Wakelist can help avoid cache bouncing and offload the overhead of waker
cpu. So far, using wakelist within the same llc only happens on
WF_ON_CPU, and this limitation could be removed to further improve
wakeup performance. The result of Unixbench Pipe-based Context Switching
can be improved up to ~10%.
The 1st patch introduces a more "correct" way to check whether the wakee
cpu is soon-to-be-idle.
The 2nd patch allows using wakelist in any case when the wakee cpu is
idle or soon-to-be-idle.
v4:
Add benchmark results about the 1st patch.
Modify some comments.
v3: https://lore.kernel.org/all/[email protected]/
Add a patch to fix the check of nr_running.
After that, we can remove WF_ON_CPU to achieve the same purpose.
Thank Valentin and Mel.
v2: https://lore.kernel.org/all/[email protected]/
Modify commit log to describe key point in detail.
Add more benchmark results on more archs.
v1: https://lore.kernel.org/all/[email protected]/
Tianchen Ding (2):
sched: Fix the check of nr_running at queue wakelist
sched: Remove the limitation of WF_ON_CPU on wakelist if wakee cpu is
idle
kernel/sched/core.c | 30 ++++++++++++++++++------------
kernel/sched/sched.h | 1 -
2 files changed, 18 insertions(+), 13 deletions(-)
--
2.27.0