1 =================================
2 Power allocator governor tunables
3 =================================
4
5 Trip points
6 -----------
7
8 The governor works optimally with the following two passive trip points:
9
10 1. "switch on" trip point: temperature above which the governor
11 control loop starts operating. This is the first passive trip
12 point of the thermal zone.
13
14 2. "desired temperature" trip point: it should be higher than the
15 "switch on" trip point. This the target temperature the governor
16 is controlling for. This is the last passive trip point of the
17 thermal zone.
18
19 PID Controller
20 --------------
21
22 The power allocator governor implements a
23 Proportional-Integral-Derivative controller (PID controller) with
24 temperature as the control input and power as the controlled output:
25
26 P_max = k_p * e + k_i * err_integral + k_d * diff_err + sustainable_power
27
28 where
29 - e = desired_temperature - current_temperature
30 - err_integral is the sum of previous errors
31 - diff_err = e - previous_error
32
33 It is similar to the one depicted below::
34
35 k_d
36 |
37 current_temp |
38 | v
39 | +----------+ +---+
40 | +----->| diff_err |-->| X |------+
41 | | +----------+ +---+ |
42 | | | tdp actor
43 | | k_i | | get_requested_power()
44 | | | | | | |
45 | | | | | | | ...
46 v | v v v v v
47 +---+ | +-------+ +---+ +---+ +---+ +----------+
48 | S |-----+----->| sum e |----->| X |--->| S |-->| S |-->|power |
49 +---+ | +-------+ +---+ +---+ +---+ |allocation|
50 ^ | ^ +----------+
51 | | | | |
52 | | +---+ | | |
53 | +------->| X |-------------------+ v v
54 | +---+ granted performance
55 desired_temperature ^
56 |
57 |
58 k_po/k_pu
59
60 Sustainable power
61 -----------------
62
63 An estimate of the sustainable dissipatable power (in mW) should be
64 provided while registering the thermal zone. This estimates the
65 sustained power that can be dissipated at the desired control
66 temperature. This is the maximum sustained power for allocation at
67 the desired maximum temperature. The actual sustained power can vary
68 for a number of reasons. The closed loop controller will take care of
69 variations such as environmental conditions, and some factors related
70 to the speed-grade of the silicon. `sustainable_power` is therefore
71 simply an estimate, and may be tuned to affect the aggressiveness of
72 the thermal ramp. For reference, the sustainable power of a 4" phone
73 is typically 2000mW, while on a 10" tablet is around 4500mW (may vary
74 depending on screen size). It is possible to have the power value
75 expressed in an abstract scale. The sustained power should be aligned
76 to the scale used by the related cooling devices.
77
78 If you are using device tree, do add it as a property of the
79 thermal-zone. For example::
80
81 thermal-zones {
82 soc_thermal {
83 polling-delay = <1000>;
84 polling-delay-passive = <100>;
85 sustainable-power = <2500>;
86 ...
87
88 Instead, if the thermal zone is registered from the platform code, pass a
89 `thermal_zone_params` that has a `sustainable_power`. If no
90 `thermal_zone_params` were being passed, then something like below
91 will suffice::
92
93 static const struct thermal_zone_params tz_params = {
94 .sustainable_power = 3500,
95 };
96
97 and then pass `tz_params` as the 5th parameter to
98 `thermal_zone_device_register()`
99
100 k_po and k_pu
101 -------------
102
103 The implementation of the PID controller in the power allocator
104 thermal governor allows the configuration of two proportional term
105 constants: `k_po` and `k_pu`. `k_po` is the proportional term
106 constant during temperature overshoot periods (current temperature is
107 above "desired temperature" trip point). Conversely, `k_pu` is the
108 proportional term constant during temperature undershoot periods
109 (current temperature below "desired temperature" trip point).
110
111 These controls are intended as the primary mechanism for configuring
112 the permitted thermal "ramp" of the system. For instance, a lower
113 `k_pu` value will provide a slower ramp, at the cost of capping
114 available capacity at a low temperature. On the other hand, a high
115 value of `k_pu` will result in the governor granting very high power
116 while temperature is low, and may lead to temperature overshooting.
117
118 The default value for `k_pu` is::
119
120 2 * sustainable_power / (desired_temperature - switch_on_temp)
121
122 This means that at `switch_on_temp` the output of the controller's
123 proportional term will be 2 * `sustainable_power`. The default value
124 for `k_po` is::
125
126 sustainable_power / (desired_temperature - switch_on_temp)
127
128 Focusing on the proportional and feed forward values of the PID
129 controller equation we have::
130
131 P_max = k_p * e + sustainable_power
132
133 The proportional term is proportional to the difference between the
134 desired temperature and the current one. When the current temperature
135 is the desired one, then the proportional component is zero and
136 `P_max` = `sustainable_power`. That is, the system should operate in
137 thermal equilibrium under constant load. `sustainable_power` is only
138 an estimate, which is the reason for closed-loop control such as this.
139
140 Expanding `k_pu` we get::
141
142 P_max = 2 * sustainable_power * (T_set - T) / (T_set - T_on) +
143 sustainable_power
144
145 where:
146
147 - T_set is the desired temperature
148 - T is the current temperature
149 - T_on is the switch on temperature
150
151 When the current temperature is the switch_on temperature, the above
152 formula becomes::
153
154 P_max = 2 * sustainable_power * (T_set - T_on) / (T_set - T_on) +
155 sustainable_power = 2 * sustainable_power + sustainable_power =
156 3 * sustainable_power
157
158 Therefore, the proportional term alone linearly decreases power from
159 3 * `sustainable_power` to `sustainable_power` as the temperature
160 rises from the switch on temperature to the desired temperature.
161
162 k_i and integral_cutoff
163 -----------------------
164
165 `k_i` configures the PID loop's integral term constant. This term
166 allows the PID controller to compensate for long term drift and for
167 the quantized nature of the output control: cooling devices can't set
168 the exact power that the governor requests. When the temperature
169 error is below `integral_cutoff`, errors are accumulated in the
170 integral term. This term is then multiplied by `k_i` and the result
171 added to the output of the controller. Typically `k_i` is set low (1
172 or 2) and `integral_cutoff` is 0.
173
174 k_d
175 ---
176
177 `k_d` configures the PID loop's derivative term constant. It's
178 recommended to leave it as the default: 0.
179
180 Cooling device power API
181 ========================
182
183 Cooling devices controlled by this governor must supply the additional
184 "power" API in their `cooling_device_ops`. It consists on three ops:
185
186 1. ::
187
188 int get_requested_power(struct thermal_cooling_device *cdev,
189 struct thermal_zone_device *tz, u32 *power);
190
191
192 @cdev:
193 The `struct thermal_cooling_device` pointer
194 @tz:
195 thermal zone in which we are currently operating
196 @power:
197 pointer in which to store the calculated power
198
199 `get_requested_power()` calculates the power requested by the device
200 in milliwatts and stores it in @power . It should return 0 on
201 success, -E* on failure. This is currently used by the power
202 allocator governor to calculate how much power to give to each cooling
203 device.
204
205 2. ::
206
207 int state2power(struct thermal_cooling_device *cdev, struct
208 thermal_zone_device *tz, unsigned long state,
209 u32 *power);
210
211 @cdev:
212 The `struct thermal_cooling_device` pointer
213 @tz:
214 thermal zone in which we are currently operating
215 @state:
216 A cooling device state
217 @power:
218 pointer in which to store the equivalent power
219
220 Convert cooling device state @state into power consumption in
221 milliwatts and store it in @power. It should return 0 on success, -E*
222 on failure. This is currently used by thermal core to calculate the
223 maximum power that an actor can consume.
224
225 3. ::
226
227 int power2state(struct thermal_cooling_device *cdev, u32 power,
228 unsigned long *state);
229
230 @cdev:
231 The `struct thermal_cooling_device` pointer
232 @power:
233 power in milliwatts
234 @state:
235 pointer in which to store the resulting state
236
237 Calculate a cooling device state that would make the device consume at
238 most @power mW and store it in @state. It should return 0 on success,
239 -E* on failure. This is currently used by the thermal core to convert
240 a given power set by the power allocator governor to a state that the
241 cooling device can set. It is a function because this conversion may
242 depend on external factors that may change so this function should the
243 best conversion given "current circumstances".
244
245 Cooling device weights
246 ----------------------
247
248 Weights are a mechanism to bias the allocation among cooling
249 devices. They express the relative power efficiency of different
250 cooling devices. Higher weight can be used to express higher power
251 efficiency. Weighting is relative such that if each cooling device
252 has a weight of one they are considered equal. This is particularly
253 useful in heterogeneous systems where two cooling devices may perform
254 the same kind of compute, but with different efficiency. For example,
255 a system with two different types of processors.
256
257 If the thermal zone is registered using
258 `thermal_zone_device_register()` (i.e., platform code), then weights
259 are passed as part of the thermal zone's `thermal_bind_parameters`.
260 If the platform is registered using device tree, then they are passed
261 as the `contribution` property of each map in the `cooling-maps` node.
262
263 Limitations of the power allocator governor
264 ===========================================
265
266 The power allocator governor's PID controller works best if there is a
267 periodic tick. If you have a driver that calls
268 `thermal_zone_device_update()` (or anything that ends up calling the
269 governor's `throttle()` function) repetitively, the governor response
270 won't be very good. Note that this is not particular to this
271 governor, step-wise will also misbehave if you call its throttle()
272 faster than the normal thermal framework tick (due to interrupts for
273 example) as it will overreact.
274
275 Energy Model requirements
276 =========================
277
278 Another important thing is the consistent scale of the power values
279 provided by the cooling devices. All of the cooling devices in a single
280 thermal zone should have power values reported either in milli-Watts
281 or scaled to the same 'abstract scale'.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.