Commit | Line | Data |
---|---|---|
e26af0e8 PZ |
1 | /* |
2 | * Disregards a certain amount of sleep time (sched_latency_ns) and | |
3 | * considers the task to be running during that period. This gives it | |
4 | * a service deficit on wakeup, allowing it to run sooner. | |
5 | */ | |
3f2aa307 | 6 | SCHED_FEAT(NEW_FAIR_SLEEPERS, 0) |
e26af0e8 PZ |
7 | |
8 | /* | |
9 | * By not normalizing the sleep time, heavy tasks get an effective | |
10 | * longer period, and lighter task an effective shorter period they | |
11 | * are considered running. | |
12 | */ | |
e52fb7c0 | 13 | SCHED_FEAT(NORMALIZED_SLEEPER, 0) |
e26af0e8 PZ |
14 | |
15 | /* | |
16 | * Place new tasks ahead so that they do not starve already running | |
17 | * tasks | |
18 | */ | |
f00b45c1 | 19 | SCHED_FEAT(START_DEBIT, 1) |
e26af0e8 PZ |
20 | |
21 | /* | |
22 | * Should wakeups try to preempt running tasks. | |
23 | */ | |
24 | SCHED_FEAT(WAKEUP_PREEMPT, 1) | |
25 | ||
26 | /* | |
27 | * Compute wakeup_gran based on task behaviour, clipped to | |
28 | * [0, sched_wakeup_gran_ns] | |
29 | */ | |
30 | SCHED_FEAT(ADAPTIVE_GRAN, 1) | |
31 | ||
32 | /* | |
33 | * When converting the wakeup granularity to virtual time, do it such | |
34 | * that heavier tasks preempting a lighter task have an edge. | |
35 | */ | |
36 | SCHED_FEAT(ASYM_GRAN, 1) | |
37 | ||
38 | /* | |
39 | * Always wakeup-preempt SYNC wakeups, see SYNC_WAKEUPS. | |
40 | */ | |
41 | SCHED_FEAT(WAKEUP_SYNC, 0) | |
42 | ||
43 | /* | |
44 | * Wakeup preempt based on task behaviour. Tasks that do not overlap | |
45 | * don't get preempted. | |
46 | */ | |
47 | SCHED_FEAT(WAKEUP_OVERLAP, 0) | |
48 | ||
49 | /* | |
50 | * Use the SYNC wakeup hint, pipes and the likes use this to indicate | |
51 | * the remote end is likely to consume the data we just wrote, and | |
52 | * therefore has cache benefit from being placed on the same cpu, see | |
53 | * also AFFINE_WAKEUPS. | |
54 | */ | |
55 | SCHED_FEAT(SYNC_WAKEUPS, 1) | |
56 | ||
57 | /* | |
58 | * Based on load and program behaviour, see if it makes sense to place | |
59 | * a newly woken task on the same cpu as the task that woke it -- | |
60 | * improve cache locality. Typically used with SYNC wakeups as | |
61 | * generated by pipes and the like, see also SYNC_WAKEUPS. | |
62 | */ | |
f00b45c1 | 63 | SCHED_FEAT(AFFINE_WAKEUPS, 1) |
e26af0e8 PZ |
64 | |
65 | /* | |
66 | * Prefer to schedule the task we woke last (assuming it failed | |
67 | * wakeup-preemption), since its likely going to consume data we | |
68 | * touched, increases cache locality. | |
69 | */ | |
70 | SCHED_FEAT(NEXT_BUDDY, 1) | |
71 | ||
72 | /* | |
73 | * Prefer to schedule the task that ran last (when we did | |
74 | * wake-preempt) as that likely will touch the same data, increases | |
75 | * cache locality. | |
76 | */ | |
77 | SCHED_FEAT(LAST_BUDDY, 1) | |
78 | ||
79 | /* | |
80 | * Consider buddies to be cache hot, decreases the likelyness of a | |
81 | * cache buddy being migrated away, increases cache locality. | |
82 | */ | |
f00b45c1 | 83 | SCHED_FEAT(CACHE_HOT_BUDDY, 1) |
e26af0e8 | 84 | |
0c4b83da | 85 | SCHED_FEAT(HRTICK, 0) |
f00b45c1 | 86 | SCHED_FEAT(DOUBLE_TICK, 0) |
efc2dead | 87 | SCHED_FEAT(LB_BIAS, 1) |
2398f2c6 | 88 | SCHED_FEAT(LB_WAKEUP_UPDATE, 1) |
f5bfb7d9 | 89 | SCHED_FEAT(ASYM_EFF_LOAD, 1) |
e26af0e8 PZ |
90 | |
91 | /* | |
92 | * Spin-wait on mutex acquisition when the mutex owner is running on | |
93 | * another cpu -- assumes that when the owner is running, it will soon | |
94 | * release the lock. Decreases scheduling overhead. | |
95 | */ | |
0d66bf6d | 96 | SCHED_FEAT(OWNER_SPIN, 1) |