CODEWITHSUNDEEP
iosarraysswiftcopy-on-writeios-multithreading
PGDev
PGDev

Reputation: 24341

Copy-on-Write in Multi-threaded environment in Swift

I've read about Copy-on-Write concept for optimization in Arrays and other data structures in Swift.

What I want to know is how Copy-on-Write works in a multi-threaded environment.

    let arr1 = [1, 2, 3, 4]
    let arr2 = arr1
    arr1.withUnsafeBytes { print("arr1:", $0.baseAddress) } //0x000060000007ee60
    arr2.withUnsafeBytes { print("arr2:", $0.baseAddress) } //0x000060000007ee60
    
    DispatchQueue.global(qos: .default).async {
        let arr3 = arr1
        arr3.withUnsafeBytes { print("arr3:", $0.baseAddress) } //0x000060000007ee60
    }

In the above code, arr1 and arr2 have same addresses initially as expected in copy-on-write. But, arr3 also shares the same memory as arr1 and arr2 although it is executed on a different thread.

As far as I know, each thread has different stack allocation. Then why arr3 is still sharing the same location?

Upvotes: 3

Views: 458

Answers (1)

Rob Napier
Rob Napier

Reputation: 299663

You're not looking at the addresses of the arrays. You're looking at the addresses of the internal backing storage of the arrays, which is shared and heap-allocated.

If you want to look at the addresses of the stack-allocated array container (the part that points to the backing storage), then you meant this:

var arr1 = [1, 2, 3, 4]
var arr2 = arr1
withUnsafePointer(to: &arr1) { print("arr1:", $0) }
withUnsafePointer(to: &arr2) { print("arr2:", $0) }

DispatchQueue.global(qos: .default).async {
    let arr3 = arr1
    withUnsafePointer(to: arr3) { print("arr3:", $0) }
}

// =>
arr1: 0x0000000122d671e0   // local stack
arr2: 0x0000000122d671e8   // local stack (next address)
arr3: 0x0000700000e48d10   // heap

I believe this is the kind of result you were expecting.

Upvotes: 5

Related Questions