sched: EAS: take cstate into account when selecting idle core
authorSrinath Sridharan <srinathsr@google.com>
Thu, 14 Jul 2016 08:57:29 +0000 (09:57 +0100)
committerJohn Stultz <john.stultz@linaro.org>
Wed, 10 Aug 2016 22:15:39 +0000 (15:15 -0700)
commit2e9abbc942e6211f9ac4dadb14debc39bb9d1418
tree682bcb1b4e811e1d103b7c53e68c78d3169dd797
parentd753e92e19660a6adcc3b8e0076a6c6078e5f59d
sched: EAS: take cstate into account when selecting idle core

Introduce a new sysctl for this option, 'sched_cstate_aware'.
When this is enabled, select_idle_sibling in CFS is modified to
choose the idle CPU in the sibling group which has the lowest
idle state index - idle state indexes are assumed to increase
as sleep depth and hence wakeup latency increase. In this way,
we attempt to minimise wakeup latency when an idle CPU is
required.

Signed-off-by: Srinath Sridharan <srinathsr@google.com>
Includes:
sched: EAS: fix select_idle_sibling

when sysctl_sched_cstate_aware is enabled, best_idle cpu will not be chosen
in the original flow because it will goto done directly

Bug: 30107557
Change-Id: Ie09c2e3960cafbb976f8d472747faefab3b4d6ac
Signed-off-by: martin_liu <martin_liu@htc.com>
include/linux/sched/sysctl.h
kernel/sched/fair.c
kernel/sysctl.c