So ive been using atomicfu for atomics for a while...

# arrow

[Racci]

So ive been using atomicfu for atomics for a while, is there a difference in performance how they work or any benefits /downsides to arrows atomics compared to atomicfu?

[simon.vergauwen]

Biggest issue atm with atomicfu is that it's not stable, and it's hard to use for libraries.

Additionally, atomicfu cannot be used to pass around as a value. So you cannot create Atomic<A> and pass it between multiple functions, or classes.

The KDoc also states that if you're only using this for a private val and don't have a need for the other things I listed, and are not a library author, it's advised to just use atomicfu.

[Racci]

well it seems i may want to make a switch then because my main projects are apis / libraries, is there any difference in memory footprint between the two?

[simon.vergauwen]

No, it uses the same impl for JS, and Native and < JDK9.

On JDK9+ atomicfu uses AtomicFieldUpdater, which imposes the limitation that it can only be used for fields.

[simon.vergauwen]

Side-not, if you're heavily relying on atomics for primitives on the JVM there might be a slight performance upgrade to implementing your own AtomicInt, etc for MPP.

If you look at the implementation of arrow's atomic you'll find that it's pretty straight forward to do.

[Racci]

alright thanks for the info, going to start the switch and ill say arrows atomics seem to be easier to work with.

[simon.vergauwen]

If you have any more questions, I'd be happy to help ☺️