[PATCH v2 0/3] sched/fair: Reduce contention on tg's load_avg

[Waiman Long <Waiman.Long@hpe.com>]

v1->v2:
 - Make a memcache for task_group to make sure that the allocated
   task_group object will always be on cacheline boundary even if
   debugging is turned on.
 - Scrap the original patch 3 and replace it with another one to
   disable load_avg update for root_task_group.

This patch series tries to reduce contention on task_group's load_avg
to improve system performance. It also tries to optimize the use of
idle_cpu() call in update_sg_lb_stats().

Waiman Long (3):
  sched/fair: Avoid redundant idle_cpu() call in update_sg_lb_stats()
  sched/fair: Move hot load_avg into its own cacheline
  sched/fair: Disable tg load_avg update for root_task_group

 kernel/sched/core.c  |   36 ++++++++++++++++++++++++++++++++++--
 kernel/sched/fair.c  |   16 +++++++++++++---
 kernel/sched/sched.h |    7 ++++++-
 3 files changed, 53 insertions(+), 6 deletions(-)