Why `Mutex` is slower than `ReentrantLock` ?

# coroutines

[feroz_baig]

Why Mutex is slower than ReentrantLock ?

[elizarov]

@feroz_baig It depends on how you compare them.

[feroz_baig]

I ran the sample code of running 1000 coroutines each performing 1000 tasks on a shared resource. We want synchronizing to be done. Context is CommonPool.
With Mutex locks, it took around 4 sec. And with ReentrantLock it took 1 sec. Why such behaviour?

[elizarov]

ReentrantLock is more efficient and it is not abortable in the same sense that Mutex is cancellable.

[elizarov]

For short CPU-consuming action you’ll almost always get better performance with ReentrantLock, because it is what RL is optimized for

[elizarov]

Mutex is designed for the cases, where critical section has suspension points inside and thus can run for longer time.

[elizarov]

Moreover, Mutex is always fair. If you run ReentrantLock in fair mode (pass true to its constructor) it should erase some of its advantage

[feroz_baig]

I will check this out! Thanks!