admin-guide/pm/cpufreq.rst

1 .. SPDX-License-Identifier: GPL-2.0
7 CPU Performance Scaling
15 The Concept of CPU Performance Scaling
20 Operating Performance Points or P-states (in ACPI terminology).  As a rule,
24 time (or the more power is drawn) by the CPU in the given P-state.  Therefore
29 as possible and then there is no reason to use any P-states different from the
30 highest one (i.e. the highest-performance frequency/voltage configuration
38 put into different P-states.
41 capacity, so as to decide which P-states to put the CPUs into.  Of course, since
44 to as CPU performance scaling or CPU frequency scaling (because it involves
48 CPU Performance Scaling in Linux
51 The Linux kernel supports CPU performance scaling by means of the ``CPUFreq``
56 interfaces for all platforms that support CPU performance scaling.  It defines
64 information on the available P-states (or P-state ranges in some cases) and
65 access platform-specific hardware interfaces to change CPU P-states as requested
70 performance scaling algorithms for P-state selection can be represented in a
71 platform-independent form in the majority of cases, so it should be possible
72 to use the same performance scaling algorithm implemented in exactly the same
76 However, that observation may not hold for performance scaling algorithms
80 platform-independent way.  For this reason, ``CPUFreq`` allows scaling drivers
81 to bypass the governor layer and implement their own performance scaling
88 In some cases the hardware interface for P-state control is shared by multiple
90 control the P-state of multiple CPUs at the same time and writing to it affects
93 Sets of CPUs sharing hardware P-state control interfaces are represented by
100 CPUs share the same hardware P-state control interface, all of the pointers
123 logical CPU may be a physical single-core processor, or a single core in a
135 Next, the scaling driver's ``->init()`` callback is invoked with the policy
137 to initialize the performance scaling hardware interface for the given CPU (or,
142 the set of supported P-states is not a continuous range), and the mask of CPUs
151 the governor's ``->init()`` callback which is expected to initialize all of the
154 invoking its ``->start()`` callback.
156 That callback is expected to register per-CPU utilization update callbacks for
162 to determine the P-state to use for the given policy going forward and to
164 the P-state selection.  The scaling driver may be invoked directly from
172 "inactive" (and is re-initialized now) instead of the default governor.
176 need to re-initialize the policy object at all.  In that case, it only is
178 into account.  That is achieved by invoking the governor's ``->stop`` and
179 ``->start()`` callbacks, in this order, for the entire policy.
182 governor layer of ``CPUFreq`` and provides its own P-state selection algorithms.
184 new policy objects.  Instead, the driver's ``->setpolicy()`` callback is invoked
185 to register per-CPU utilization update callbacks for each policy.  These
187 governors, but in the |intel_pstate| case they both determine the P-state to
210 in :file:`/sys/devices/system/cpu/cpufreq` each contain policy-specific
217 also add driver-specific attributes to the policy directories in ``sysfs`` to
218 control policy-specific aspects of driver behavior.
225 	performance scaling interface represented by the ``policyX`` policy
235 	BIOS/HW-based mechanisms.
261 	P-state to another, in nanoseconds.
263 	If unknown or if known to be so high that the scaling driver does not
264 	work with the `ondemand`_ governor, -1 (:c:macro:`CPUFREQ_ETERNAL`)
283 	In the majority of cases, this is the frequency of the last P-state
302 	This attribute is read-write and writing to it will cause a new scaling
313 	This attribute is read-write and writing a string representing an
321 	This attribute is read-write and writing a string representing a
322 	non-negative integer to it will cause a new limit to be set (it must not
338 parametrized, performance scaling algorithm.
347 Some governors expose ``sysfs`` attributes to control or fine-tune the scaling
349 tunables, can be either global (system-wide) or per-policy, depending on the
351 per-policy, they are located in a subdirectory of each policy directory.
357 ``performance``
358 ---------------
364 ``performance`` and whenever the ``scaling_max_freq`` or ``scaling_min_freq``
368 -------------
378 -------------
385 -------------
401 Per-Entity Load Tracking (PELT) metric for the root control group of the
402 given CPU as the CPU utilization estimate (see the *Per-entity load tracking*
410 policy (if the PELT number is frequency-invariant), or the current CPU frequency
415 "IO-wait boosting".  That happens when the :c:macro:`SCHED_CPUFREQ_IOWAIT` flag
438 ------------
444 time in which the given CPU was not idle.  The ratio of the non-idle (active)
452 invoked asynchronously (via a workqueue) and CPU P-states are updated from
455 relatively often and the CPU P-state updates triggered by it can be relatively
480 	If this tunable is per-policy, the following shell command sets the time
481 	represented by it to be 750 times as high as the transition latency::
524 		f * (1 - ``powersave_bias`` / 1000)
535 	hardware.  That value can be used to estimate how the performance of the
538 	The performance of a workload with the sensitivity of 0 (memory-bound or
539 	IO-bound) is not expected to increase at all as a result of increasing
541 	(CPU-bound) are expected to perform much better if the CPU frequency is
547 	target, so as to avoid over-provisioning workloads that will not benefit
551 ----------------
560 battery-powered).  To achieve that, it changes the frequency in relatively
561 small steps, one step at a time, up or down - depending on whether or not a
594 ----------------
596 The CPUfreq governor `interactive` is designed for latency-sensitive,
605         this long before raising speed in response to continued high load.
619         If non-zero, immediately boost speed of all CPUs to at least
628         hispeed_freq according to load as usual. Its a write-only file.
640         An intermediate "high speed" at which to initially ramp
642         stays high for the amount of time specified in above_hispeed_delay,
656 ----------
665 "Turbo-Core" or (in technical documentation) "Core Performance Boost" and so on.
670 The frequency boost mechanism may be either hardware-based or software-based.
671 If it is hardware-based (e.g. on x86), the decision to trigger the boosting is
674 limits).  If it is software-based (e.g. on ARM), the scaling driver decides
678 -------------------------------
683 but provides a driver-specific interface for controlling it, like
688 trigger boosting (in the hardware-based case), or the software is allowed to
689 trigger boosting (in the software-based case).  It does not mean that boosting
700 --------------------------------
703 CPU performance on time scales below software resolution (e.g. below the
721      performance or energy consumption (or both) and the ability to disable
730      single-thread performance may vary because of it which may lead to
736 -----------------------
738 The AMD powernow-k8 scaling driver supports a ``sysfs`` knob very similar to
740 Performance Boost" feature of some AMD processors.
745 implementation, however, works on the system-wide basis and setting that knob
765 .. [1] Jonathan Corbet, *Per-entity load tracking*,