xref: /openbmc/linux/kernel/sched/pelt.h (revision 8365a898)
1 #ifdef CONFIG_SMP
2 #include "sched-pelt.h"
3 
4 int __update_load_avg_blocked_se(u64 now, struct sched_entity *se);
5 int __update_load_avg_se(u64 now, struct cfs_rq *cfs_rq, struct sched_entity *se);
6 int __update_load_avg_cfs_rq(u64 now, struct cfs_rq *cfs_rq);
7 int update_rt_rq_load_avg(u64 now, struct rq *rq, int running);
8 int update_dl_rq_load_avg(u64 now, struct rq *rq, int running);
9 
10 #ifdef CONFIG_SCHED_THERMAL_PRESSURE
11 int update_thermal_load_avg(u64 now, struct rq *rq, u64 capacity);
12 
13 static inline u64 thermal_load_avg(struct rq *rq)
14 {
15 	return READ_ONCE(rq->avg_thermal.load_avg);
16 }
17 #else
18 static inline int
19 update_thermal_load_avg(u64 now, struct rq *rq, u64 capacity)
20 {
21 	return 0;
22 }
23 
24 static inline u64 thermal_load_avg(struct rq *rq)
25 {
26 	return 0;
27 }
28 #endif
29 
30 #ifdef CONFIG_HAVE_SCHED_AVG_IRQ
31 int update_irq_load_avg(struct rq *rq, u64 running);
32 #else
33 static inline int
34 update_irq_load_avg(struct rq *rq, u64 running)
35 {
36 	return 0;
37 }
38 #endif
39 
40 /*
41  * When a task is dequeued, its estimated utilization should not be update if
42  * its util_avg has not been updated at least once.
43  * This flag is used to synchronize util_avg updates with util_est updates.
44  * We map this information into the LSB bit of the utilization saved at
45  * dequeue time (i.e. util_est.dequeued).
46  */
47 #define UTIL_AVG_UNCHANGED 0x1
48 
49 static inline void cfs_se_util_change(struct sched_avg *avg)
50 {
51 	unsigned int enqueued;
52 
53 	if (!sched_feat(UTIL_EST))
54 		return;
55 
56 	/* Avoid store if the flag has been already set */
57 	enqueued = avg->util_est.enqueued;
58 	if (!(enqueued & UTIL_AVG_UNCHANGED))
59 		return;
60 
61 	/* Reset flag to report util_avg has been updated */
62 	enqueued &= ~UTIL_AVG_UNCHANGED;
63 	WRITE_ONCE(avg->util_est.enqueued, enqueued);
64 }
65 
66 /*
67  * The clock_pelt scales the time to reflect the effective amount of
68  * computation done during the running delta time but then sync back to
69  * clock_task when rq is idle.
70  *
71  *
72  * absolute time   | 1| 2| 3| 4| 5| 6| 7| 8| 9|10|11|12|13|14|15|16
73  * @ max capacity  ------******---------------******---------------
74  * @ half capacity ------************---------************---------
75  * clock pelt      | 1| 2|    3|    4| 7| 8| 9|   10|   11|14|15|16
76  *
77  */
78 static inline void update_rq_clock_pelt(struct rq *rq, s64 delta)
79 {
80 	if (unlikely(is_idle_task(rq->curr))) {
81 		/* The rq is idle, we can sync to clock_task */
82 		rq->clock_pelt  = rq_clock_task(rq);
83 		return;
84 	}
85 
86 	/*
87 	 * When a rq runs at a lower compute capacity, it will need
88 	 * more time to do the same amount of work than at max
89 	 * capacity. In order to be invariant, we scale the delta to
90 	 * reflect how much work has been really done.
91 	 * Running longer results in stealing idle time that will
92 	 * disturb the load signal compared to max capacity. This
93 	 * stolen idle time will be automatically reflected when the
94 	 * rq will be idle and the clock will be synced with
95 	 * rq_clock_task.
96 	 */
97 
98 	/*
99 	 * Scale the elapsed time to reflect the real amount of
100 	 * computation
101 	 */
102 	delta = cap_scale(delta, arch_scale_cpu_capacity(cpu_of(rq)));
103 	delta = cap_scale(delta, arch_scale_freq_capacity(cpu_of(rq)));
104 
105 	rq->clock_pelt += delta;
106 }
107 
108 /*
109  * When rq becomes idle, we have to check if it has lost idle time
110  * because it was fully busy. A rq is fully used when the /Sum util_sum
111  * is greater or equal to:
112  * (LOAD_AVG_MAX - 1024 + rq->cfs.avg.period_contrib) << SCHED_CAPACITY_SHIFT;
113  * For optimization and computing rounding purpose, we don't take into account
114  * the position in the current window (period_contrib) and we use the higher
115  * bound of util_sum to decide.
116  */
117 static inline void update_idle_rq_clock_pelt(struct rq *rq)
118 {
119 	u32 divider = ((LOAD_AVG_MAX - 1024) << SCHED_CAPACITY_SHIFT) - LOAD_AVG_MAX;
120 	u32 util_sum = rq->cfs.avg.util_sum;
121 	util_sum += rq->avg_rt.util_sum;
122 	util_sum += rq->avg_dl.util_sum;
123 
124 	/*
125 	 * Reflecting stolen time makes sense only if the idle
126 	 * phase would be present at max capacity. As soon as the
127 	 * utilization of a rq has reached the maximum value, it is
128 	 * considered as an always runnig rq without idle time to
129 	 * steal. This potential idle time is considered as lost in
130 	 * this case. We keep track of this lost idle time compare to
131 	 * rq's clock_task.
132 	 */
133 	if (util_sum >= divider)
134 		rq->lost_idle_time += rq_clock_task(rq) - rq->clock_pelt;
135 }
136 
137 static inline u64 rq_clock_pelt(struct rq *rq)
138 {
139 	lockdep_assert_held(&rq->lock);
140 	assert_clock_updated(rq);
141 
142 	return rq->clock_pelt - rq->lost_idle_time;
143 }
144 
145 #ifdef CONFIG_CFS_BANDWIDTH
146 /* rq->task_clock normalized against any time this cfs_rq has spent throttled */
147 static inline u64 cfs_rq_clock_pelt(struct cfs_rq *cfs_rq)
148 {
149 	if (unlikely(cfs_rq->throttle_count))
150 		return cfs_rq->throttled_clock_task - cfs_rq->throttled_clock_task_time;
151 
152 	return rq_clock_pelt(rq_of(cfs_rq)) - cfs_rq->throttled_clock_task_time;
153 }
154 #else
155 static inline u64 cfs_rq_clock_pelt(struct cfs_rq *cfs_rq)
156 {
157 	return rq_clock_pelt(rq_of(cfs_rq));
158 }
159 #endif
160 
161 #else
162 
163 static inline int
164 update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq)
165 {
166 	return 0;
167 }
168 
169 static inline int
170 update_rt_rq_load_avg(u64 now, struct rq *rq, int running)
171 {
172 	return 0;
173 }
174 
175 static inline int
176 update_dl_rq_load_avg(u64 now, struct rq *rq, int running)
177 {
178 	return 0;
179 }
180 
181 static inline int
182 update_thermal_load_avg(u64 now, struct rq *rq, u64 capacity)
183 {
184 	return 0;
185 }
186 
187 static inline u64 thermal_load_avg(struct rq *rq)
188 {
189 	return 0;
190 }
191 
192 static inline int
193 update_irq_load_avg(struct rq *rq, u64 running)
194 {
195 	return 0;
196 }
197 
198 static inline u64 rq_clock_pelt(struct rq *rq)
199 {
200 	return rq_clock_task(rq);
201 }
202 
203 static inline void
204 update_rq_clock_pelt(struct rq *rq, s64 delta) { }
205 
206 static inline void
207 update_idle_rq_clock_pelt(struct rq *rq) { }
208 
209 #endif
210 
211 
212