You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently the benches run for mg::serverbox::TaskScheduler and for a trivial lock-based scheduler without any features. Would be interesting to implement a scheduler similar to mg::serverbox::TaskScheduler in terms of feature set but also lock-based. At least its queues should be lock-based. Don't even need to actually implement those features. Should be enough to split front- and ready-queues and have a dedicated sched-role.
Another idea would be to introduce a version of the task scheduler which has a dedicated thread just for scheduling. Without the migrating sched-role like now. So its thread count is always +1 compared to mg::serverbox::TaskScheduler and this additional thread just manages the queues.
Both these new schedulers would be quite close to the serverbox's one in a sense that potentially they could have all the same features like a binary heap with waiting tasks, task wakeups, and all.
The text was updated successfully, but these errors were encountered:
Currently the benches run for
mg::serverbox::TaskScheduler
and for a trivial lock-based scheduler without any features. Would be interesting to implement a scheduler similar tomg::serverbox::TaskScheduler
in terms of feature set but also lock-based. At least its queues should be lock-based. Don't even need to actually implement those features. Should be enough to split front- and ready-queues and have a dedicated sched-role.Another idea would be to introduce a version of the task scheduler which has a dedicated thread just for scheduling. Without the migrating sched-role like now. So its thread count is always +1 compared to
mg::serverbox::TaskScheduler
and this additional thread just manages the queues.Both these new schedulers would be quite close to the serverbox's one in a sense that potentially they could have all the same features like a binary heap with waiting tasks, task wakeups, and all.
The text was updated successfully, but these errors were encountered: