When are lock free data structures less performant than mutual exclusion (mutexes)?

Question

I read somewhere (can\'t find the page anymore) that lock free data structures are more efficient \"for certain workloads\" which seems to imply that sometimes they\'re actually

Answer 1

lockless data structures will, one way or another, use atomic semantics from your architecture to perform its core operations. When you do this, you can be using the machines entire internal exclusion mechanisms to ensure correct ordering or fencing of data. A mutex or critical section also does this, but it only does it once for a single flag. Where the mutex or critical section is slow, is when the the lock acquisition fails (there is contention). In this case, the OS also invokes the scheduler to suspend the thread until the exclusion object has been released.

So it seems logical that whenever your lock-less data structure uses multiple atomic operations per core method when a single lock shielding a critical section could provide the same semantics AND there tends to be very little contention, in practice, for the data structure in question, then in fact, it does make more sense to use an OS-provided locking mechanism, than trying to build your own.