1 .. SPDX-License-Identifier: GPL-2.0 <<
2 <<
3 ================ 1 ================
4 CPU Idle Cooling 2 CPU Idle Cooling
5 ================ 3 ================
6 4
7 Situation: 5 Situation:
8 ---------- 6 ----------
9 7
10 Under certain circumstances a SoC can reach a 8 Under certain circumstances a SoC can reach a critical temperature
11 limit and is unable to stabilize the temperatu 9 limit and is unable to stabilize the temperature around a temperature
12 control. When the SoC has to stabilize the tem 10 control. When the SoC has to stabilize the temperature, the kernel can
13 act on a cooling device to mitigate the dissip 11 act on a cooling device to mitigate the dissipated power. When the
14 critical temperature is reached, a decision mu 12 critical temperature is reached, a decision must be taken to reduce
15 the temperature, that, in turn impacts perform 13 the temperature, that, in turn impacts performance.
16 14
17 Another situation is when the silicon temperat 15 Another situation is when the silicon temperature continues to
18 increase even after the dynamic leakage is red 16 increase even after the dynamic leakage is reduced to its minimum by
19 clock gating the component. This runaway pheno 17 clock gating the component. This runaway phenomenon can continue due
20 to the static leakage. The only solution is to 18 to the static leakage. The only solution is to power down the
21 component, thus dropping the dynamic and stati 19 component, thus dropping the dynamic and static leakage that will
22 allow the component to cool down. 20 allow the component to cool down.
23 21
24 Last but not least, the system can ask for a s 22 Last but not least, the system can ask for a specific power budget but
25 because of the OPP density, we can only choose 23 because of the OPP density, we can only choose an OPP with a power
26 budget lower than the requested one and under- 24 budget lower than the requested one and under-utilize the CPU, thus
27 losing performance. In other words, one OPP un 25 losing performance. In other words, one OPP under-utilizes the CPU
28 with a power less than the requested power bud 26 with a power less than the requested power budget and the next OPP
29 exceeds the power budget. An intermediate OPP 27 exceeds the power budget. An intermediate OPP could have been used if
30 it were present. 28 it were present.
31 29
32 Solutions: 30 Solutions:
33 ---------- 31 ----------
34 32
35 If we can remove the static and the dynamic le 33 If we can remove the static and the dynamic leakage for a specific
36 duration in a controlled period, the SoC tempe 34 duration in a controlled period, the SoC temperature will
37 decrease. Acting on the idle state duration or 35 decrease. Acting on the idle state duration or the idle cycle
38 injection period, we can mitigate the temperat 36 injection period, we can mitigate the temperature by modulating the
39 power budget. 37 power budget.
40 38
41 The Operating Performance Point (OPP) density 39 The Operating Performance Point (OPP) density has a great influence on
42 the control precision of cpufreq, however diff 40 the control precision of cpufreq, however different vendors have a
43 plethora of OPP density, and some have large p 41 plethora of OPP density, and some have large power gap between OPPs,
44 that will result in loss of performance during 42 that will result in loss of performance during thermal control and
45 loss of power in other scenarios. 43 loss of power in other scenarios.
46 44
47 At a specific OPP, we can assume that injectin 45 At a specific OPP, we can assume that injecting idle cycle on all CPUs
48 belong to the same cluster, with a duration gr 46 belong to the same cluster, with a duration greater than the cluster
49 idle state target residency, we lead to droppi 47 idle state target residency, we lead to dropping the static and the
50 dynamic leakage for this period (modulo the en 48 dynamic leakage for this period (modulo the energy needed to enter
51 this state). So the sustainable power with idl 49 this state). So the sustainable power with idle cycles has a linear
52 relation with the OPP’s sustainable power an 50 relation with the OPP’s sustainable power and can be computed with a
53 coefficient similar to:: !! 51 coefficient similar to:
54 52
55 Power(IdleCycle) = Coef x Power(OP 53 Power(IdleCycle) = Coef x Power(OPP)
56 54
57 Idle Injection: 55 Idle Injection:
58 --------------- 56 ---------------
59 57
60 The base concept of the idle injection is to f 58 The base concept of the idle injection is to force the CPU to go to an
61 idle state for a specified time each control c 59 idle state for a specified time each control cycle, it provides
62 another way to control CPU power and heat in a 60 another way to control CPU power and heat in addition to
63 cpufreq. Ideally, if all CPUs belonging to the 61 cpufreq. Ideally, if all CPUs belonging to the same cluster, inject
64 their idle cycles synchronously, the cluster c 62 their idle cycles synchronously, the cluster can reach its power down
65 state with a minimum power consumption and red 63 state with a minimum power consumption and reduce the static leakage
66 to almost zero. However, these idle cycles in 64 to almost zero. However, these idle cycles injection will add extra
67 latencies as the CPUs will have to wakeup from 65 latencies as the CPUs will have to wakeup from a deep sleep state.
68 66
69 We use a fixed duration of idle injection that 67 We use a fixed duration of idle injection that gives an acceptable
70 performance penalty and a fixed latency. Mitig 68 performance penalty and a fixed latency. Mitigation can be increased
71 or decreased by modulating the duty cycle of t 69 or decreased by modulating the duty cycle of the idle injection.
72 70
73 :: 71 ::
74 72
75 ^ 73 ^
76 | 74 |
77 | 75 |
78 |------- ------- 76 |------- -------
79 |_______|_______________________|_______| 77 |_______|_______________________|_______|___________
80 78
81 <------> 79 <------>
82 idle <----------------------> 80 idle <---------------------->
83 running 81 running
84 82
85 <-----------------------------> 83 <----------------------------->
86 duty cycle 25% 84 duty cycle 25%
87 85
88 86
89 The implementation of the cooling device bases 87 The implementation of the cooling device bases the number of states on
90 the duty cycle percentage. When no mitigation 88 the duty cycle percentage. When no mitigation is happening the cooling
91 device state is zero, meaning the duty cycle i 89 device state is zero, meaning the duty cycle is 0%.
92 90
93 When the mitigation begins, depending on the g 91 When the mitigation begins, depending on the governor's policy, a
94 starting state is selected. With a fixed idle 92 starting state is selected. With a fixed idle duration and the duty
95 cycle (aka the cooling device state), the runn 93 cycle (aka the cooling device state), the running duration can be
96 computed. 94 computed.
97 95
98 The governor will change the cooling device st 96 The governor will change the cooling device state thus the duty cycle
99 and this variation will modulate the cooling e 97 and this variation will modulate the cooling effect.
100 98
101 :: 99 ::
102 100
103 ^ 101 ^
104 | 102 |
105 | 103 |
106 |------- ------- 104 |------- -------
107 |_______|_______________|_______|________ 105 |_______|_______________|_______|___________
108 106
109 <------> 107 <------>
110 idle <--------------> 108 idle <-------------->
111 running 109 running
112 110
113 <---------------------> 111 <--------------------->
114 duty cycle 33% 112 duty cycle 33%
115 113
116 114
117 ^ 115 ^
118 | 116 |
119 | 117 |
120 |------- ------- 118 |------- -------
121 |_______|_______|_______|___________ 119 |_______|_______|_______|___________
122 120
123 <------> 121 <------>
124 idle <------> 122 idle <------>
125 running 123 running
126 124
127 <-------------> 125 <------------->
128 duty cycle 50% 126 duty cycle 50%
129 127
130 The idle injection duration value must comply 128 The idle injection duration value must comply with the constraints:
131 129
132 - It is less than or equal to the latency we t 130 - It is less than or equal to the latency we tolerate when the
133 mitigation begins. It is platform dependent 131 mitigation begins. It is platform dependent and will depend on the
134 user experience, reactivity vs performance t 132 user experience, reactivity vs performance trade off we want. This
135 value should be specified. 133 value should be specified.
136 134
137 - It is greater than the idle state’s target 135 - It is greater than the idle state’s target residency we want to go
138 for thermal mitigation, otherwise we end up 136 for thermal mitigation, otherwise we end up consuming more energy.
139 137
140 Power considerations 138 Power considerations
141 -------------------- 139 --------------------
142 140
143 When we reach the thermal trip point, we have 141 When we reach the thermal trip point, we have to sustain a specified
144 power for a specific temperature but at this t !! 142 power for a specific temperature but at this time we consume:
145 143
146 Power = Capacitance x Voltage^2 x Frequency x 144 Power = Capacitance x Voltage^2 x Frequency x Utilisation
147 145
148 ... which is more than the sustainable power ( 146 ... which is more than the sustainable power (or there is something
149 wrong in the system setup). The ‘Capacitance 147 wrong in the system setup). The ‘Capacitance’ and ‘Utilisation’ are a
150 fixed value, ‘Voltage’ and the ‘Frequenc 148 fixed value, ‘Voltage’ and the ‘Frequency’ are fixed artificially
151 because we don’t want to change the OPP. We 149 because we don’t want to change the OPP. We can group the
152 ‘Capacitance’ and the ‘Utilisation’ in 150 ‘Capacitance’ and the ‘Utilisation’ into a single term which is the
153 ‘Dynamic Power Coefficient (Cdyn)’ Simplif !! 151 ‘Dynamic Power Coefficient (Cdyn)’ Simplifying the above, we have:
154 152
155 Pdyn = Cdyn x Voltage^2 x Frequency 153 Pdyn = Cdyn x Voltage^2 x Frequency
156 154
157 The power allocator governor will ask us someh 155 The power allocator governor will ask us somehow to reduce our power
158 in order to target the sustainable power defin 156 in order to target the sustainable power defined in the device
159 tree. So with the idle injection mechanism, we 157 tree. So with the idle injection mechanism, we want an average power
160 (Ptarget) resulting in an amount of time runni 158 (Ptarget) resulting in an amount of time running at full power on a
161 specific OPP and idle another amount of time. 159 specific OPP and idle another amount of time. That could be put in a
162 equation:: !! 160 equation:
163 161
164 P(opp)target = ((Trunning x (P(opp)running) + 162 P(opp)target = ((Trunning x (P(opp)running) + (Tidle x P(opp)idle)) /
165 (Trunning + Tidle) 163 (Trunning + Tidle)
166 164
167 ... 165 ...
168 166
169 Tidle = Trunning x ((P(opp)running / P(opp)ta 167 Tidle = Trunning x ((P(opp)running / P(opp)target) - 1)
170 168
171 At this point if we know the running period fo 169 At this point if we know the running period for the CPU, that gives us
172 the idle injection we need. Alternatively if w 170 the idle injection we need. Alternatively if we have the idle
173 injection duration, we can compute the running !! 171 injection duration, we can compute the running duration with:
174 172
175 Trunning = Tidle / ((P(opp)running / P(opp)ta 173 Trunning = Tidle / ((P(opp)running / P(opp)target) - 1)
176 174
177 Practically, if the running power is less than 175 Practically, if the running power is less than the targeted power, we
178 end up with a negative time value, so obviousl 176 end up with a negative time value, so obviously the equation usage is
179 bound to a power reduction, hence a higher OPP 177 bound to a power reduction, hence a higher OPP is needed to have the
180 running power greater than the targeted power. 178 running power greater than the targeted power.
181 179
182 However, in this demonstration we ignore three 180 However, in this demonstration we ignore three aspects:
183 181
184 * The static leakage is not defined here, we 182 * The static leakage is not defined here, we can introduce it in the
185 equation but assuming it will be zero most 183 equation but assuming it will be zero most of the time as it is
186 difficult to get the values from the SoC ve 184 difficult to get the values from the SoC vendors
187 185
188 * The idle state wake up latency (or entry + 186 * The idle state wake up latency (or entry + exit latency) is not
189 taken into account, it must be added in the 187 taken into account, it must be added in the equation in order to
190 rigorously compute the idle injection 188 rigorously compute the idle injection
191 189
192 * The injected idle duration must be greater 190 * The injected idle duration must be greater than the idle state
193 target residency, otherwise we end up consu 191 target residency, otherwise we end up consuming more energy and
194 potentially invert the mitigation effect 192 potentially invert the mitigation effect
195 193
196 So the final equation is:: !! 194 So the final equation is:
197 195
198 Trunning = (Tidle - Twakeup ) x 196 Trunning = (Tidle - Twakeup ) x
199 (((P(opp)dyn + P(opp)static ) 197 (((P(opp)dyn + P(opp)static ) - P(opp)target) / P(opp)target )
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.