trace:sched: Make util_avg in load_avg trace reflect PELT/WALT as used
authorChris Redpath <chris.redpath@arm.com>
Wed, 8 Feb 2017 14:25:35 +0000 (14:25 +0000)
committerAmit Pundir <amit.pundir@linaro.org>
Wed, 21 Jun 2017 11:07:44 +0000 (16:37 +0530)
commitf25df5eba6e86f2214112873ff5d2874fb681ea0
tree289224b1c05eda5732a0996c54bcd1c41f514706
parent48d818f6e34adeac57fed5ce55a92e577f3c5897
trace:sched: Make util_avg in load_avg trace reflect PELT/WALT as used

With the ability to choose between WALT and PELT for utilisation tracking
we can have the situation where we're using WALT to make all the
decisions and reporting PELT figures in the sched_load_avg_(cpu|task)
trace points. This is not too much of an issue, but when analysing trace
it is nice to see numbers representing what the scheduler is using rather
than needing to add in additional sched_walt_* traces to figure it out.

Add reporting for both types, and make the util_avg member reflect what
will be seen from cpu or task_util functions in the scheduler.

Change-Id: I2abbd2c5fa70822096d0f3372b4c12b1c6af1590
Signed-off-by: Chris Redpath <chris.redpath@arm.com>
include/trace/events/sched.h
kernel/sched/fair.c