1===============================================
2``intel_pstate`` CPU Performance Scaling Driver
3===============================================
4
5::
6
7 Copyright (c) 2017 Intel Corp., Rafael J. Wysocki <rafael.j.wysocki@intel.com>
8
9
10General Information
11===================
12
13``intel_pstate`` is a part of the
14:doc:`CPU performance scaling subsystem <cpufreq>` in the Linux kernel
15(``CPUFreq``).  It is a scaling driver for the Sandy Bridge and later
16generations of Intel processors.  Note, however, that some of those processors
17may not be supported.  [To understand ``intel_pstate`` it is necessary to know
18how ``CPUFreq`` works in general, so this is the time to read :doc:`cpufreq` if
19you have not done that yet.]
20
21For the processors supported by ``intel_pstate``, the P-state concept is broader
22than just an operating frequency or an operating performance point (see the
23`LinuxCon Europe 2015 presentation by Kristen Accardi <LCEU2015_>`_ for more
24information about that).  For this reason, the representation of P-states used
25by ``intel_pstate`` internally follows the hardware specification (for details
26refer to `Intel® 64 and IA-32 Architectures Software Developer’s Manual
27Volume 3: System Programming Guide <SDM_>`_).  However, the ``CPUFreq`` core
28uses frequencies for identifying operating performance points of CPUs and
29frequencies are involved in the user space interface exposed by it, so
30``intel_pstate`` maps its internal representation of P-states to frequencies too
31(fortunately, that mapping is unambiguous).  At the same time, it would not be
32practical for ``intel_pstate`` to supply the ``CPUFreq`` core with a table of
33available frequencies due to the possible size of it, so the driver does not do
34that.  Some functionality of the core is limited by that.
35
36Since the hardware P-state selection interface used by ``intel_pstate`` is
37available at the logical CPU level, the driver always works with individual
38CPUs.  Consequently, if ``intel_pstate`` is in use, every ``CPUFreq`` policy
39object corresponds to one logical CPU and ``CPUFreq`` policies are effectively
40equivalent to CPUs.  In particular, this means that they become "inactive" every
41time the corresponding CPU is taken offline and need to be re-initialized when
42it goes back online.
43
44``intel_pstate`` is not modular, so it cannot be unloaded, which means that the
45only way to pass early-configuration-time parameters to it is via the kernel
46command line.  However, its configuration can be adjusted via ``sysfs`` to a
47great extent.  In some configurations it even is possible to unregister it via
48``sysfs`` which allows another ``CPUFreq`` scaling driver to be loaded and
49registered (see `below <status_attr_>`_).
50
51
52Operation Modes
53===============
54
55``intel_pstate`` can operate in three different modes: in the active mode with
56or without hardware-managed P-states support and in the passive mode.  Which of
57them will be in effect depends on what kernel command line options are used and
58on the capabilities of the processor.
59
60Active Mode
61-----------
62
63This is the default operation mode of ``intel_pstate``.  If it works in this
64mode, the ``scaling_driver`` policy attribute in ``sysfs`` for all ``CPUFreq``
65policies contains the string "intel_pstate".
66
67In this mode the driver bypasses the scaling governors layer of ``CPUFreq`` and
68provides its own scaling algorithms for P-state selection.  Those algorithms
69can be applied to ``CPUFreq`` policies in the same way as generic scaling
70governors (that is, through the ``scaling_governor`` policy attribute in
71``sysfs``).  [Note that different P-state selection algorithms may be chosen for
72different policies, but that is not recommended.]
73
74They are not generic scaling governors, but their names are the same as the
75names of some of those governors.  Moreover, confusingly enough, they generally
76do not work in the same way as the generic governors they share the names with.
77For example, the ``powersave`` P-state selection algorithm provided by
78``intel_pstate`` is not a counterpart of the generic ``powersave`` governor
79(roughly, it corresponds to the ``schedutil`` and ``ondemand`` governors).
80
81There are two P-state selection algorithms provided by ``intel_pstate`` in the
82active mode: ``powersave`` and ``performance``.  The way they both operate
83depends on whether or not the hardware-managed P-states (HWP) feature has been
84enabled in the processor and possibly on the processor model.
85
86Which of the P-state selection algorithms is used by default depends on the
87:c:macro:`CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE` kernel configuration option.
88Namely, if that option is set, the ``performance`` algorithm will be used by
89default, and the other one will be used by default if it is not set.
90
91Active Mode With HWP
92~~~~~~~~~~~~~~~~~~~~
93
94If the processor supports the HWP feature, it will be enabled during the
95processor initialization and cannot be disabled after that.  It is possible
96to avoid enabling it by passing the ``intel_pstate=no_hwp`` argument to the
97kernel in the command line.
98
99If the HWP feature has been enabled, ``intel_pstate`` relies on the processor to
100select P-states by itself, but still it can give hints to the processor's
101internal P-state selection logic.  What those hints are depends on which P-state
102selection algorithm has been applied to the given policy (or to the CPU it
103corresponds to).
104
105Even though the P-state selection is carried out by the processor automatically,
106``intel_pstate`` registers utilization update callbacks with the CPU scheduler
107in this mode.  However, they are not used for running a P-state selection
108algorithm, but for periodic updates of the current CPU frequency information to
109be made available from the ``scaling_cur_freq`` policy attribute in ``sysfs``.
110
111HWP + ``performance``
112.....................
113
114In this configuration ``intel_pstate`` will write 0 to the processor's
115Energy-Performance Preference (EPP) knob (if supported) or its
116Energy-Performance Bias (EPB) knob (otherwise), which means that the processor's
117internal P-state selection logic is expected to focus entirely on performance.
118
119This will override the EPP/EPB setting coming from the ``sysfs`` interface
120(see `Energy vs Performance Hints`_ below).
121
122Also, in this configuration the range of P-states available to the processor's
123internal P-state selection logic is always restricted to the upper boundary
124(that is, the maximum P-state that the driver is allowed to use).
125
126HWP + ``powersave``
127...................
128
129In this configuration ``intel_pstate`` will set the processor's
130Energy-Performance Preference (EPP) knob (if supported) or its
131Energy-Performance Bias (EPB) knob (otherwise) to whatever value it was
132previously set to via ``sysfs`` (or whatever default value it was
133set to by the platform firmware).  This usually causes the processor's
134internal P-state selection logic to be less performance-focused.
135
136Active Mode Without HWP
137~~~~~~~~~~~~~~~~~~~~~~~
138
139This is the default operation mode for processors that do not support the HWP
140feature.  It also is used by default with the ``intel_pstate=no_hwp`` argument
141in the kernel command line.  However, in this mode ``intel_pstate`` may refuse
142to work with the given processor if it does not recognize it.  [Note that
143``intel_pstate`` will never refuse to work with any processor with the HWP
144feature enabled.]
145
146In this mode ``intel_pstate`` registers utilization update callbacks with the
147CPU scheduler in order to run a P-state selection algorithm, either
148``powersave`` or ``performance``, depending on the ``scaling_cur_freq`` policy
149setting in ``sysfs``.  The current CPU frequency information to be made
150available from the ``scaling_cur_freq`` policy attribute in ``sysfs`` is
151periodically updated by those utilization update callbacks too.
152
153``performance``
154...............
155
156Without HWP, this P-state selection algorithm is always the same regardless of
157the processor model and platform configuration.
158
159It selects the maximum P-state it is allowed to use, subject to limits set via
160``sysfs``, every time the P-state selection computations are carried out by the
161driver's utilization update callback for the given CPU (that does not happen
162more often than every 10 ms), but the hardware configuration will not be changed
163if the new P-state is the same as the current one.
164
165This is the default P-state selection algorithm if the
166:c:macro:`CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE` kernel configuration option
167is set.
168
169``powersave``
170.............
171
172Without HWP, this P-state selection algorithm generally depends on the
173processor model and/or the system profile setting in the ACPI tables and there
174are two variants of it.
175
176One of them is used with processors from the Atom line and (regardless of the
177processor model) on platforms with the system profile in the ACPI tables set to
178"mobile" (laptops mostly), "tablet", "appliance PC", "desktop", or
179"workstation".  It is also used with processors supporting the HWP feature if
180that feature has not been enabled (that is, with the ``intel_pstate=no_hwp``
181argument in the kernel command line).  It is similar to the algorithm
182implemented by the generic ``schedutil`` scaling governor except that the
183utilization metric used by it is based on numbers coming from feedback
184registers of the CPU.  It generally selects P-states proportional to the
185current CPU utilization, so it is referred to as the "proportional" algorithm.
186
187The second variant of the ``powersave`` P-state selection algorithm, used in all
188of the other cases (generally, on processors from the Core line, so it is
189referred to as the "Core" algorithm), is based on the values read from the APERF
190and MPERF feedback registers and the previously requested target P-state.
191It does not really take CPU utilization into account explicitly, but as a rule
192it causes the CPU P-state to ramp up very quickly in response to increased
193utilization which is generally desirable in server environments.
194
195Regardless of the variant, this algorithm is run by the driver's utilization
196update callback for the given CPU when it is invoked by the CPU scheduler, but
197not more often than every 10 ms (that can be tweaked via ``debugfs`` in `this
198particular case <Tuning Interface in debugfs_>`_).  Like in the ``performance``
199case, the hardware configuration is not touched if the new P-state turns out to
200be the same as the current one.
201
202This is the default P-state selection algorithm if the
203:c:macro:`CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE` kernel configuration option
204is not set.
205
206Passive Mode
207------------
208
209This mode is used if the ``intel_pstate=passive`` argument is passed to the
210kernel in the command line (it implies the ``intel_pstate=no_hwp`` setting too).
211Like in the active mode without HWP support, in this mode ``intel_pstate`` may
212refuse to work with the given processor if it does not recognize it.
213
214If the driver works in this mode, the ``scaling_driver`` policy attribute in
215``sysfs`` for all ``CPUFreq`` policies contains the string "intel_cpufreq".
216Then, the driver behaves like a regular ``CPUFreq`` scaling driver.  That is,
217it is invoked by generic scaling governors when necessary to talk to the
218hardware in order to change the P-state of a CPU (in particular, the
219``schedutil`` governor can invoke it directly from scheduler context).
220
221While in this mode, ``intel_pstate`` can be used with all of the (generic)
222scaling governors listed by the ``scaling_available_governors`` policy attribute
223in ``sysfs`` (and the P-state selection algorithms described above are not
224used).  Then, it is responsible for the configuration of policy objects
225corresponding to CPUs and provides the ``CPUFreq`` core (and the scaling
226governors attached to the policy objects) with accurate information on the
227maximum and minimum operating frequencies supported by the hardware (including
228the so-called "turbo" frequency ranges).  In other words, in the passive mode
229the entire range of available P-states is exposed by ``intel_pstate`` to the
230``CPUFreq`` core.  However, in this mode the driver does not register
231utilization update callbacks with the CPU scheduler and the ``scaling_cur_freq``
232information comes from the ``CPUFreq`` core (and is the last frequency selected
233by the current scaling governor for the given policy).
234
235
236.. _turbo:
237
238Turbo P-states Support
239======================
240
241In the majority of cases, the entire range of P-states available to
242``intel_pstate`` can be divided into two sub-ranges that correspond to
243different types of processor behavior, above and below a boundary that
244will be referred to as the "turbo threshold" in what follows.
245
246The P-states above the turbo threshold are referred to as "turbo P-states" and
247the whole sub-range of P-states they belong to is referred to as the "turbo
248range".  These names are related to the Turbo Boost technology allowing a
249multicore processor to opportunistically increase the P-state of one or more
250cores if there is enough power to do that and if that is not going to cause the
251thermal envelope of the processor package to be exceeded.
252
253Specifically, if software sets the P-state of a CPU core within the turbo range
254(that is, above the turbo threshold), the processor is permitted to take over
255performance scaling control for that core and put it into turbo P-states of its
256choice going forward.  However, that permission is interpreted differently by
257different processor generations.  Namely, the Sandy Bridge generation of
258processors will never use any P-states above the last one set by software for
259the given core, even if it is within the turbo range, whereas all of the later
260processor generations will take it as a license to use any P-states from the
261turbo range, even above the one set by software.  In other words, on those
262processors setting any P-state from the turbo range will enable the processor
263to put the given core into all turbo P-states up to and including the maximum
264supported one as it sees fit.
265
266One important property of turbo P-states is that they are not sustainable.  More
267precisely, there is no guarantee that any CPUs will be able to stay in any of
268those states indefinitely, because the power distribution within the processor
269package may change over time  or the thermal envelope it was designed for might
270be exceeded if a turbo P-state was used for too long.
271
272In turn, the P-states below the turbo threshold generally are sustainable.  In
273fact, if one of them is set by software, the processor is not expected to change
274it to a lower one unless in a thermal stress or a power limit violation
275situation (a higher P-state may still be used if it is set for another CPU in
276the same package at the same time, for example).
277
278Some processors allow multiple cores to be in turbo P-states at the same time,
279but the maximum P-state that can be set for them generally depends on the number
280of cores running concurrently.  The maximum turbo P-state that can be set for 3
281cores at the same time usually is lower than the analogous maximum P-state for
2822 cores, which in turn usually is lower than the maximum turbo P-state that can
283be set for 1 core.  The one-core maximum turbo P-state is thus the maximum
284supported one overall.
285
286The maximum supported turbo P-state, the turbo threshold (the maximum supported
287non-turbo P-state) and the minimum supported P-state are specific to the
288processor model and can be determined by reading the processor's model-specific
289registers (MSRs).  Moreover, some processors support the Configurable TDP
290(Thermal Design Power) feature and, when that feature is enabled, the turbo
291threshold effectively becomes a configurable value that can be set by the
292platform firmware.
293
294Unlike ``_PSS`` objects in the ACPI tables, ``intel_pstate`` always exposes
295the entire range of available P-states, including the whole turbo range, to the
296``CPUFreq`` core and (in the passive mode) to generic scaling governors.  This
297generally causes turbo P-states to be set more often when ``intel_pstate`` is
298used relative to ACPI-based CPU performance scaling (see `below <acpi-cpufreq_>`_
299for more information).
300
301Moreover, since ``intel_pstate`` always knows what the real turbo threshold is
302(even if the Configurable TDP feature is enabled in the processor), its
303``no_turbo`` attribute in ``sysfs`` (described `below <no_turbo_attr_>`_) should
304work as expected in all cases (that is, if set to disable turbo P-states, it
305always should prevent ``intel_pstate`` from using them).
306
307
308Processor Support
309=================
310
311To handle a given processor ``intel_pstate`` requires a number of different
312pieces of information on it to be known, including:
313
314 * The minimum supported P-state.
315
316 * The maximum supported `non-turbo P-state <turbo_>`_.
317
318 * Whether or not turbo P-states are supported at all.
319
320 * The maximum supported `one-core turbo P-state <turbo_>`_ (if turbo P-states
321   are supported).
322
323 * The scaling formula to translate the driver's internal representation
324   of P-states into frequencies and the other way around.
325
326Generally, ways to obtain that information are specific to the processor model
327or family.  Although it often is possible to obtain all of it from the processor
328itself (using model-specific registers), there are cases in which hardware
329manuals need to be consulted to get to it too.
330
331For this reason, there is a list of supported processors in ``intel_pstate`` and
332the driver initialization will fail if the detected processor is not in that
333list, unless it supports the `HWP feature <Active Mode_>`_.  [The interface to
334obtain all of the information listed above is the same for all of the processors
335supporting the HWP feature, which is why they all are supported by
336``intel_pstate``.]
337
338
339User Space Interface in ``sysfs``
340=================================
341
342Global Attributes
343-----------------
344
345``intel_pstate`` exposes several global attributes (files) in ``sysfs`` to
346control its functionality at the system level.  They are located in the
347``/sys/devices/system/cpu/cpufreq/intel_pstate/`` directory and affect all
348CPUs.
349
350Some of them are not present if the ``intel_pstate=per_cpu_perf_limits``
351argument is passed to the kernel in the command line.
352
353``max_perf_pct``
354	Maximum P-state the driver is allowed to set in percent of the
355	maximum supported performance level (the highest supported `turbo
356	P-state <turbo_>`_).
357
358	This attribute will not be exposed if the
359	``intel_pstate=per_cpu_perf_limits`` argument is present in the kernel
360	command line.
361
362``min_perf_pct``
363	Minimum P-state the driver is allowed to set in percent of the
364	maximum supported performance level (the highest supported `turbo
365	P-state <turbo_>`_).
366
367	This attribute will not be exposed if the
368	``intel_pstate=per_cpu_perf_limits`` argument is present in the kernel
369	command line.
370
371``num_pstates``
372	Number of P-states supported by the processor (between 0 and 255
373	inclusive) including both turbo and non-turbo P-states (see
374	`Turbo P-states Support`_).
375
376	The value of this attribute is not affected by the ``no_turbo``
377	setting described `below <no_turbo_attr_>`_.
378
379	This attribute is read-only.
380
381``turbo_pct``
382	Ratio of the `turbo range <turbo_>`_ size to the size of the entire
383	range of supported P-states, in percent.
384
385	This attribute is read-only.
386
387.. _no_turbo_attr:
388
389``no_turbo``
390	If set (equal to 1), the driver is not allowed to set any turbo P-states
391	(see `Turbo P-states Support`_).  If unset (equalt to 0, which is the
392	default), turbo P-states can be set by the driver.
393	[Note that ``intel_pstate`` does not support the general ``boost``
394	attribute (supported by some other scaling drivers) which is replaced
395	by this one.]
396
397	This attrubute does not affect the maximum supported frequency value
398	supplied to the ``CPUFreq`` core and exposed via the policy interface,
399	but it affects the maximum possible value of per-policy P-state	limits
400	(see `Interpretation of Policy Attributes`_ below for details).
401
402.. _status_attr:
403
404``status``
405	Operation mode of the driver: "active", "passive" or "off".
406
407	"active"
408		The driver is functional and in the `active mode
409		<Active Mode_>`_.
410
411	"passive"
412		The driver is functional and in the `passive mode
413		<Passive Mode_>`_.
414
415	"off"
416		The driver is not functional (it is not registered as a scaling
417		driver with the ``CPUFreq`` core).
418
419	This attribute can be written to in order to change the driver's
420	operation mode or to unregister it.  The string written to it must be
421	one of the possible values of it and, if successful, the write will
422	cause the driver to switch over to the operation mode represented by
423	that string - or to be unregistered in the "off" case.  [Actually,
424	switching over from the active mode to the passive mode or the other
425	way around causes the driver to be unregistered and registered again
426	with a different set of callbacks, so all of its settings (the global
427	as well as the per-policy ones) are then reset to their default
428	values, possibly depending on the target operation mode.]
429
430	That only is supported in some configurations, though (for example, if
431	the `HWP feature is enabled in the processor <Active Mode With HWP_>`_,
432	the operation mode of the driver cannot be changed), and if it is not
433	supported in the current configuration, writes to this attribute with
434	fail with an appropriate error.
435
436Interpretation of Policy Attributes
437-----------------------------------
438
439The interpretation of some ``CPUFreq`` policy attributes described in
440:doc:`cpufreq` is special with ``intel_pstate`` as the current scaling driver
441and it generally depends on the driver's `operation mode <Operation Modes_>`_.
442
443First of all, the values of the ``cpuinfo_max_freq``, ``cpuinfo_min_freq`` and
444``scaling_cur_freq`` attributes are produced by applying a processor-specific
445multiplier to the internal P-state representation used by ``intel_pstate``.
446Also, the values of the ``scaling_max_freq`` and ``scaling_min_freq``
447attributes are capped by the frequency corresponding to the maximum P-state that
448the driver is allowed to set.
449
450If the ``no_turbo`` `global attribute <no_turbo_attr_>`_ is set, the driver is
451not allowed to use turbo P-states, so the maximum value of ``scaling_max_freq``
452and ``scaling_min_freq`` is limited to the maximum non-turbo P-state frequency.
453Accordingly, setting ``no_turbo`` causes ``scaling_max_freq`` and
454``scaling_min_freq`` to go down to that value if they were above it before.
455However, the old values of ``scaling_max_freq`` and ``scaling_min_freq`` will be
456restored after unsetting ``no_turbo``, unless these attributes have been written
457to after ``no_turbo`` was set.
458
459If ``no_turbo`` is not set, the maximum possible value of ``scaling_max_freq``
460and ``scaling_min_freq`` corresponds to the maximum supported turbo P-state,
461which also is the value of ``cpuinfo_max_freq`` in either case.
462
463Next, the following policy attributes have special meaning if
464``intel_pstate`` works in the `active mode <Active Mode_>`_:
465
466``scaling_available_governors``
467	List of P-state selection algorithms provided by ``intel_pstate``.
468
469``scaling_governor``
470	P-state selection algorithm provided by ``intel_pstate`` currently in
471	use with the given policy.
472
473``scaling_cur_freq``
474	Frequency of the average P-state of the CPU represented by the given
475	policy for the time interval between the last two invocations of the
476	driver's utilization update callback by the CPU scheduler for that CPU.
477
478The meaning of these attributes in the `passive mode <Passive Mode_>`_ is the
479same as for other scaling drivers.
480
481Additionally, the value of the ``scaling_driver`` attribute for ``intel_pstate``
482depends on the operation mode of the driver.  Namely, it is either
483"intel_pstate" (in the `active mode <Active Mode_>`_) or "intel_cpufreq" (in the
484`passive mode <Passive Mode_>`_).
485
486Coordination of P-State Limits
487------------------------------
488
489``intel_pstate`` allows P-state limits to be set in two ways: with the help of
490the ``max_perf_pct`` and ``min_perf_pct`` `global attributes
491<Global Attributes_>`_ or via the ``scaling_max_freq`` and ``scaling_min_freq``
492``CPUFreq`` policy attributes.  The coordination between those limits is based
493on the following rules, regardless of the current operation mode of the driver:
494
495 1. All CPUs are affected by the global limits (that is, none of them can be
496    requested to run faster than the global maximum and none of them can be
497    requested to run slower than the global minimum).
498
499 2. Each individual CPU is affected by its own per-policy limits (that is, it
500    cannot be requested to run faster than its own per-policy maximum and it
501    cannot be requested to run slower than its own per-policy minimum).
502
503 3. The global and per-policy limits can be set independently.
504
505If the `HWP feature is enabled in the processor <Active Mode With HWP_>`_, the
506resulting effective values are written into its registers whenever the limits
507change in order to request its internal P-state selection logic to always set
508P-states within these limits.  Otherwise, the limits are taken into account by
509scaling governors (in the `passive mode <Passive Mode_>`_) and by the driver
510every time before setting a new P-state for a CPU.
511
512Additionally, if the ``intel_pstate=per_cpu_perf_limits`` command line argument
513is passed to the kernel, ``max_perf_pct`` and ``min_perf_pct`` are not exposed
514at all and the only way to set the limits is by using the policy attributes.
515
516
517Energy vs Performance Hints
518---------------------------
519
520If ``intel_pstate`` works in the `active mode with the HWP feature enabled
521<Active Mode With HWP_>`_ in the processor, additional attributes are present
522in every ``CPUFreq`` policy directory in ``sysfs``.  They are intended to allow
523user space to help ``intel_pstate`` to adjust the processor's internal P-state
524selection logic by focusing it on performance or on energy-efficiency, or
525somewhere between the two extremes:
526
527``energy_performance_preference``
528	Current value of the energy vs performance hint for the given policy
529	(or the CPU represented by it).
530
531	The hint can be changed by writing to this attribute.
532
533``energy_performance_available_preferences``
534	List of strings that can be written to the
535	``energy_performance_preference`` attribute.
536
537	They represent different energy vs performance hints and should be
538	self-explanatory, except that ``default`` represents whatever hint
539	value was set by the platform firmware.
540
541Strings written to the ``energy_performance_preference`` attribute are
542internally translated to integer values written to the processor's
543Energy-Performance Preference (EPP) knob (if supported) or its
544Energy-Performance Bias (EPB) knob.
545
546[Note that tasks may by migrated from one CPU to another by the scheduler's
547load-balancing algorithm and if different energy vs performance hints are
548set for those CPUs, that may lead to undesirable outcomes.  To avoid such
549issues it is better to set the same energy vs performance hint for all CPUs
550or to pin every task potentially sensitive to them to a specific CPU.]
551
552.. _acpi-cpufreq:
553
554``intel_pstate`` vs ``acpi-cpufreq``
555====================================
556
557On the majority of systems supported by ``intel_pstate``, the ACPI tables
558provided by the platform firmware contain ``_PSS`` objects returning information
559that can be used for CPU performance scaling (refer to the `ACPI specification`_
560for details on the ``_PSS`` objects and the format of the information returned
561by them).
562
563The information returned by the ACPI ``_PSS`` objects is used by the
564``acpi-cpufreq`` scaling driver.  On systems supported by ``intel_pstate``
565the ``acpi-cpufreq`` driver uses the same hardware CPU performance scaling
566interface, but the set of P-states it can use is limited by the ``_PSS``
567output.
568
569On those systems each ``_PSS`` object returns a list of P-states supported by
570the corresponding CPU which basically is a subset of the P-states range that can
571be used by ``intel_pstate`` on the same system, with one exception: the whole
572`turbo range <turbo_>`_ is represented by one item in it (the topmost one).  By
573convention, the frequency returned by ``_PSS`` for that item is greater by 1 MHz
574than the frequency of the highest non-turbo P-state listed by it, but the
575corresponding P-state representation (following the hardware specification)
576returned for it matches the maximum supported turbo P-state (or is the
577special value 255 meaning essentially "go as high as you can get").
578
579The list of P-states returned by ``_PSS`` is reflected by the table of
580available frequencies supplied by ``acpi-cpufreq`` to the ``CPUFreq`` core and
581scaling governors and the minimum and maximum supported frequencies reported by
582it come from that list as well.  In particular, given the special representation
583of the turbo range described above, this means that the maximum supported
584frequency reported by ``acpi-cpufreq`` is higher by 1 MHz than the frequency
585of the highest supported non-turbo P-state listed by ``_PSS`` which, of course,
586affects decisions made by the scaling governors, except for ``powersave`` and
587``performance``.
588
589For example, if a given governor attempts to select a frequency proportional to
590estimated CPU load and maps the load of 100% to the maximum supported frequency
591(possibly multiplied by a constant), then it will tend to choose P-states below
592the turbo threshold if ``acpi-cpufreq`` is used as the scaling driver, because
593in that case the turbo range corresponds to a small fraction of the frequency
594band it can use (1 MHz vs 1 GHz or more).  In consequence, it will only go to
595the turbo range for the highest loads and the other loads above 50% that might
596benefit from running at turbo frequencies will be given non-turbo P-states
597instead.
598
599One more issue related to that may appear on systems supporting the
600`Configurable TDP feature <turbo_>`_ allowing the platform firmware to set the
601turbo threshold.  Namely, if that is not coordinated with the lists of P-states
602returned by ``_PSS`` properly, there may be more than one item corresponding to
603a turbo P-state in those lists and there may be a problem with avoiding the
604turbo range (if desirable or necessary).  Usually, to avoid using turbo
605P-states overall, ``acpi-cpufreq`` simply avoids using the topmost state listed
606by ``_PSS``, but that is not sufficient when there are other turbo P-states in
607the list returned by it.
608
609Apart from the above, ``acpi-cpufreq`` works like ``intel_pstate`` in the
610`passive mode <Passive Mode_>`_, except that the number of P-states it can set
611is limited to the ones listed by the ACPI ``_PSS`` objects.
612
613
614Kernel Command Line Options for ``intel_pstate``
615================================================
616
617Several kernel command line options can be used to pass early-configuration-time
618parameters to ``intel_pstate`` in order to enforce specific behavior of it.  All
619of them have to be prepended with the ``intel_pstate=`` prefix.
620
621``disable``
622	Do not register ``intel_pstate`` as the scaling driver even if the
623	processor is supported by it.
624
625``passive``
626	Register ``intel_pstate`` in the `passive mode <Passive Mode_>`_ to
627	start with.
628
629	This option implies the ``no_hwp`` one described below.
630
631``force``
632	Register ``intel_pstate`` as the scaling driver instead of
633	``acpi-cpufreq`` even if the latter is preferred on the given system.
634
635	This may prevent some platform features (such as thermal controls and
636	power capping) that rely on the availability of ACPI P-states
637	information from functioning as expected, so it should be used with
638	caution.
639
640	This option does not work with processors that are not supported by
641	``intel_pstate`` and on platforms where the ``pcc-cpufreq`` scaling
642	driver is used instead of ``acpi-cpufreq``.
643
644``no_hwp``
645	Do not enable the `hardware-managed P-states (HWP) feature
646	<Active Mode With HWP_>`_ even if it is supported by the processor.
647
648``hwp_only``
649	Register ``intel_pstate`` as the scaling driver only if the
650	`hardware-managed P-states (HWP) feature <Active Mode With HWP_>`_ is
651	supported by the processor.
652
653``support_acpi_ppc``
654	Take ACPI ``_PPC`` performance limits into account.
655
656	If the preferred power management profile in the FADT (Fixed ACPI
657	Description Table) is set to "Enterprise Server" or "Performance
658	Server", the ACPI ``_PPC`` limits are taken into account by default
659	and this option has no effect.
660
661``per_cpu_perf_limits``
662	Use per-logical-CPU P-State limits (see `Coordination of P-state
663	Limits`_ for details).
664
665
666Diagnostics and Tuning
667======================
668
669Trace Events
670------------
671
672There are two static trace events that can be used for ``intel_pstate``
673diagnostics.  One of them is the ``cpu_frequency`` trace event generally used
674by ``CPUFreq``, and the other one is the ``pstate_sample`` trace event specific
675to ``intel_pstate``.  Both of them are triggered by ``intel_pstate`` only if
676it works in the `active mode <Active Mode_>`_.
677
678The following sequence of shell commands can be used to enable them and see
679their output (if the kernel is generally configured to support event tracing)::
680
681 # cd /sys/kernel/debug/tracing/
682 # echo 1 > events/power/pstate_sample/enable
683 # echo 1 > events/power/cpu_frequency/enable
684 # cat trace
685 gnome-terminal--4510  [001] ..s.  1177.680733: pstate_sample: core_busy=107 scaled=94 from=26 to=26 mperf=1143818 aperf=1230607 tsc=29838618 freq=2474476
686 cat-5235  [002] ..s.  1177.681723: cpu_frequency: state=2900000 cpu_id=2
687
688If ``intel_pstate`` works in the `passive mode <Passive Mode_>`_, the
689``cpu_frequency`` trace event will be triggered either by the ``schedutil``
690scaling governor (for the policies it is attached to), or by the ``CPUFreq``
691core (for the policies with other scaling governors).
692
693``ftrace``
694----------
695
696The ``ftrace`` interface can be used for low-level diagnostics of
697``intel_pstate``.  For example, to check how often the function to set a
698P-state is called, the ``ftrace`` filter can be set to to
699:c:func:`intel_pstate_set_pstate`::
700
701 # cd /sys/kernel/debug/tracing/
702 # cat available_filter_functions | grep -i pstate
703 intel_pstate_set_pstate
704 intel_pstate_cpu_init
705 ...
706 # echo intel_pstate_set_pstate > set_ftrace_filter
707 # echo function > current_tracer
708 # cat trace | head -15
709 # tracer: function
710 #
711 # entries-in-buffer/entries-written: 80/80   #P:4
712 #
713 #                              _-----=> irqs-off
714 #                             / _----=> need-resched
715 #                            | / _---=> hardirq/softirq
716 #                            || / _--=> preempt-depth
717 #                            ||| /     delay
718 #           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
719 #              | |       |   ||||       |         |
720             Xorg-3129  [000] ..s.  2537.644844: intel_pstate_set_pstate <-intel_pstate_timer_func
721  gnome-terminal--4510  [002] ..s.  2537.649844: intel_pstate_set_pstate <-intel_pstate_timer_func
722      gnome-shell-3409  [001] ..s.  2537.650850: intel_pstate_set_pstate <-intel_pstate_timer_func
723           <idle>-0     [000] ..s.  2537.654843: intel_pstate_set_pstate <-intel_pstate_timer_func
724
725Tuning Interface in ``debugfs``
726-------------------------------
727
728The ``powersave`` algorithm provided by ``intel_pstate`` for `the Core line of
729processors in the active mode <powersave_>`_ is based on a `PID controller`_
730whose parameters were chosen to address a number of different use cases at the
731same time.  However, it still is possible to fine-tune it to a specific workload
732and the ``debugfs`` interface under ``/sys/kernel/debug/pstate_snb/`` is
733provided for this purpose.  [Note that the ``pstate_snb`` directory will be
734present only if the specific P-state selection algorithm matching the interface
735in it actually is in use.]
736
737The following files present in that directory can be used to modify the PID
738controller parameters at run time:
739
740| ``deadband``
741| ``d_gain_pct``
742| ``i_gain_pct``
743| ``p_gain_pct``
744| ``sample_rate_ms``
745| ``setpoint``
746
747Note, however, that achieving desirable results this way generally requires
748expert-level understanding of the power vs performance tradeoff, so extra care
749is recommended when attempting to do that.
750
751
752.. _LCEU2015: http://events.linuxfoundation.org/sites/events/files/slides/LinuxConEurope_2015.pdf
753.. _SDM: http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-software-developer-system-programming-manual-325384.html
754.. _ACPI specification: http://www.uefi.org/sites/default/files/resources/ACPI_6_1.pdf
755.. _PID controller: https://en.wikipedia.org/wiki/PID_controller
756