Comment on 3.7X speedup from removing a call to sleepparentComments−cperciva14yIn FreeBSD the vast majority of sleep lock acquisitions are uncontended and end up being a single compare-and-exchange -- very fast. Spin locks are slightly more work, since we disable interrupts during them.
Comments
In FreeBSD the vast majority of sleep lock acquisitions are uncontended and end up being a single compare-and-exchange -- very fast. Spin locks are slightly more work, since we disable interrupts during them.