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