Closed romange closed 11 months ago
All modified and coverable lines are covered by tests :white_check_mark:
Comparison is base (
6b3c2c5
) 79.29% compared to head (8df1d77
) 79.27%.
:exclamation: Your organization needs to install the Codecov GitHub app to enable full functionality.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
wait_for_cqe
as small as possible to increase the chances that another thread will notify the proactor in time. That should decrease the amount of false positives where proactor blocks and then immediately wakes up do to notifications via kernel.Some naive tests showed 5-10% reduction in proactor polling counts.
I also compared this PR (with all the other changes in helio) vs dragonfly v1.13:
run:
memtier_benchmark -s 172.31.36.182 --command=PING -t 32 -c 10 -n 300000 --hide-histogram --distinct-client-seed
from a 32 cpu machine while running dragonfly on a 8-cpu server (kernel 6.5).Dragonfly v1.13
This PR