]> Git Repo - linux.git/commit
sched/fair: Update util_est only on util_avg updates
authorPatrick Bellasi <[email protected]>
Fri, 9 Mar 2018 09:52:45 +0000 (09:52 +0000)
committerIngo Molnar <[email protected]>
Tue, 20 Mar 2018 07:11:09 +0000 (08:11 +0100)
commitd519329f72a6f36bc4f2b85452640cfe583b4f81
tree8be90588a6d222f84108687cbdaa5cb46702e270
parenta07630b8b2c16f82fd5b71d890079f4dd7599c1d
sched/fair: Update util_est only on util_avg updates

The estimated utilization of a task is currently updated every time the
task is dequeued. However, to keep overheads under control, PELT signals
are effectively updated at maximum once every 1ms.

Thus, for really short running tasks, it can happen that their util_avg
value has not been updates since their last enqueue.  If such tasks are
also frequently running tasks (e.g. the kind of workload generated by
hackbench) it can also happen that their util_avg is updated only every
few activations.

This means that updating util_est at every dequeue potentially introduces
not necessary overheads and it's also conceptually wrong if the util_avg
signal has never been updated during a task activation.

Let's introduce a throttling mechanism on task's util_est updates
to sync them with util_avg updates. To make the solution memory
efficient, both in terms of space and load/store operations, we encode a
synchronization flag into the LSB of util_est.enqueued.
This makes util_est an even values only metric, which is still
considered good enough for its purpose.
The synchronization bit is (re)set by __update_load_avg_se() once the
PELT signal of a task has been updated during its last activation.

Such a throttling mechanism allows to keep under control util_est
overheads in the wakeup hot path, thus making it a suitable mechanism
which can be enabled also on high-intensity workload systems.
Thus, this now switches on by default the estimation utilization
scheduler feature.

Suggested-by: Chris Redpath <[email protected]>
Signed-off-by: Patrick Bellasi <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Dietmar Eggemann <[email protected]>
Cc: Joel Fernandes <[email protected]>
Cc: Juri Lelli <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Morten Rasmussen <[email protected]>
Cc: Paul Turner <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Rafael J . Wysocki <[email protected]>
Cc: Steve Muckle <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Todd Kjos <[email protected]>
Cc: Vincent Guittot <[email protected]>
Cc: Viresh Kumar <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
kernel/sched/fair.c
kernel/sched/features.h
This page took 0.057046 seconds and 4 git commands to generate.