realistically which of these inlines cleaner? `val foo by delegate::foo` or `inline val foo get(...

# compiler

[jimn]

realistically which of these inlines cleaner?

val foo by delegate::foo or inline val foo get()=delegate.foo

I know which one shows up in the icicle graph like a bag of cinder blocks

[Youssef Shoaib [MOD]]

most likely the second one since it becomes a direct call to delegate.foo

[jimn]

this is counter-intuitive to the postulate :: > ()

[Youssef Shoaib [MOD]]

I'm not sure where you got the information that :: is more performant than () . In fact, usually it isn't unless you're calling an inline function with :: as a lambda argument. I think most Kotlin devs immediately recognise :: as a sign that there isn't huge performance optimisations done to said code because (in some cases) :: can result in a function reference which uses costly reflection.

[jimn]

when we're discussing a delegate on a vtable, or when 30 years ago we would be, :: > (*a.b) and () fast forward to present day where an interface delegation is a shim and an inline is a faster shim you would still wonder about :: costing more than any declarative keyword magic.

[ephemient]

https://youtrack.jetbrains.com/issue/KT-16997

[ephemient]

until that's taken care of, by :: will necessarily involve more machinery

[udalov]

In the upcoming Kotlin 1.6, we've optimized properties which delegate directly to other properties, so val foo by delegate::foo will almost literally mean val foo get() = delegate.foo.
https://youtrack.jetbrains.com/issue/KT-39054

[jimn]

is there a gap list of kotlin IR vs. llvm ir features? last i knew this was the first and foremost assumption of llvm optimization unless you fancy -O0