Some of the strategy used doesn't have comments to explain what they outght to solve
e.g.
https://github.com/franz1981/Netty-VirtualThread-Scheduler/blob/master/bootstrap/src/main/java/io/netty/loom/scheduler/EventLoopScheduler.java#L368
For example this signaling while the carrier loop seems to cover for the case where a single continuation hog the loop leaving a backlog of pending continuations, and request someone to help draining them.
This is one among other N cases where it's not clear the additional signalling what's solving and it's trade-off
e.g. if a pinned poller enqueue new work and there has been some progress before it, it would over signal asking for help. It has the benefits that the carrier receive some help, but still what define being in need?
As it stands now, it is pretty much for all the work pending while the carrier is running, but it doesn't mean is chocking, it could just be well utilized.
Some of the strategy used doesn't have comments to explain what they outght to solve
e.g.
https://github.com/franz1981/Netty-VirtualThread-Scheduler/blob/master/bootstrap/src/main/java/io/netty/loom/scheduler/EventLoopScheduler.java#L368
For example this signaling while the carrier loop seems to cover for the case where a single continuation hog the loop leaving a backlog of pending continuations, and request someone to help draining them.
This is one among other N cases where it's not clear the additional signalling what's solving and it's trade-off
e.g. if a pinned poller enqueue new work and there has been some progress before it, it would over signal asking for help. It has the benefits that the carrier receive some help, but still what define being in need?
As it stands now, it is pretty much for all the work pending while the carrier is running, but it doesn't mean is chocking, it could just be well utilized.