1 .. SPDX-License-Identifier: GPL-2.0
2 .. include:: <isonum.txt>
4 .. |struct cpufreq_policy| replace:: :c:type:`struct cpufreq_policy <cpufreq_policy>`
5 .. |intel_pstate| replace:: :doc:`intel_pstate <intel_pstate>`
7 =======================
8 CPU Performance Scaling
9 =======================
11 :Copyright: |copy| 2017 Intel Corporation
13 :Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
16 The Concept of CPU Performance Scaling
17 ======================================
19 The majority of modern processors are capable of operating in a number of
20 different clock frequency and voltage configurations, often referred to as
21 Operating Performance Points or P-states (in ACPI terminology). As a rule,
22 the higher the clock frequency and the higher the voltage, the more instructions
23 can be retired by the CPU over a unit of time, but also the higher the clock
24 frequency and the higher the voltage, the more energy is consumed over a unit of
25 time (or the more power is drawn) by the CPU in the given P-state. Therefore
26 there is a natural tradeoff between the CPU capacity (the number of instructions
27 that can be executed over a unit of time) and the power drawn by the CPU.
29 In some situations it is desirable or even necessary to run the program as fast
30 as possible and then there is no reason to use any P-states different from the
31 highest one (i.e. the highest-performance frequency/voltage configuration
32 available). In some other cases, however, it may not be necessary to execute
33 instructions so quickly and maintaining the highest available CPU capacity for a
34 relatively long time without utilizing it entirely may be regarded as wasteful.
35 It also may not be physically possible to maintain maximum CPU capacity for too
36 long for thermal or power supply capacity reasons or similar. To cover those
37 cases, there are hardware interfaces allowing CPUs to be switched between
38 different frequency/voltage configurations or (in the ACPI terminology) to be
39 put into different P-states.
41 Typically, they are used along with algorithms to estimate the required CPU
42 capacity, so as to decide which P-states to put the CPUs into. Of course, since
43 the utilization of the system generally changes over time, that has to be done
44 repeatedly on a regular basis. The activity by which this happens is referred
45 to as CPU performance scaling or CPU frequency scaling (because it involves
46 adjusting the CPU clock frequency).
49 CPU Performance Scaling in Linux
50 ================================
52 The Linux kernel supports CPU performance scaling by means of the ``CPUFreq``
53 (CPU Frequency scaling) subsystem that consists of three layers of code: the
54 core, scaling governors and scaling drivers.
56 The ``CPUFreq`` core provides the common code infrastructure and user space
57 interfaces for all platforms that support CPU performance scaling. It defines
58 the basic framework in which the other components operate.
60 Scaling governors implement algorithms to estimate the required CPU capacity.
61 As a rule, each governor implements one, possibly parametrized, scaling
64 Scaling drivers talk to the hardware. They provide scaling governors with
65 information on the available P-states (or P-state ranges in some cases) and
66 access platform-specific hardware interfaces to change CPU P-states as requested
69 In principle, all available scaling governors can be used with every scaling
70 driver. That design is based on the observation that the information used by
71 performance scaling algorithms for P-state selection can be represented in a
72 platform-independent form in the majority of cases, so it should be possible
73 to use the same performance scaling algorithm implemented in exactly the same
74 way regardless of which scaling driver is used. Consequently, the same set of
75 scaling governors should be suitable for every supported platform.
77 However, that observation may not hold for performance scaling algorithms
78 based on information provided by the hardware itself, for example through
79 feedback registers, as that information is typically specific to the hardware
80 interface it comes from and may not be easily represented in an abstract,
81 platform-independent way. For this reason, ``CPUFreq`` allows scaling drivers
82 to bypass the governor layer and implement their own performance scaling
83 algorithms. That is done by the |intel_pstate| scaling driver.
86 ``CPUFreq`` Policy Objects
87 ==========================
89 In some cases the hardware interface for P-state control is shared by multiple
90 CPUs. That is, for example, the same register (or set of registers) is used to
91 control the P-state of multiple CPUs at the same time and writing to it affects
92 all of those CPUs simultaneously.
94 Sets of CPUs sharing hardware P-state control interfaces are represented by
95 ``CPUFreq`` as |struct cpufreq_policy| objects. For consistency,
96 |struct cpufreq_policy| is also used when there is only one CPU in the given
99 The ``CPUFreq`` core maintains a pointer to a |struct cpufreq_policy| object for
100 every CPU in the system, including CPUs that are currently offline. If multiple
101 CPUs share the same hardware P-state control interface, all of the pointers
102 corresponding to them point to the same |struct cpufreq_policy| object.
104 ``CPUFreq`` uses |struct cpufreq_policy| as its basic data type and the design
105 of its user space interface is based on the policy concept.
111 First of all, a scaling driver has to be registered for ``CPUFreq`` to work.
112 It is only possible to register one scaling driver at a time, so the scaling
113 driver is expected to be able to handle all CPUs in the system.
115 The scaling driver may be registered before or after CPU registration. If
116 CPUs are registered earlier, the driver core invokes the ``CPUFreq`` core to
117 take a note of all of the already registered CPUs during the registration of the
118 scaling driver. In turn, if any CPUs are registered after the registration of
119 the scaling driver, the ``CPUFreq`` core will be invoked to take note of them
120 at their registration time.
122 In any case, the ``CPUFreq`` core is invoked to take note of any logical CPU it
123 has not seen so far as soon as it is ready to handle that CPU. [Note that the
124 logical CPU may be a physical single-core processor, or a single core in a
125 multicore processor, or a hardware thread in a physical processor or processor
126 core. In what follows "CPU" always means "logical CPU" unless explicitly stated
127 otherwise and the word "processor" is used to refer to the physical part
128 possibly including multiple logical CPUs.]
130 Once invoked, the ``CPUFreq`` core checks if the policy pointer is already set
131 for the given CPU and if so, it skips the policy object creation. Otherwise,
132 a new policy object is created and initialized, which involves the creation of
133 a new policy directory in ``sysfs``, and the policy pointer corresponding to
134 the given CPU is set to the new policy object's address in memory.
136 Next, the scaling driver's ``->init()`` callback is invoked with the policy
137 pointer of the new CPU passed to it as the argument. That callback is expected
138 to initialize the performance scaling hardware interface for the given CPU (or,
139 more precisely, for the set of CPUs sharing the hardware interface it belongs
140 to, represented by its policy object) and, if the policy object it has been
141 called for is new, to set parameters of the policy, like the minimum and maximum
142 frequencies supported by the hardware, the table of available frequencies (if
143 the set of supported P-states is not a continuous range), and the mask of CPUs
144 that belong to the same policy (including both online and offline CPUs). That
145 mask is then used by the core to populate the policy pointers for all of the
148 The next major initialization step for a new policy object is to attach a
149 scaling governor to it (to begin with, that is the default scaling governor
150 determined by the kernel configuration, but it may be changed later
151 via ``sysfs``). First, a pointer to the new policy object is passed to the
152 governor's ``->init()`` callback which is expected to initialize all of the
153 data structures necessary to handle the given policy and, possibly, to add
154 a governor ``sysfs`` interface to it. Next, the governor is started by
155 invoking its ``->start()`` callback.
157 That callback is expected to register per-CPU utilization update callbacks for
158 all of the online CPUs belonging to the given policy with the CPU scheduler.
159 The utilization update callbacks will be invoked by the CPU scheduler on
160 important events, like task enqueue and dequeue, on every iteration of the
161 scheduler tick or generally whenever the CPU utilization may change (from the
162 scheduler's perspective). They are expected to carry out computations needed
163 to determine the P-state to use for the given policy going forward and to
164 invoke the scaling driver to make changes to the hardware in accordance with
165 the P-state selection. The scaling driver may be invoked directly from
166 scheduler context or asynchronously, via a kernel thread or workqueue, depending
167 on the configuration and capabilities of the scaling driver and the governor.
169 Similar steps are taken for policy objects that are not new, but were "inactive"
170 previously, meaning that all of the CPUs belonging to them were offline. The
171 only practical difference in that case is that the ``CPUFreq`` core will attempt
172 to use the scaling governor previously used with the policy that became
173 "inactive" (and is re-initialized now) instead of the default governor.
175 In turn, if a previously offline CPU is being brought back online, but some
176 other CPUs sharing the policy object with it are online already, there is no
177 need to re-initialize the policy object at all. In that case, it only is
178 necessary to restart the scaling governor so that it can take the new online CPU
179 into account. That is achieved by invoking the governor's ``->stop`` and
180 ``->start()`` callbacks, in this order, for the entire policy.
182 As mentioned before, the |intel_pstate| scaling driver bypasses the scaling
183 governor layer of ``CPUFreq`` and provides its own P-state selection algorithms.
184 Consequently, if |intel_pstate| is used, scaling governors are not attached to
185 new policy objects. Instead, the driver's ``->setpolicy()`` callback is invoked
186 to register per-CPU utilization update callbacks for each policy. These
187 callbacks are invoked by the CPU scheduler in the same way as for scaling
188 governors, but in the |intel_pstate| case they both determine the P-state to
189 use and change the hardware configuration accordingly in one go from scheduler
192 The policy objects created during CPU initialization and other data structures
193 associated with them are torn down when the scaling driver is unregistered
194 (which happens when the kernel module containing it is unloaded, for example) or
195 when the last CPU belonging to the given policy in unregistered.
198 Policy Interface in ``sysfs``
199 =============================
201 During the initialization of the kernel, the ``CPUFreq`` core creates a
202 ``sysfs`` directory (kobject) called ``cpufreq`` under
203 :file:`/sys/devices/system/cpu/`.
205 That directory contains a ``policyX`` subdirectory (where ``X`` represents an
206 integer number) for every policy object maintained by the ``CPUFreq`` core.
207 Each ``policyX`` directory is pointed to by ``cpufreq`` symbolic links
208 under :file:`/sys/devices/system/cpu/cpuY/` (where ``Y`` represents an integer
209 that may be different from the one represented by ``X``) for all of the CPUs
210 associated with (or belonging to) the given policy. The ``policyX`` directories
211 in :file:`/sys/devices/system/cpu/cpufreq` each contain policy-specific
212 attributes (files) to control ``CPUFreq`` behavior for the corresponding policy
213 objects (that is, for all of the CPUs associated with them).
215 Some of those attributes are generic. They are created by the ``CPUFreq`` core
216 and their behavior generally does not depend on what scaling driver is in use
217 and what scaling governor is attached to the given policy. Some scaling drivers
218 also add driver-specific attributes to the policy directories in ``sysfs`` to
219 control policy-specific aspects of driver behavior.
221 The generic attributes under :file:`/sys/devices/system/cpu/cpufreq/policyX/`
225 List of online CPUs belonging to this policy (i.e. sharing the hardware
226 performance scaling interface represented by the ``policyX`` policy
230 If the platform firmware (BIOS) tells the OS to apply an upper limit to
231 CPU frequencies, that limit will be reported through this attribute (if
234 The existence of the limit may be a result of some (often unintentional)
235 BIOS settings, restrictions coming from a service processor or another
236 BIOS/HW-based mechanisms.
238 This does not cover ACPI thermal limitations which can be discovered
239 through a generic thermal driver.
241 This attribute is not present if the scaling driver in use does not
245 Current frequency of the CPUs belonging to this policy as obtained from
246 the hardware (in KHz).
248 This is expected to be the frequency the hardware actually runs at.
249 If that frequency cannot be determined, this attribute should not
253 Maximum possible operating frequency the CPUs belonging to this policy
257 Minimum possible operating frequency the CPUs belonging to this policy
260 ``cpuinfo_transition_latency``
261 The time it takes to switch the CPUs belonging to this policy from one
262 P-state to another, in nanoseconds.
264 If unknown or if known to be so high that the scaling driver does not
265 work with the `ondemand`_ governor, -1 (:c:macro:`CPUFREQ_ETERNAL`)
266 will be returned by reads from this attribute.
269 List of all (online and offline) CPUs belonging to this policy.
271 ``scaling_available_governors``
272 List of ``CPUFreq`` scaling governors present in the kernel that can
273 be attached to this policy or (if the |intel_pstate| scaling driver is
274 in use) list of scaling algorithms provided by the driver that can be
275 applied to this policy.
277 [Note that some governors are modular and it may be necessary to load a
278 kernel module for the governor held by it to become available and be
279 listed by this attribute.]
282 Current frequency of all of the CPUs belonging to this policy (in kHz).
284 In the majority of cases, this is the frequency of the last P-state
285 requested by the scaling driver from the hardware using the scaling
286 interface provided by it, which may or may not reflect the frequency
287 the CPU is actually running at (due to hardware design and other
290 Some architectures (e.g. ``x86``) may attempt to provide information
291 more precisely reflecting the current CPU frequency through this
292 attribute, but that still may not be the exact current CPU frequency as
293 seen by the hardware at the moment.
296 The scaling driver currently in use.
299 The scaling governor currently attached to this policy or (if the
300 |intel_pstate| scaling driver is in use) the scaling algorithm
301 provided by the driver that is currently applied to this policy.
303 This attribute is read-write and writing to it will cause a new scaling
304 governor to be attached to this policy or a new scaling algorithm
305 provided by the scaling driver to be applied to it (in the
306 |intel_pstate| case), as indicated by the string written to this
307 attribute (which must be one of the names listed by the
308 ``scaling_available_governors`` attribute described above).
311 Maximum frequency the CPUs belonging to this policy are allowed to be
314 This attribute is read-write and writing a string representing an
315 integer to it will cause a new limit to be set (it must not be lower
316 than the value of the ``scaling_min_freq`` attribute).
319 Minimum frequency the CPUs belonging to this policy are allowed to be
322 This attribute is read-write and writing a string representing a
323 non-negative integer to it will cause a new limit to be set (it must not
324 be higher than the value of the ``scaling_max_freq`` attribute).
327 This attribute is functional only if the `userspace`_ scaling governor
328 is attached to the given policy.
330 It returns the last frequency requested by the governor (in kHz) or can
331 be written to in order to set a new frequency for the policy.
334 Generic Scaling Governors
335 =========================
337 ``CPUFreq`` provides generic scaling governors that can be used with all
338 scaling drivers. As stated before, each of them implements a single, possibly
339 parametrized, performance scaling algorithm.
341 Scaling governors are attached to policy objects and different policy objects
342 can be handled by different scaling governors at the same time (although that
343 may lead to suboptimal results in some cases).
345 The scaling governor for a given policy object can be changed at any time with
346 the help of the ``scaling_governor`` policy attribute in ``sysfs``.
348 Some governors expose ``sysfs`` attributes to control or fine-tune the scaling
349 algorithms implemented by them. Those attributes, referred to as governor
350 tunables, can be either global (system-wide) or per-policy, depending on the
351 scaling driver in use. If the driver requires governor tunables to be
352 per-policy, they are located in a subdirectory of each policy directory.
353 Otherwise, they are located in a subdirectory under
354 :file:`/sys/devices/system/cpu/cpufreq/`. In either case the name of the
355 subdirectory containing the governor tunables is the name of the governor
361 When attached to a policy object, this governor causes the highest frequency,
362 within the ``scaling_max_freq`` policy limit, to be requested for that policy.
364 The request is made once at that time the governor for the policy is set to
365 ``performance`` and whenever the ``scaling_max_freq`` or ``scaling_min_freq``
366 policy limits change after that.
371 When attached to a policy object, this governor causes the lowest frequency,
372 within the ``scaling_min_freq`` policy limit, to be requested for that policy.
374 The request is made once at that time the governor for the policy is set to
375 ``powersave`` and whenever the ``scaling_max_freq`` or ``scaling_min_freq``
376 policy limits change after that.
381 This governor does not do anything by itself. Instead, it allows user space
382 to set the CPU frequency for the policy it is attached to by writing to the
383 ``scaling_setspeed`` attribute of that policy.
388 This governor uses CPU utilization data available from the CPU scheduler. It
389 generally is regarded as a part of the CPU scheduler, so it can access the
390 scheduler's internal data structures directly.
392 It runs entirely in scheduler context, although in some cases it may need to
393 invoke the scaling driver asynchronously when it decides that the CPU frequency
394 should be changed for a given policy (that depends on whether or not the driver
395 is capable of changing the CPU frequency from scheduler context).
397 The actions of this governor for a particular CPU depend on the scheduling class
398 invoking its utilization update callback for that CPU. If it is invoked by the
399 RT or deadline scheduling classes, the governor will increase the frequency to
400 the allowed maximum (that is, the ``scaling_max_freq`` policy limit). In turn,
401 if it is invoked by the CFS scheduling class, the governor will use the
402 Per-Entity Load Tracking (PELT) metric for the root control group of the
403 given CPU as the CPU utilization estimate (see the *Per-entity load tracking*
404 LWN.net article [1]_ for a description of the PELT mechanism). Then, the new
405 CPU frequency to apply is computed in accordance with the formula
407 f = 1.25 * ``f_0`` * ``util`` / ``max``
409 where ``util`` is the PELT number, ``max`` is the theoretical maximum of
410 ``util``, and ``f_0`` is either the maximum possible CPU frequency for the given
411 policy (if the PELT number is frequency-invariant), or the current CPU frequency
414 This governor also employs a mechanism allowing it to temporarily bump up the
415 CPU frequency for tasks that have been waiting on I/O most recently, called
416 "IO-wait boosting". That happens when the :c:macro:`SCHED_CPUFREQ_IOWAIT` flag
417 is passed by the scheduler to the governor callback which causes the frequency
418 to go up to the allowed maximum immediately and then draw back to the value
419 returned by the above formula over time.
421 This governor exposes only one tunable:
424 Minimum time (in microseconds) that has to pass between two consecutive
425 runs of governor computations (default: 1000 times the scaling driver's
428 The purpose of this tunable is to reduce the scheduler context overhead
429 of the governor which might be excessive without it.
431 This governor generally is regarded as a replacement for the older `ondemand`_
432 and `conservative`_ governors (described below), as it is simpler and more
433 tightly integrated with the CPU scheduler, its overhead in terms of CPU context
434 switches and similar is less significant, and it uses the scheduler's own CPU
435 utilization metric, so in principle its decisions should not contradict the
436 decisions made by the other parts of the scheduler.
441 This governor uses CPU load as a CPU frequency selection metric.
443 In order to estimate the current CPU load, it measures the time elapsed between
444 consecutive invocations of its worker routine and computes the fraction of that
445 time in which the given CPU was not idle. The ratio of the non-idle (active)
446 time to the total CPU time is taken as an estimate of the load.
448 If this governor is attached to a policy shared by multiple CPUs, the load is
449 estimated for all of them and the greatest result is taken as the load estimate
450 for the entire policy.
452 The worker routine of this governor has to run in process context, so it is
453 invoked asynchronously (via a workqueue) and CPU P-states are updated from
454 there if necessary. As a result, the scheduler context overhead from this
455 governor is minimum, but it causes additional CPU context switches to happen
456 relatively often and the CPU P-state updates triggered by it can be relatively
457 irregular. Also, it affects its own CPU load metric by running code that
458 reduces the CPU idle time (even though the CPU idle time is only reduced very
461 It generally selects CPU frequencies proportional to the estimated load, so that
462 the value of the ``cpuinfo_max_freq`` policy attribute corresponds to the load of
463 1 (or 100%), and the value of the ``cpuinfo_min_freq`` policy attribute
464 corresponds to the load of 0, unless when the load exceeds a (configurable)
465 speedup threshold, in which case it will go straight for the highest frequency
466 it is allowed to use (the ``scaling_max_freq`` policy limit).
468 This governor exposes the following tunables:
471 This is how often the governor's worker routine should run, in
474 Typically, it is set to values of the order of 10000 (10 ms). Its
475 default value is equal to the value of ``cpuinfo_transition_latency``
476 for each policy this governor is attached to (but since the unit here
477 is greater by 1000, this means that the time represented by
478 ``sampling_rate`` is 1000 times greater than the transition latency by
481 If this tunable is per-policy, the following shell command sets the time
482 represented by it to be 750 times as high as the transition latency::
484 # echo `$(($(cat cpuinfo_transition_latency) * 750 / 1000)) > ondemand/sampling_rate
487 If the estimated CPU load is above this value (in percent), the governor
488 will set the frequency to the maximum value allowed for the policy.
489 Otherwise, the selected frequency will be proportional to the estimated
493 If set to 1 (default 0), it will cause the CPU load estimation code to
494 treat the CPU time spent on executing tasks with "nice" levels greater
495 than 0 as CPU idle time.
497 This may be useful if there are tasks in the system that should not be
498 taken into account when deciding what frequency to run the CPUs at.
499 Then, to make that happen it is sufficient to increase the "nice" level
500 of those tasks above 0 and set this attribute to 1.
502 ``sampling_down_factor``
503 Temporary multiplier, between 1 (default) and 100 inclusive, to apply to
504 the ``sampling_rate`` value if the CPU load goes above ``up_threshold``.
506 This causes the next execution of the governor's worker routine (after
507 setting the frequency to the allowed maximum) to be delayed, so the
508 frequency stays at the maximum level for a longer time.
510 Frequency fluctuations in some bursty workloads may be avoided this way
511 at the cost of additional energy spent on maintaining the maximum CPU
515 Reduction factor to apply to the original frequency target of the
516 governor (including the maximum value used when the ``up_threshold``
517 value is exceeded by the estimated CPU load) or sensitivity threshold
518 for the AMD frequency sensitivity powersave bias driver
519 (:file:`drivers/cpufreq/amd_freq_sensitivity.c`), between 0 and 1000
522 If the AMD frequency sensitivity powersave bias driver is not loaded,
523 the effective frequency to apply is given by
525 f * (1 - ``powersave_bias`` / 1000)
527 where f is the governor's original frequency target. The default value
528 of this attribute is 0 in that case.
530 If the AMD frequency sensitivity powersave bias driver is loaded, the
531 value of this attribute is 400 by default and it is used in a different
534 On Family 16h (and later) AMD processors there is a mechanism to get a
535 measured workload sensitivity, between 0 and 100% inclusive, from the
536 hardware. That value can be used to estimate how the performance of the
537 workload running on a CPU will change in response to frequency changes.
539 The performance of a workload with the sensitivity of 0 (memory-bound or
540 IO-bound) is not expected to increase at all as a result of increasing
541 the CPU frequency, whereas workloads with the sensitivity of 100%
542 (CPU-bound) are expected to perform much better if the CPU frequency is
545 If the workload sensitivity is less than the threshold represented by
546 the ``powersave_bias`` value, the sensitivity powersave bias driver
547 will cause the governor to select a frequency lower than its original
548 target, so as to avoid over-provisioning workloads that will not benefit
549 from running at higher CPU frequencies.
554 This governor uses CPU load as a CPU frequency selection metric.
556 It estimates the CPU load in the same way as the `ondemand`_ governor described
557 above, but the CPU frequency selection algorithm implemented by it is different.
559 Namely, it avoids changing the frequency significantly over short time intervals
560 which may not be suitable for systems with limited power supply capacity (e.g.
561 battery-powered). To achieve that, it changes the frequency in relatively
562 small steps, one step at a time, up or down - depending on whether or not a
563 (configurable) threshold has been exceeded by the estimated CPU load.
565 This governor exposes the following tunables:
568 Frequency step in percent of the maximum frequency the governor is
569 allowed to set (the ``scaling_max_freq`` policy limit), between 0 and
572 This is how much the frequency is allowed to change in one go. Setting
573 it to 0 will cause the default frequency step (5 percent) to be used
574 and setting it to 100 effectively causes the governor to periodically
575 switch the frequency between the ``scaling_min_freq`` and
576 ``scaling_max_freq`` policy limits.
579 Threshold value (in percent, 20 by default) used to determine the
580 frequency change direction.
582 If the estimated CPU load is greater than this value, the frequency will
583 go up (by ``freq_step``). If the load is less than this value (and the
584 ``sampling_down_factor`` mechanism is not in effect), the frequency will
585 go down. Otherwise, the frequency will not be changed.
587 ``sampling_down_factor``
588 Frequency decrease deferral factor, between 1 (default) and 10
591 It effectively causes the frequency to go down ``sampling_down_factor``
592 times slower than it ramps up.
595 Frequency Boost Support
596 =======================
601 Some processors support a mechanism to raise the operating frequency of some
602 cores in a multicore package temporarily (and above the sustainable frequency
603 threshold for the whole package) under certain conditions, for example if the
604 whole chip is not fully utilized and below its intended thermal or power budget.
606 Different names are used by different vendors to refer to this functionality.
607 For Intel processors it is referred to as "Turbo Boost", AMD calls it
608 "Turbo-Core" or (in technical documentation) "Core Performance Boost" and so on.
609 As a rule, it also is implemented differently by different vendors. The simple
610 term "frequency boost" is used here for brevity to refer to all of those
613 The frequency boost mechanism may be either hardware-based or software-based.
614 If it is hardware-based (e.g. on x86), the decision to trigger the boosting is
615 made by the hardware (although in general it requires the hardware to be put
616 into a special state in which it can control the CPU frequency within certain
617 limits). If it is software-based (e.g. on ARM), the scaling driver decides
618 whether or not to trigger boosting and when to do that.
620 The ``boost`` File in ``sysfs``
621 -------------------------------
623 This file is located under :file:`/sys/devices/system/cpu/cpufreq/` and controls
624 the "boost" setting for the whole system. It is not present if the underlying
625 scaling driver does not support the frequency boost mechanism (or supports it,
626 but provides a driver-specific interface for controlling it, like
629 If the value in this file is 1, the frequency boost mechanism is enabled. This
630 means that either the hardware can be put into states in which it is able to
631 trigger boosting (in the hardware-based case), or the software is allowed to
632 trigger boosting (in the software-based case). It does not mean that boosting
633 is actually in use at the moment on any CPUs in the system. It only means a
634 permission to use the frequency boost mechanism (which still may never be used
637 If the value in this file is 0, the frequency boost mechanism is disabled and
638 cannot be used at all.
640 The only values that can be written to this file are 0 and 1.
642 Rationale for Boost Control Knob
643 --------------------------------
645 The frequency boost mechanism is generally intended to help to achieve optimum
646 CPU performance on time scales below software resolution (e.g. below the
647 scheduler tick interval) and it is demonstrably suitable for many workloads, but
648 it may lead to problems in certain situations.
650 For this reason, many systems make it possible to disable the frequency boost
651 mechanism in the platform firmware (BIOS) setup, but that requires the system to
652 be restarted for the setting to be adjusted as desired, which may not be
653 practical at least in some cases. For example:
655 1. Boosting means overclocking the processor, although under controlled
656 conditions. Generally, the processor's energy consumption increases
657 as a result of increasing its frequency and voltage, even temporarily.
658 That may not be desirable on systems that switch to power sources of
659 limited capacity, such as batteries, so the ability to disable the boost
660 mechanism while the system is running may help there (but that depends on
663 2. In some situations deterministic behavior is more important than
664 performance or energy consumption (or both) and the ability to disable
665 boosting while the system is running may be useful then.
667 3. To examine the impact of the frequency boost mechanism itself, it is useful
668 to be able to run tests with and without boosting, preferably without
669 restarting the system in the meantime.
671 4. Reproducible results are important when running benchmarks. Since
672 the boosting functionality depends on the load of the whole package,
673 single-thread performance may vary because of it which may lead to
674 unreproducible results sometimes. That can be avoided by disabling the
675 frequency boost mechanism before running benchmarks sensitive to that
678 Legacy AMD ``cpb`` Knob
679 -----------------------
681 The AMD powernow-k8 scaling driver supports a ``sysfs`` knob very similar to
682 the global ``boost`` one. It is used for disabling/enabling the "Core
683 Performance Boost" feature of some AMD processors.
685 If present, that knob is located in every ``CPUFreq`` policy directory in
686 ``sysfs`` (:file:`/sys/devices/system/cpu/cpufreq/policyX/`) and is called
687 ``cpb``, which indicates a more fine grained control interface. The actual
688 implementation, however, works on the system-wide basis and setting that knob
689 for one policy causes the same value of it to be set for all of the other
690 policies at the same time.
692 That knob is still supported on AMD processors that support its underlying
693 hardware feature, but it may be configured out of the kernel (via the
694 :c:macro:`CONFIG_X86_ACPI_CPUFREQ_CPB` configuration option) and the global
695 ``boost`` knob is present regardless. Thus it is always possible use the
696 ``boost`` knob instead of the ``cpb`` one which is highly recommended, as that
697 is more consistent with what all of the other systems do (and the ``cpb`` knob
698 may not be supported any more in the future).
700 The ``cpb`` knob is never present for any processors without the underlying
701 hardware feature (e.g. all Intel ones), even if the
702 :c:macro:`CONFIG_X86_ACPI_CPUFREQ_CPB` configuration option is set.
708 .. [1] Jonathan Corbet, *Per-entity load tracking*,
709 https://lwn.net/Articles/531853/
projects / linux-2.6-microblaze.git / blob