00
sg_cpu->flags = 0;
sched: schedutil: Add slack_timer and pm_qos handler 1. the slack_timer works to prevent the fequency from being held unnecessarily high when the cpu sleeps long in a high utilization. 2. The Pm_qos handler tells the governor if the frequency that was held high by qos could fall. Change-Id: I5f2d5be73b8d8294a35b2810cc9721264ec55964 Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com>
2017-11-02 20:47:24 +09:00
sugov_start_slack(cpu);
cpufreq: schedutil: Fix per-CPU structure initialization in sugov_start() commit 4296f23ed49a15d36949458adcc66ff993dee2a8 upstream. sugov_start() only initializes struct sugov_cpu per-CPU structures for shared policies, but it should do that for single-CPU policies too. That in particular makes the IO-wait boost mechanism work in the cases when cpufreq policies correspond to individual CPUs. Fixes: 21ca6d2c52f8 (cpufreq: schedutil: Add iowait boosting) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Amit Pundir <amit.pundir@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-03-19 14:30:02 +01:00
sg_cpu->iowait_boost_max = policy->cpuinfo.max_freq;
cpufreq: schedutil: Fix sugov_start versus sugov_update_shared race With a shared policy in place, when one of the CPUs in the policy is hotplugged out and then brought back online, sugov_stop and sugov_start are called in order. sugov_stop removes utilization hooks for each CPU in the policy and does nothing else in the for_each_cpu loop. sugov_start on the other hand iterates through the CPUs in the policy and re-initializes the per-cpu structure _and_ adds the utilization hook. This implies that the scheduler is allowed to invoke a CPU's utilization update hook when the rest of the per-cpu structures have yet to be re-inited. Apart from some strange values in tracepoints this doesn't cause a problem, but if we do end up accessing a pointer from the per-cpu sugov_cpu structure somewhere in the sugov_update_shared path, we will likely see crashes since the memset for another CPU in the policy is free to race with sugov_update_shared from the CPU that is ready to go. So let's fix this now to first init all per-cpu structures, and then add the per-cpu utilization update hooks all at once. Change-Id: I399e0e159b3db3ae3258843c9231f92312fe18ef Signed-off-by: Vikram Mulukutla <markivx@codeaurora.org>
2017-07-06 10:05:52 -07:00
}
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
sg_policy->be_stochastic = true;
#endif
cpufreq: schedutil: Fix sugov_start versus sugov_update_shared race With a shared policy in place, when one of the CPUs in the policy is hotplugged out and then brought back online, sugov_stop and sugov_start are called in order. sugov_stop removes utilization hooks for each CPU in the policy and does nothing else in the for_each_cpu loop. sugov_start on the other hand iterates through the CPUs in the policy and re-initializes the per-cpu structure _and_ adds the utilization hook. This implies that the scheduler is allowed to invoke a CPU's utilization update hook when the rest of the per-cpu structures have yet to be re-inited. Apart from some strange values in tracepoints this doesn't cause a problem, but if we do end up accessing a pointer from the per-cpu sugov_cpu structure somewhere in the sugov_update_shared path, we will likely see crashes since the memset for another CPU in the policy is free to race with sugov_update_shared from the CPU that is ready to go. So let's fix this now to first init all per-cpu structures, and then add the per-cpu utilization update hooks all at once. Change-Id: I399e0e159b3db3ae3258843c9231f92312fe18ef Signed-off-by: Vikram Mulukutla <markivx@codeaurora.org>
2017-07-06 10:05:52 -07:00
for_each_cpu(cpu, policy->cpus) {
struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
cpufreq: schedutil: Fix per-CPU structure initialization in sugov_start() commit 4296f23ed49a15d36949458adcc66ff993dee2a8 upstream. sugov_start() only initializes struct sugov_cpu per-CPU structures for shared policies, but it should do that for single-CPU policies too. That in particular makes the IO-wait boost mechanism work in the cases when cpufreq policies correspond to individual CPUs. Fixes: 21ca6d2c52f8 (cpufreq: schedutil: Add iowait boosting) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Amit Pundir <amit.pundir@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-03-19 14:30:02 +01:00
cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
policy_is_shared(policy) ?
sugov_update_shared :
sugov_update_single);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
return 0;
}
cpufreq: governor: Get rid of governor events The design of the cpufreq governor API is not very straightforward, as struct cpufreq_governor provides only one callback to be invoked from different code paths for different purposes. The purpose it is invoked for is determined by its second "event" argument, causing it to act as a "callback multiplexer" of sorts. Unfortunately, that leads to extra complexity in governors, some of which implement the ->governor() callback as a switch statement that simply checks the event argument and invokes a separate function to handle that specific event. That extra complexity can be eliminated by replacing the all-purpose ->governor() callback with a family of callbacks to carry out specific governor operations: initialization and exit, start and stop and policy limits updates. That also turns out to reduce the code size too, so do it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-06-02 23:24:15 +02:00
static void sugov_stop(struct cpufreq_policy *policy)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
struct sugov_policy *sg_policy = policy->governor_data;
unsigned int cpu;
sched: schedutil: Add slack_timer and pm_qos handler 1. the slack_timer works to prevent the fequency from being held unnecessarily high when the cpu sleeps long in a high utilization. 2. The Pm_qos handler tells the governor if the frequency that was held high by qos could fall. Change-Id: I5f2d5be73b8d8294a35b2810cc9721264ec55964 Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com>
2017-11-02 20:47:24 +09:00
for_each_cpu(cpu, policy->cpus) {
sugov_stop_slack(cpu);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
cpufreq_remove_update_util_hook(cpu);
sched: schedutil: Add slack_timer and pm_qos handler 1. the slack_timer works to prevent the fequency from being held unnecessarily high when the cpu sleeps long in a high utilization. 2. The Pm_qos handler tells the governor if the frequency that was held high by qos could fall. Change-Id: I5f2d5be73b8d8294a35b2810cc9721264ec55964 Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com>
2017-11-02 20:47:24 +09:00
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
synchronize_sched();
exynos9810: schedutil/KAIR: Sync schedutil to latest patches from Samsung Taken from Note 10 5G & S10 [Exynos 9825/9820] and fully merged with KAIR feature as they quote "AI based Resource Control"
2020-09-18 16:40:46 +03:00
#ifdef CONFIG_SCHED_KAIR_GLUE
for_each_cpu(cpu, policy->cpus) {
struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
if (sg_cpu->util_vessel) {
sg_cpu->util_vessel->stopper(sg_cpu->util_vessel);
}
}
#endif
universal9810: Reset scheduler to android-4.9-q state We will adapt new scheduler mechanism after this Signed-off-by: Diep Quynh <remilia.1505@gmail.com>
2020-04-04 13:06:34 +07:00
if (!policy->fast_switch_enabled) {
irq_work_sync(&sg_policy->irq_work);
kthread_cancel_work_sync(&sg_policy->work);
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
cpufreq: governor: Get rid of governor events The design of the cpufreq governor API is not very straightforward, as struct cpufreq_governor provides only one callback to be invoked from different code paths for different purposes. The purpose it is invoked for is determined by its second "event" argument, causing it to act as a "callback multiplexer" of sorts. Unfortunately, that leads to extra complexity in governors, some of which implement the ->governor() callback as a switch statement that simply checks the event argument and invokes a separate function to handle that specific event. That extra complexity can be eliminated by replacing the all-purpose ->governor() callback with a family of callbacks to carry out specific governor operations: initialization and exit, start and stop and policy limits updates. That also turns out to reduce the code size too, so do it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-06-02 23:24:15 +02:00
static void sugov_limits(struct cpufreq_policy *policy)
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
{
struct sugov_policy *sg_policy = policy->governor_data;
cpufreq: prevent to access NULL sg_policy Change-Id: If6a532db0dc3a7fe7b3146fd48c03550d5e16fbd Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2017-08-01 19:36:30 +09:00
mutex_lock(&global_tunables_lock);
if (!sg_policy) {
mutex_unlock(&global_tunables_lock);
return;
}
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
if (!policy->fast_switch_enabled) {
mutex_lock(&sg_policy->work_lock);
cpufreq: governor: Create cpufreq_policy_apply_limits() Create a new helper to avoid code duplication across governors. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-05-18 17:55:31 +05:30
cpufreq_policy_apply_limits(policy);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
mutex_unlock(&sg_policy->work_lock);
}
sched: schedutil: Add slack_timer and pm_qos handler 1. the slack_timer works to prevent the fequency from being held unnecessarily high when the cpu sleeps long in a high utilization. 2. The Pm_qos handler tells the governor if the frequency that was held high by qos could fall. Change-Id: I5f2d5be73b8d8294a35b2810cc9721264ec55964 Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com>
2017-11-02 20:47:24 +09:00
sugov_update_min(policy);
cpufreq: schedutil: Don't skip freq update when limits change To avoid reducing the frequency of a CPU prematurely, we skip reducing the frequency if the CPU had been busy recently. This should not be done when the limits of the policy are changed, for example due to thermal throttling. We should always get the frequency within the new limits as soon as possible. Trying to fix this by using only one flag, i.e. need_freq_update, can lead to a race condition where the flag gets cleared without forcing us to change the frequency at least once. And so this patch introduces another flag to avoid that race condition. Fixes: ecd288429126 ("cpufreq: schedutil: Don't set next_freq to UINT_MAX") Cc: v4.18+ <stable@vger.kernel.org> # v4.18+ Reported-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-08-07 12:36:01 +05:30
sg_policy->limits_changed = true;
cpufreq: prevent to access NULL sg_policy Change-Id: If6a532db0dc3a7fe7b3146fd48c03550d5e16fbd Signed-off-by: Park Bumgyu <bumgyu.park@samsung.com>
2017-08-01 19:36:30 +09:00
mutex_unlock(&global_tunables_lock);
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
}
static struct cpufreq_governor schedutil_gov = {
.name = "schedutil",
.owner = THIS_MODULE,
cpufreq: governor: Get rid of governor events The design of the cpufreq governor API is not very straightforward, as struct cpufreq_governor provides only one callback to be invoked from different code paths for different purposes. The purpose it is invoked for is determined by its second "event" argument, causing it to act as a "callback multiplexer" of sorts. Unfortunately, that leads to extra complexity in governors, some of which implement the ->governor() callback as a switch statement that simply checks the event argument and invokes a separate function to handle that specific event. That extra complexity can be eliminated by replacing the all-purpose ->governor() callback with a family of callbacks to carry out specific governor operations: initialization and exit, start and stop and policy limits updates. That also turns out to reduce the code size too, so do it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-06-02 23:24:15 +02:00
.init = sugov_init,
.exit = sugov_exit,
.start = sugov_start,
.stop = sugov_stop,
.limits = sugov_limits,
cpufreq: schedutil: New governor based on scheduler utilization data Add a new cpufreq scaling governor, called "schedutil", that uses scheduler-provided CPU utilization information as input for making its decisions. Doing that is possible after commit 34e2c555f3e1 (cpufreq: Add mechanism for registering utilization update callbacks) that introduced cpufreq_update_util() called by the scheduler on utilization changes (from CFS) and RT/DL task status updates. In particular, CPU frequency scaling decisions may be based on the the utilization data passed to cpufreq_update_util() by CFS. The new governor is relatively simple. The frequency selection formula used by it depends on whether or not the utilization is frequency-invariant. In the frequency-invariant case the new CPU frequency is given by next_freq = 1.25 * max_freq * util / max where util and max are the last two arguments of cpufreq_update_util(). In turn, if util is not frequency-invariant, the maximum frequency in the above formula is replaced with the current frequency of the CPU: next_freq = 1.25 * curr_freq * util / max The coefficient 1.25 corresponds to the frequency tipping point at (util / max) = 0.8. All of the computations are carried out in the utilization update handlers provided by the new governor. One of those handlers is used for cpufreq policies shared between multiple CPUs and the other one is for policies with one CPU only (and therefore it doesn't need to use any extra synchronization means). The governor supports fast frequency switching if that is supported by the cpufreq driver in use and possible for the given policy. In the fast switching case, all operations of the governor take place in its utilization update handlers. If fast switching cannot be used, the frequency switch operations are carried out with the help of a work item which only calls __cpufreq_driver_target() (under a mutex) to trigger a frequency update (to a value already computed beforehand in one of the utilization update handlers). Currently, the governor treats all of the RT and DL tasks as "unknown utilization" and sets the frequency to the allowed maximum when updated from the RT or DL sched classes. That heavy-handed approach should be replaced with something more subtle and specifically targeted at RT and DL tasks. The governor shares some tunables management code with the "ondemand" and "conservative" governors and uses some common definitions from cpufreq_governor.h, but apart from that it is stand-alone. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
2016-04-02 01:09:12 +02:00
};
#ifdef CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL
struct cpufreq_governor *cpufreq_default_governor(void)
{
return &schedutil_gov;
}
#endif
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
sched: schedutil: Add slack_timer and pm_qos handler 1. the slack_timer works to prevent the fequency from being held unnecessarily high when the cpu sleeps long in a high utilization. 2. The Pm_qos handler tells the governor if the frequency that was held high by qos could fall. Change-Id: I5f2d5be73b8d8294a35b2810cc9721264ec55964 Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com>
2017-11-02 20:47:24 +09:00
static void sugov_update_min(struct cpufreq_policy *policy)
{
int cpu, max_cap;
struct sugov_exynos *sg_exynos;
int min_cap;
max_cap = arch_scale_cpu_capacity(NULL, policy->cpu);
/* min_cap is minimum value making higher frequency than policy->min */
min_cap = max_cap * policy->min / policy->max;
min_cap = (min_cap * 4 / 5) + 1;
for_each_cpu(cpu, policy->cpus) {
sg_exynos = &per_cpu(sugov_exynos, cpu);
sg_exynos->min = min_cap;
}
}
static void sugov_nop_timer(unsigned long data)
{
/*
* The purpose of slack-timer is to wake up the CPU from IDLE, in order
* to decrease its frequency if it is not set to minimum already.
*
* This is important for platforms where CPU with higher frequencies
* consume higher power even at IDLE.
*/
trace_sugov_slack_func(smp_processor_id());
}
static void sugov_start_slack(int cpu)
{
struct sugov_exynos *sg_exynos = &per_cpu(sugov_exynos, cpu);
if (!sg_exynos->enabled)
return;
sg_exynos->min = ULONG_MAX;
sg_exynos->started = true;
}
static void sugov_stop_slack(int cpu)
{
struct sugov_exynos *sg_exynos = &per_cpu(sugov_exynos, cpu);
sg_exynos->started = false;
if (timer_pending(&sg_exynos->timer))
del_timer_sync(&sg_exynos->timer);
}
static s64 get_next_event_time_ms(void)
{
return ktime_to_us(tick_nohz_get_sleep_length());
}
static int sugov_need_slack_timer(unsigned int cpu)
{
struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
struct sugov_exynos *sg_exynos = &per_cpu(sugov_exynos, cpu);
if (sg_cpu->util > sg_exynos->min &&
get_next_event_time_ms() > sg_exynos->expired_time)
return 1;
return 0;
}
static int sugov_pm_notifier(struct notifier_block *self,
unsigned long action, void *v)
{
unsigned int cpu = raw_smp_processor_id();
struct sugov_exynos *sg_exynos = &per_cpu(sugov_exynos, cpu);
struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
struct timer_list *timer = &sg_exynos->timer;
if (!sg_exynos->started)
return NOTIFY_OK;
switch (action) {
case CPU_PM_ENTER_PREPARE:
if (timer_pending(timer))
del_timer_sync(timer);
if (sugov_need_slack_timer(cpu)) {
timer->expires = jiffies + msecs_to_jiffies(sg_exynos->expired_time);
add_timer_on(timer, cpu);
trace_sugov_slack(cpu, sg_cpu->util, sg_exynos->min, action, 1);
}
break;
case CPU_PM_ENTER:
if (timer_pending(timer) && !sugov_need_slack_timer(cpu)) {
del_timer_sync(timer);
trace_sugov_slack(cpu, sg_cpu->util, sg_exynos->min, action, -1);
}
break;
case CPU_PM_EXIT_POST:
if (timer_pending(timer) && (time_after(timer->expires, jiffies))) {
del_timer_sync(timer);
trace_sugov_slack(cpu, sg_cpu->util, sg_exynos->min, action, -1);
}
break;
}
return NOTIFY_OK;
}
static struct notifier_block sugov_pm_nb = {
.notifier_call = sugov_pm_notifier,
};
static int find_cpu_pm_qos_class(int pm_qos_class)
{
int cpu;
for_each_possible_cpu(cpu) {
struct sugov_exynos *sg_exynos = &per_cpu(sugov_exynos, cpu);
if ((sg_exynos->qos_min_class == pm_qos_class) &&
cpumask_test_cpu(cpu, cpu_active_mask))
return cpu;
}
pr_err("cannot find cpu of PM QoS class\n");
return -EINVAL;
}
static int sugov_pm_qos_callback(struct notifier_block *nb,
unsigned long val, void *v)
{
struct sugov_cpu *sg_cpu;
struct cpufreq_policy *policy;
int pm_qos_class = *((int *)v);
unsigned int next_freq;
int cpu;
cpu = find_cpu_pm_qos_class(pm_qos_class);
if (cpu < 0)
return NOTIFY_BAD;
sg_cpu = &per_cpu(sugov_cpu, cpu);
if (!sg_cpu || !sg_cpu->sg_policy || !sg_cpu->sg_policy->policy)
return NOTIFY_BAD;
next_freq = sg_cpu->sg_policy->next_freq;
policy = cpufreq_cpu_get(cpu);
if (!policy)
return NOTIFY_BAD;
if (val >= policy->cur) {
cpufreq_cpu_put(policy);
return NOTIFY_BAD;
}
__cpufreq_driver_target(policy, next_freq, CPUFREQ_RELATION_L);
cpufreq_cpu_put(policy);
return NOTIFY_OK;
}
static struct notifier_block sugov_min_qos_notifier = {
.notifier_call = sugov_pm_qos_callback,
.priority = INT_MIN,
};
static int __init sugov_parse_dt(struct device_node *dn, int cpu)
{
struct sugov_exynos *sg_exynos = &per_cpu(sugov_exynos, cpu);
/* parsing slack info */
if (of_property_read_u32(dn, "enabled", &sg_exynos->enabled))
return -EINVAL;
if (sg_exynos->enabled)
if (of_property_read_u32(dn, "expired_time", &sg_exynos->expired_time))
sg_exynos->expired_time = DEFAULT_EXPIRED_TIME;
/* parsing pm_qos_class info */
if (of_property_read_u32(dn, "qos_min_class", &sg_exynos->qos_min_class))
return -EINVAL;
return 0;
}
static void __init sugov_exynos_init(void)
{
int cpu, ret;
struct device_node *dn = NULL;
const char *buf;
while ((dn = of_find_node_by_type(dn, "schedutil-domain"))) {
struct cpumask shared_mask;
/* Get shared cpus */
ret = of_property_read_string(dn, "shared-cpus", &buf);
if (ret)
goto exit;
cpulist_parse(buf, &shared_mask);
for_each_cpu(cpu, &shared_mask)
if (sugov_parse_dt(dn, cpu))
goto exit;
}
for_each_possible_cpu(cpu) {
struct sugov_exynos *sg_exynos = &per_cpu(sugov_exynos, cpu);
if (!sg_exynos->enabled)
continue;
/* Initialize slack-timer */
init_timer_pinned(&sg_exynos->timer);
sg_exynos->timer.function = sugov_nop_timer;
}
pm_qos_add_notifier(PM_QOS_CLUSTER0_FREQ_MIN, &sugov_min_qos_notifier);
pm_qos_add_notifier(PM_QOS_CLUSTER1_FREQ_MIN, &sugov_min_qos_notifier);
cpu_pm_register_notifier(&sugov_pm_nb);
return;
exit:
pr_info("%s: failed to initialized slack_timer, pm_qos handler\n", __func__);
}
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
static int __init sugov_register(void)
{
sched: schedutil: Add slack_timer and pm_qos handler 1. the slack_timer works to prevent the fequency from being held unnecessarily high when the cpu sleeps long in a high utilization. 2. The Pm_qos handler tells the governor if the frequency that was held high by qos could fall. Change-Id: I5f2d5be73b8d8294a35b2810cc9721264ec55964 Signed-off-by: Youngtae Lee <yt0729.lee@samsung.com>
2017-11-02 20:47:24 +09:00
sugov_exynos_init();
cpufreq / sched: Pass flags to cpufreq_update_util() It is useful to know the reason why cpufreq_update_util() has just been called and that can be passed as flags to cpufreq_update_util() and to the ->func() callback in struct update_util_data. However, doing that in addition to passing the util and max arguments they already take would be clumsy, so avoid it. Instead, use the observation that the schedutil governor is part of the scheduler proper, so it can access scheduler data directly. This allows the util and max arguments of cpufreq_update_util() and the ->func() callback in struct update_util_data to be replaced with a flags one, but schedutil has to be modified to follow. Thus make the schedutil governor obtain the CFS utilization information from the scheduler and use the "RT" and "DL" flags instead of the special utilization value of ULONG_MAX to track updates from the RT and DL sched classes. Make it non-modular too to avoid having to export scheduler variables to modules at large. Next, update all of the other users of cpufreq_update_util() and the ->func() callback in struct update_util_data accordingly. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2016-08-16 22:14:55 +02:00
return cpufreq_register_governor(&schedutil_gov);
}
fs_initcall(sugov_register);
Reference in a new issue Copy permalink