Question

Possible Duplicate:
Objective-C: With ARC, what's better? alloc or autorelease initializers?

Does ARC automatically turn the autoreleased versions of class initializers into the appropriate non-autorelased versions, or are they still technically being autoreleased?

I don’t want to keep memory around any longer than it’s absolutely required, so I’ve gotten in the habit of using alloc/init in almost all circumstances. Now in ARC, I’m wondering if I can just start using the “autorelease” initializers and expect them to act like a non-autorelased versions would behave…

Does anyone have any documentation on where I can find out what happens to autoreleased methods under ARC?

Answer 1

When you get an autoreleased object, ARC will manage to avoid the autorelease pool, as long as both your code and the called method/function are compiled with ARC.

ARC adds a call to objc_retainAutoreleasedReturnValue in your code and a call to objc_autoreleaseReturnValue in the called function/method. At runtime when objc_autoreleaseReturnValue sees that the returned value will be retained by objc_retainAutoreleaseReturnValue, it doesn’t autorelease the object and sets a flag to tell objc_retainAutoreleaseReturnValue not to retain the object. Thus you get no (perceptible) extra cost for using a convenient creation method rather that alloc/init.

For more information about that mechanism, you may read How does objc_retainAutoreleasedReturnValue work? by Matt Galloway.

In conclusion, just use the method you prefer, Apple engineers will ensure it runs fast.