3 NOTE; all this assumes a linear relation between frequency and work capacity,
4 we know this is flawed, but it is the best workable approximation.
7 PELT (Per Entity Load Tracking)
8 -------------------------------
10 With PELT we track some metrics across the various scheduler entities, from
11 individual tasks to task-group slices to CPU runqueues. As the basis for this
12 we use an Exponentially Weighted Moving Average (EWMA), each period (1024us)
13 is decayed such that y^32 = 0.5. That is, the most recent 32ms contribute
14 half, while the rest of history contribute the other half.
18 ewma_sum(u) := u_0 + u_1*y + u_2*y^2 + ...
20 ewma(u) = ewma_sum(u) / ewma_sum(1)
22 Since this is essentially a progression of an infinite geometric series, the
23 results are composable, that is ewma(A) + ewma(B) = ewma(A+B). This property
24 is key, since it gives the ability to recompose the averages when tasks move
27 Note that blocked tasks still contribute to the aggregates (task-group slices
28 and CPU runqueues), which reflects their expected contribution when they
31 Using this we track 2 key metrics: 'running' and 'runnable'. 'Running'
32 reflects the time an entity spends on the CPU, while 'runnable' reflects the
33 time an entity spends on the runqueue. When there is only a single task these
34 two metrics are the same, but once there is contention for the CPU 'running'
35 will decrease to reflect the fraction of time each task spends on the CPU
36 while 'runnable' will increase to reflect the amount of contention.
38 For more detail see: kernel/sched/pelt.c
41 Frequency- / CPU Invariance
42 ---------------------------
44 Because consuming the CPU for 50% at 1GHz is not the same as consuming the CPU
45 for 50% at 2GHz, nor is running 50% on a LITTLE CPU the same as running 50% on
46 a big CPU, we allow architectures to scale the time delta with two ratios, one
47 Dynamic Voltage and Frequency Scaling (DVFS) ratio and one microarch ratio.
49 For simple DVFS architectures (where software is in full control) we trivially
56 For more dynamic systems where the hardware is in control of DVFS we use
57 hardware counters (Intel APERF/MPERF, ARMv8.4-AMU) to provide us this ratio.
58 For Intel specifically, we use:
64 4C-turbo; if available and turbo enabled
65 f_max := { 1C-turbo; if turbo enabled
69 r_dvfs := min( 1, ----- )
72 We pick 4C turbo over 1C turbo to make it slightly more sustainable.
74 r_cpu is determined as the ratio of highest performance level of the current
75 CPU vs the highest performance level of any other CPU in the system.
77 r_tot = r_dvfs * r_cpu
79 The result is that the above 'running' and 'runnable' metrics become invariant
80 of DVFS and CPU type. IOW. we can transfer and compare them between CPUs.
84 - kernel/sched/pelt.h:update_rq_clock_pelt()
85 - arch/x86/kernel/smpboot.c:"APERF/MPERF frequency ratio computation."
86 - Documentation/scheduler/sched-capacity.rst:"1. CPU Capacity + 2. Task utilization"
89 UTIL_EST / UTIL_EST_FASTUP
90 --------------------------
92 Because periodic tasks have their averages decayed while they sleep, even
93 though when running their expected utilization will be the same, they suffer a
94 (DVFS) ramp-up after they are running again.
96 To alleviate this (a default enabled option) UTIL_EST drives an Infinite
97 Impulse Response (IIR) EWMA with the 'running' value on dequeue -- when it is
98 highest. A further default enabled option UTIL_EST_FASTUP modifies the IIR
99 filter to instantly increase and only decay on decrease.
101 A further runqueue wide sum (of runnable tasks) is maintained of:
103 util_est := \Sum_t max( t_running, t_util_est_ewma )
105 For more detail see: kernel/sched/fair.c:util_est_dequeue()
111 It is possible to set effective u_min and u_max clamps on each CFS or RT task;
112 the runqueue keeps an max aggregate of these clamps for all running tasks.
114 For more detail see: include/uapi/linux/sched/types.h
120 Every time the scheduler load tracking is updated (task wakeup, task
121 migration, time progression) we call out to schedutil to update the hardware
124 The basis is the CPU runqueue's 'running' metric, which per the above it is
125 the frequency invariant utilization estimate of the CPU. From this we compute
126 a desired frequency like:
128 max( running, util_est ); if UTIL_EST
129 u_cfs := { running; otherwise
131 clamp( u_cfs + u_rt , u_min, u_max ); if UCLAMP_TASK
132 u_clamp := { u_cfs + u_rt; otherwise
134 u := u_clamp + u_irq + u_dl; [approx. see source for more detail]
136 f_des := min( f_max, 1.25 u * f_max )
138 XXX IO-wait; when the update is due to a task wakeup from IO-completion we
141 This frequency is then used to select a P-state/OPP or directly munged into a
142 CPPC style request to the hardware.
144 XXX: deadline tasks (Sporadic Task Model) allows us to calculate a hard f_min
145 required to satisfy the workload.
147 Because these callbacks are directly from the scheduler, the DVFS hardware
148 interaction should be 'fast' and non-blocking. Schedutil supports
149 rate-limiting DVFS requests for when hardware interaction is slow and
150 expensive, this reduces effectiveness.
152 For more information see: kernel/sched/cpufreq_schedutil.c
158 - On low-load scenarios, where DVFS is most relevant, the 'running' numbers
159 will closely reflect utilization.
161 - In saturated scenarios task movement will cause some transient dips,
162 suppose we have a CPU saturated with 4 tasks, then when we migrate a task
163 to an idle CPU, the old CPU will have a 'running' value of 0.75 while the
164 new CPU will gain 0.25. This is inevitable and time progression will
165 correct this. XXX do we still guarantee f_max due to no idle-time?
167 - Much of the above is about avoiding DVFS dips, and independent DVFS domains
168 having to re-learn / ramp-up when load shifts.