Join GitHub today
GitHub is home to over 36 million developers working together to host and review code, manage projects, and build software together.Sign up
Address CAF spinlock problem #380
Using the work-stealing CAF scheduler, we've observed the spinlock used by its double-ended-queue being a cpu hog. Unclear exactly why, but introducing a 1 microsecond sleep to the spinlock resolved the issue. Either need to coordinate to get that improved in upstream CAF and patched for Zeek 3.0 or else try switching to the work-sharing CAF scheduler, which doesn't use a spinlock (the work-stealing I don't think best fits our default use-case anyway: we don't need a high level of concurrency or particularly low latency).