xref: /openbmc/linux/kernel/sched/pelt.h (revision 710b797c)
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 static inline u32 get_pelt_divider(struct sched_avg *avg)
41 {
42 	return LOAD_AVG_MAX - 1024 + avg->period_contrib;
43 }
44 
45 /*
46  * When a task is dequeued, its estimated utilization should not be update if
47  * its util_avg has not been updated at least once.
48  * This flag is used to synchronize util_avg updates with util_est updates.
49  * We map this information into the LSB bit of the utilization saved at
50  * dequeue time (i.e. util_est.dequeued).
51  */
52 #define UTIL_AVG_UNCHANGED 0x1
53 
54 static inline void cfs_se_util_change(struct sched_avg *avg)
55 {
56 	unsigned int enqueued;
57 
58 	if (!sched_feat(UTIL_EST))
59 		return;
60 
61 	/* Avoid store if the flag has been already set */
62 	enqueued = avg->util_est.enqueued;
63 	if (!(enqueued & UTIL_AVG_UNCHANGED))
64 		return;
65 
66 	/* Reset flag to report util_avg has been updated */
67 	enqueued &= ~UTIL_AVG_UNCHANGED;
68 	WRITE_ONCE(avg->util_est.enqueued, enqueued);
69 }
70 
71 /*
72  * The clock_pelt scales the time to reflect the effective amount of
73  * computation done during the running delta time but then sync back to
74  * clock_task when rq is idle.
75  *
76  *
77  * absolute time   | 1| 2| 3| 4| 5| 6| 7| 8| 9|10|11|12|13|14|15|16
78  * @ max capacity  ------******---------------******---------------
79  * @ half capacity ------************---------************---------
80  * clock pelt      | 1| 2|    3|    4| 7| 8| 9|   10|   11|14|15|16
81  *
82  */
83 static inline void update_rq_clock_pelt(struct rq *rq, s64 delta)
84 {
85 	if (unlikely(is_idle_task(rq->curr))) {
86 		/* The rq is idle, we can sync to clock_task */
87 		rq->clock_pelt  = rq_clock_task(rq);
88 		return;
89 	}
90 
91 	/*
92 	 * When a rq runs at a lower compute capacity, it will need
93 	 * more time to do the same amount of work than at max
94 	 * capacity. In order to be invariant, we scale the delta to
95 	 * reflect how much work has been really done.
96 	 * Running longer results in stealing idle time that will
97 	 * disturb the load signal compared to max capacity. This
98 	 * stolen idle time will be automatically reflected when the
99 	 * rq will be idle and the clock will be synced with
100 	 * rq_clock_task.
101 	 */
102 
103 	/*
104 	 * Scale the elapsed time to reflect the real amount of
105 	 * computation
106 	 */
107 	delta = cap_scale(delta, arch_scale_cpu_capacity(cpu_of(rq)));
108 	delta = cap_scale(delta, arch_scale_freq_capacity(cpu_of(rq)));
109 
110 	rq->clock_pelt += delta;
111 }
112 
113 /*
114  * When rq becomes idle, we have to check if it has lost idle time
115  * because it was fully busy. A rq is fully used when the /Sum util_sum
116  * is greater or equal to:
117  * (LOAD_AVG_MAX - 1024 + rq->cfs.avg.period_contrib) << SCHED_CAPACITY_SHIFT;
118  * For optimization and computing rounding purpose, we don't take into account
119  * the position in the current window (period_contrib) and we use the higher
120  * bound of util_sum to decide.
121  */
122 static inline void update_idle_rq_clock_pelt(struct rq *rq)
123 {
124 	u32 divider = ((LOAD_AVG_MAX - 1024) << SCHED_CAPACITY_SHIFT) - LOAD_AVG_MAX;
125 	u32 util_sum = rq->cfs.avg.util_sum;
126 	util_sum += rq->avg_rt.util_sum;
127 	util_sum += rq->avg_dl.util_sum;
128 
129 	/*
130 	 * Reflecting stolen time makes sense only if the idle
131 	 * phase would be present at max capacity. As soon as the
132 	 * utilization of a rq has reached the maximum value, it is
133 	 * considered as an always running rq without idle time to
134 	 * steal. This potential idle time is considered as lost in
135 	 * this case. We keep track of this lost idle time compare to
136 	 * rq's clock_task.
137 	 */
138 	if (util_sum >= divider)
139 		rq->lost_idle_time += rq_clock_task(rq) - rq->clock_pelt;
140 }
141 
142 static inline u64 rq_clock_pelt(struct rq *rq)
143 {
144 	lockdep_assert_held(&rq->lock);
145 	assert_clock_updated(rq);
146 
147 	return rq->clock_pelt - rq->lost_idle_time;
148 }
149 
150 #ifdef CONFIG_CFS_BANDWIDTH
151 /* rq->task_clock normalized against any time this cfs_rq has spent throttled */
152 static inline u64 cfs_rq_clock_pelt(struct cfs_rq *cfs_rq)
153 {
154 	if (unlikely(cfs_rq->throttle_count))
155 		return cfs_rq->throttled_clock_task - cfs_rq->throttled_clock_task_time;
156 
157 	return rq_clock_pelt(rq_of(cfs_rq)) - cfs_rq->throttled_clock_task_time;
158 }
159 #else
160 static inline u64 cfs_rq_clock_pelt(struct cfs_rq *cfs_rq)
161 {
162 	return rq_clock_pelt(rq_of(cfs_rq));
163 }
164 #endif
165 
166 #else
167 
168 static inline int
169 update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq)
170 {
171 	return 0;
172 }
173 
174 static inline int
175 update_rt_rq_load_avg(u64 now, struct rq *rq, int running)
176 {
177 	return 0;
178 }
179 
180 static inline int
181 update_dl_rq_load_avg(u64 now, struct rq *rq, int running)
182 {
183 	return 0;
184 }
185 
186 static inline int
187 update_thermal_load_avg(u64 now, struct rq *rq, u64 capacity)
188 {
189 	return 0;
190 }
191 
192 static inline u64 thermal_load_avg(struct rq *rq)
193 {
194 	return 0;
195 }
196 
197 static inline int
198 update_irq_load_avg(struct rq *rq, u64 running)
199 {
200 	return 0;
201 }
202 
203 static inline u64 rq_clock_pelt(struct rq *rq)
204 {
205 	return rq_clock_task(rq);
206 }
207 
208 static inline void
209 update_rq_clock_pelt(struct rq *rq, s64 delta) { }
210 
211 static inline void
212 update_idle_rq_clock_pelt(struct rq *rq) { }
213 
214 #endif
215 
216 
217