blob: a1ce2235f121c29f3520e5504d06c22c60030f99 [file] [log] [blame]
Javi Merino6b775e82015-03-02 17:17:19 +00001Power allocator governor tunables
2=================================
3
4Trip points
5-----------
6
Javi Merino8b7b3902015-09-14 14:23:52 +01007The governor works optimally with the following two passive trip points:
Javi Merino6b775e82015-03-02 17:17:19 +00008
91. "switch on" trip point: temperature above which the governor
10 control loop starts operating. This is the first passive trip
11 point of the thermal zone.
12
132. "desired temperature" trip point: it should be higher than the
14 "switch on" trip point. This the target temperature the governor
15 is controlling for. This is the last passive trip point of the
16 thermal zone.
17
18PID Controller
19--------------
20
21The power allocator governor implements a
22Proportional-Integral-Derivative controller (PID controller) with
23temperature as the control input and power as the controlled output:
24
25 P_max = k_p * e + k_i * err_integral + k_d * diff_err + sustainable_power
26
27where
28 e = desired_temperature - current_temperature
29 err_integral is the sum of previous errors
30 diff_err = e - previous_error
31
32It is similar to the one depicted below:
33
34 k_d
35 |
36current_temp |
37 | v
38 | +----------+ +---+
39 | +----->| diff_err |-->| X |------+
40 | | +----------+ +---+ |
41 | | | tdp actor
42 | | k_i | | get_requested_power()
43 | | | | | | |
44 | | | | | | | ...
45 v | v v v v v
46 +---+ | +-------+ +---+ +---+ +---+ +----------+
47 | S |-------+----->| sum e |----->| X |--->| S |-->| S |-->|power |
48 +---+ | +-------+ +---+ +---+ +---+ |allocation|
49 ^ | ^ +----------+
50 | | | | |
51 | | +---+ | | |
52 | +------->| X |-------------------+ v v
53 | +---+ granted performance
54desired_temperature ^
55 |
56 |
57 k_po/k_pu
58
59Sustainable power
60-----------------
61
62An estimate of the sustainable dissipatable power (in mW) should be
63provided while registering the thermal zone. This estimates the
64sustained power that can be dissipated at the desired control
65temperature. This is the maximum sustained power for allocation at
66the desired maximum temperature. The actual sustained power can vary
67for a number of reasons. The closed loop controller will take care of
68variations such as environmental conditions, and some factors related
69to the speed-grade of the silicon. `sustainable_power` is therefore
70simply an estimate, and may be tuned to affect the aggressiveness of
71the thermal ramp. For reference, the sustainable power of a 4" phone
72is typically 2000mW, while on a 10" tablet is around 4500mW (may vary
73depending on screen size).
74
75If you are using device tree, do add it as a property of the
76thermal-zone. For example:
77
78 thermal-zones {
79 soc_thermal {
80 polling-delay = <1000>;
81 polling-delay-passive = <100>;
82 sustainable-power = <2500>;
83 ...
84
85Instead, if the thermal zone is registered from the platform code, pass a
86`thermal_zone_params` that has a `sustainable_power`. If no
87`thermal_zone_params` were being passed, then something like below
88will suffice:
89
90 static const struct thermal_zone_params tz_params = {
91 .sustainable_power = 3500,
92 };
93
94and then pass `tz_params` as the 5th parameter to
95`thermal_zone_device_register()`
96
97k_po and k_pu
98-------------
99
100The implementation of the PID controller in the power allocator
101thermal governor allows the configuration of two proportional term
102constants: `k_po` and `k_pu`. `k_po` is the proportional term
103constant during temperature overshoot periods (current temperature is
104above "desired temperature" trip point). Conversely, `k_pu` is the
105proportional term constant during temperature undershoot periods
106(current temperature below "desired temperature" trip point).
107
108These controls are intended as the primary mechanism for configuring
109the permitted thermal "ramp" of the system. For instance, a lower
110`k_pu` value will provide a slower ramp, at the cost of capping
111available capacity at a low temperature. On the other hand, a high
112value of `k_pu` will result in the governor granting very high power
113whilst temperature is low, and may lead to temperature overshooting.
114
115The default value for `k_pu` is:
116
117 2 * sustainable_power / (desired_temperature - switch_on_temp)
118
119This means that at `switch_on_temp` the output of the controller's
120proportional term will be 2 * `sustainable_power`. The default value
121for `k_po` is:
122
123 sustainable_power / (desired_temperature - switch_on_temp)
124
125Focusing on the proportional and feed forward values of the PID
126controller equation we have:
127
128 P_max = k_p * e + sustainable_power
129
130The proportional term is proportional to the difference between the
131desired temperature and the current one. When the current temperature
132is the desired one, then the proportional component is zero and
133`P_max` = `sustainable_power`. That is, the system should operate in
134thermal equilibrium under constant load. `sustainable_power` is only
135an estimate, which is the reason for closed-loop control such as this.
136
137Expanding `k_pu` we get:
138 P_max = 2 * sustainable_power * (T_set - T) / (T_set - T_on) +
139 sustainable_power
140
141where
142 T_set is the desired temperature
143 T is the current temperature
144 T_on is the switch on temperature
145
146When the current temperature is the switch_on temperature, the above
147formula becomes:
148
149 P_max = 2 * sustainable_power * (T_set - T_on) / (T_set - T_on) +
150 sustainable_power = 2 * sustainable_power + sustainable_power =
151 3 * sustainable_power
152
153Therefore, the proportional term alone linearly decreases power from
1543 * `sustainable_power` to `sustainable_power` as the temperature
155rises from the switch on temperature to the desired temperature.
156
157k_i and integral_cutoff
158-----------------------
159
160`k_i` configures the PID loop's integral term constant. This term
161allows the PID controller to compensate for long term drift and for
162the quantized nature of the output control: cooling devices can't set
163the exact power that the governor requests. When the temperature
164error is below `integral_cutoff`, errors are accumulated in the
165integral term. This term is then multiplied by `k_i` and the result
166added to the output of the controller. Typically `k_i` is set low (1
167or 2) and `integral_cutoff` is 0.
168
169k_d
170---
171
172`k_d` configures the PID loop's derivative term constant. It's
173recommended to leave it as the default: 0.
174
175Cooling device power API
176========================
177
178Cooling devices controlled by this governor must supply the additional
179"power" API in their `cooling_device_ops`. It consists on three ops:
180
1811. int get_requested_power(struct thermal_cooling_device *cdev,
182 struct thermal_zone_device *tz, u32 *power);
183@cdev: The `struct thermal_cooling_device` pointer
184@tz: thermal zone in which we are currently operating
185@power: pointer in which to store the calculated power
186
187`get_requested_power()` calculates the power requested by the device
188in milliwatts and stores it in @power . It should return 0 on
189success, -E* on failure. This is currently used by the power
190allocator governor to calculate how much power to give to each cooling
191device.
192
1932. int state2power(struct thermal_cooling_device *cdev, struct
194 thermal_zone_device *tz, unsigned long state, u32 *power);
195@cdev: The `struct thermal_cooling_device` pointer
196@tz: thermal zone in which we are currently operating
197@state: A cooling device state
198@power: pointer in which to store the equivalent power
199
200Convert cooling device state @state into power consumption in
201milliwatts and store it in @power. It should return 0 on success, -E*
202on failure. This is currently used by thermal core to calculate the
203maximum power that an actor can consume.
204
2053. int power2state(struct thermal_cooling_device *cdev, u32 power,
206 unsigned long *state);
207@cdev: The `struct thermal_cooling_device` pointer
208@power: power in milliwatts
209@state: pointer in which to store the resulting state
210
211Calculate a cooling device state that would make the device consume at
212most @power mW and store it in @state. It should return 0 on success,
213-E* on failure. This is currently used by the thermal core to convert
214a given power set by the power allocator governor to a state that the
215cooling device can set. It is a function because this conversion may
216depend on external factors that may change so this function should the
217best conversion given "current circumstances".
218
219Cooling device weights
220----------------------
221
222Weights are a mechanism to bias the allocation among cooling
223devices. They express the relative power efficiency of different
224cooling devices. Higher weight can be used to express higher power
225efficiency. Weighting is relative such that if each cooling device
226has a weight of one they are considered equal. This is particularly
227useful in heterogeneous systems where two cooling devices may perform
228the same kind of compute, but with different efficiency. For example,
229a system with two different types of processors.
230
231If the thermal zone is registered using
232`thermal_zone_device_register()` (i.e., platform code), then weights
233are passed as part of the thermal zone's `thermal_bind_parameters`.
234If the platform is registered using device tree, then they are passed
235as the `contribution` property of each map in the `cooling-maps` node.
236
237Limitations of the power allocator governor
238===========================================
239
240The power allocator governor's PID controller works best if there is a
241periodic tick. If you have a driver that calls
242`thermal_zone_device_update()` (or anything that ends up calling the
243governor's `throttle()` function) repetitively, the governor response
244won't be very good. Note that this is not particular to this
245governor, step-wise will also misbehave if you call its throttle()
246faster than the normal thermal framework tick (due to interrupts for
247example) as it will overreact.