xref: /dpdk/doc/guides/prog_guide/power_man.rst (revision b3477a6bae7e889955edd8b82f359868dc0abb01)
1..  SPDX-License-Identifier: BSD-3-Clause
2    Copyright(c) 2010-2014 Intel Corporation.
3
4Power Management
5================
6
7The DPDK Power Management feature allows users space applications to save power
8by dynamically adjusting CPU frequency or entering into different C-States.
9
10*   Adjusting the CPU frequency dynamically according to the utilization of RX queue.
11
12*   Entering into different deeper C-States according to the adaptive algorithms to speculate
13    brief periods of time suspending the application if no packets are received.
14
15The interfaces for adjusting the operating CPU frequency are in the power management library.
16C-State control is implemented in applications according to the different use cases.
17
18CPU Frequency Scaling
19---------------------
20
21The Linux kernel provides a cpufreq module for CPU frequency scaling for each lcore.
22For example, for cpuX, /sys/devices/system/cpu/cpuX/cpufreq/ has the following sys files for frequency scaling:
23
24*   affected_cpus
25
26*   bios_limit
27
28*   cpuinfo_cur_freq
29
30*   cpuinfo_max_freq
31
32*   cpuinfo_min_freq
33
34*   cpuinfo_transition_latency
35
36*   related_cpus
37
38*   scaling_available_frequencies
39
40*   scaling_available_governors
41
42*   scaling_cur_freq
43
44*   scaling_driver
45
46*   scaling_governor
47
48*   scaling_max_freq
49
50*   scaling_min_freq
51
52*   scaling_setspeed
53
54In the DPDK, scaling_governor is configured in user space.
55Then, a user space application can prompt the kernel by writing scaling_setspeed to adjust the CPU frequency
56according to the strategies defined by the user space application.
57
58Core-load Throttling through C-States
59-------------------------------------
60
61Core state can be altered by speculative sleeps whenever the specified lcore has nothing to do.
62In the DPDK, if no packet is received after polling,
63speculative sleeps can be triggered according the strategies defined by the user space application.
64
65Per-core Turbo Boost
66--------------------
67
68Individual cores can be allowed to enter a Turbo Boost state on a per-core
69basis. This is achieved by enabling Turbo Boost Technology in the BIOS, then
70looping through the relevant cores and enabling/disabling Turbo Boost on each
71core.
72
73Use of Power Library in a Hyper-Threaded Environment
74----------------------------------------------------
75
76In the case where the power library is in use on a system with Hyper-Threading enabled,
77the frequency on the physical core is set to the highest frequency of the Hyper-Thread siblings.
78So even though an application may request a scale down, the core frequency will
79remain at the highest frequency until all Hyper-Threads on that core request a scale down.
80
81API Overview of the Power Library
82---------------------------------
83
84The main methods exported by power library are for CPU frequency scaling and include the following:
85
86*   **Freq up**: Prompt the kernel to scale up the frequency of the specific lcore.
87
88*   **Freq down**: Prompt the kernel to scale down the frequency of the specific lcore.
89
90*   **Freq max**: Prompt the kernel to scale up the frequency of the specific lcore to the maximum.
91
92*   **Freq min**: Prompt the kernel to scale down the frequency of the specific lcore to the minimum.
93
94*   **Get available freqs**: Read the available frequencies of the specific lcore from the sys file.
95
96*   **Freq get**: Get the current frequency of the specific lcore.
97
98*   **Freq set**: Prompt the kernel to set the frequency for the specific lcore.
99
100*   **Enable turbo**: Prompt the kernel to enable Turbo Boost for the specific lcore.
101
102*   **Disable turbo**: Prompt the kernel to disable Turbo Boost for the specific lcore.
103
104User Cases
105----------
106
107The power management mechanism is used to save power when performing L3 forwarding.
108
109
110PM QoS
111------
112
113The ``/sys/devices/system/cpu/cpuX/power/pm_qos_resume_latency_us``
114sysfs interface is used to set and get the resume latency limit
115on the cpuX for userspace.
116Each cpuidle governor in Linux selects which idle state to enter
117based on this CPU resume latency in their idle task.
118
119The deeper the idle state, the lower the power consumption,
120but the longer the resume time.
121Some services are latency sensitive and request a low resume time,
122like interrupt packet receiving mode.
123
124Applications can set and get the CPU resume latency with
125``rte_power_qos_set_cpu_resume_latency()``
126and ``rte_power_qos_get_cpu_resume_latency()`` respectively.
127Applications can set a strict resume latency (zero value)
128to lower the resume latency and get better performance
129(instead, the power consumption of platform may increase).
130
131
132Ethernet PMD Power Management API
133---------------------------------
134
135Abstract
136~~~~~~~~
137
138Existing power management mechanisms require developers to change application
139design or change code to make use of it. The PMD power management API provides a
140convenient alternative by utilizing Ethernet PMD RX callbacks, and triggering
141power saving whenever empty poll count reaches a certain number.
142
143* Monitor
144   This power saving scheme will put the CPU into optimized power state and
145   monitor the Ethernet PMD RX descriptor address, waking the CPU up whenever
146   there's new traffic. Support for this scheme may not be available on all
147   platforms, and further limitations may apply (see below).
148
149* Pause
150   This power saving scheme will avoid busy polling by either entering
151   power-optimized sleep state with ``rte_power_pause()`` function, or, if it's
152   not supported by the underlying platform, use ``rte_pause()``.
153
154* Frequency scaling
155   This power saving scheme will use ``librte_power`` library functionality to
156   scale the core frequency up/down depending on traffic volume.
157   The reaction time of the frequency scaling mode is longer
158   than the pause and monitor mode.
159
160The "monitor" mode is only supported in the following configurations and scenarios:
161
162* On Linux* x86_64, `rte_power_monitor()` requires WAITPKG instruction set being
163  supported by the CPU, while `rte_power_monitor_multi()` requires WAITPKG and
164  RTM instruction sets being supported by the CPU. RTM instruction set may also
165  require booting the Linux with `tsx=on` command line parameter. Please refer
166  to your platform documentation for further information.
167
168* If ``rte_cpu_get_intrinsics_support()`` function indicates that
169  ``rte_power_monitor_multi()`` function is supported by the platform, then
170  monitoring multiple Ethernet Rx queues for traffic will be supported.
171
172* If ``rte_cpu_get_intrinsics_support()`` function indicates that only
173  ``rte_power_monitor()`` is supported by the platform, then monitoring will be
174  limited to a mapping of 1 core 1 queue (thus, each Rx queue will have to be
175  monitored from a different lcore).
176
177* If ``rte_cpu_get_intrinsics_support()`` function indicates that neither of the
178  two monitoring functions are supported, then monitor mode will not be supported.
179
180* Not all Ethernet drivers support monitoring, even if the underlying
181  platform may support the necessary CPU instructions. Please refer to
182  :doc:`../nics/overview` for more information.
183
184
185API Overview for Ethernet PMD Power Management
186~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
187
188* **Queue Enable**: Enable specific power scheme for certain queue/port/core.
189
190* **Queue Disable**: Disable power scheme for certain queue/port/core.
191
192* **Get Emptypoll Max**: Get the configured number of empty polls to wait before
193  entering sleep state.
194
195* **Set Emptypoll Max**: Set the number of empty polls to wait before entering
196  sleep state.
197
198* **Get Pause Duration**: Get the configured duration (microseconds) to be used
199  in the Pause callback.
200
201* **Set Pause Duration**: Set the duration of the pause (microseconds) used in
202  the Pause mode callback.
203
204* **Get Scaling Min Freq**: Get the configured minimum frequency (kHz) to be used
205  in Frequency Scaling mode.
206
207* **Set Scaling Min Freq**: Set the minimum frequency (kHz) to be used in Frequency
208  Scaling mode.
209
210* **Get Scaling Max Freq**: Get the configured maximum frequency (kHz) to be used
211  in Frequency Scaling mode.
212
213* **Set Scaling Max Freq**: Set the maximum frequency (kHz) to be used in Frequency
214  Scaling mode.
215
216Uncore API
217----------
218
219Abstract
220~~~~~~~~
221
222Uncore is a term used by Intel to describe the functions of a microprocessor
223that are not in the core, but which must be closely connected to the core
224to achieve high performance: L3 cache, on-die memory controller, etc.
225Significant power savings can be achieved by reducing the uncore frequency
226to its lowest value.
227
228Intel Uncore
229~~~~~~~~~~~~
230
231The Linux kernel provides the driver "intel-uncore-frequency"
232to control the uncore frequency limits for x86 platform.
233The driver is available from kernel version 5.6 and above.
234Also CONFIG_INTEL_UNCORE_FREQ_CONTROL will need to be enabled in the kernel,
235which was added in 5.6.
236This manipulates the context of MSR 0x620,
237which sets min/max of the uncore for the SKU.
238
239AMD EPYC Uncore
240~~~~~~~~~~~~~~~
241
242On AMD EPYC platforms, the Host System Management Port (HSMP) kernel module
243facilitates user-level access to HSMP mailboxes,
244which are implemented by the firmware in the System Management Unit (SMU).
245The AMD HSMP driver is available starting from kernel version 5.18.
246Please ensure that ``CONFIG_AMD_HSMP`` is enabled in your kernel configuration.
247
248Additionally, the EPYC System Management Interface In-band Library for Linux
249offers essential API, enabling user-space software
250to effectively manage system functions.
251
252Uncore API Overview
253~~~~~~~~~~~~~~~~~~~
254
255Overview of each function in the Uncore API,
256with explanation of what they do.
257Each function should not be called in the fast path.
258
259Uncore Power Init
260  Initialize uncore power, populate frequency array
261  and record original min & max for die on pkg.
262
263Uncore Power Exit
264  Exit uncore power, restoring original min & max for die on pkg.
265
266Get Uncore Power Freq
267  Get current uncore freq index for die on pkg.
268
269Set Uncore Power Freq
270  Set min & max uncore freq index for die on pkg
271  to specified index value (min and max will be the same).
272
273Uncore Power Max
274  Set min & max uncore freq to maximum frequency index for die on pkg
275  (min and max will be the same).
276
277Uncore Power Min
278  Set min & max uncore freq to minimum frequency index for die on pkg
279  (min and max will be the same).
280
281Get Num Freqs
282  Get the number of frequencies in the index array.
283
284Get Num Pkgs
285  Get the number of packages (CPU's) on the system.
286
287Get Num Dies
288  Get the number of die's on a given package.
289
290References
291----------
292
293*   The :doc:`../sample_app_ug/l3_forward_power_man`
294    chapter in the :doc:`../sample_app_ug/index` section.
295
296*   The :doc:`../sample_app_ug/vm_power_management`
297    chapter in the :doc:`../sample_app_ug/index` section.
298
299*   The :doc:`../nics/overview` chapter in the :doc:`../nics/index` section
300