Swift Resistance Explored

I am also still exploring Swift, so forgive me if I got this wrong, but isn't the problem here the following:

Your RenderGradient() takes a RenderBuffer struct. Structs are passed by value. So two things happen in your RenderBuffer function:

1) You get a copy of the RenderBuffer, which means it and its contents (a lot of Pixel instances - Arrays are passed by value too!) will be copied every iteration

2) The work you do on the pixels in the RenderBuffer copy has no effect: you change the values of a copy and then ignore the copy. You don't return the modified RenderBuffer it from RenderGradient() or do anything with the result within RenderGradient().

So I think the reason you see such a big difference in this test between optimization and no-optimization is that in the first case, with optimization, the compiler knows that the work you do in RenderBuffer is a waste of time and simply does not do it.

While in the debug build the compiler does not apply optimization so your useless code is still executed.

You can confirm this by looking at the generated code. Or maybe simpler, render the RenderBuffer in an actual buffer or image for a visual inspection.

I can understand the frustration. However I think it's a little premature to cast aspersions on the whole language because it isn't good at this particular test right now.

We still have ObjC. The two languages work together. (we also have C and asm). Pick the right tool for the job, and have reasonable expectations about the maturity and progress of Swift.

I wonder if creating a wrapper struct (let's call it Buffer) which combines an UnsafeMutablePointer and an UnsafeMutableBufferPointer would improve the language ergonomics while keeping the performance of the 'Swift*' version.

Buffer would have a constructor taking the number of items to store, and would upon creation create a buffer using UnsafeMutablePointer.alloc(). It would then use the UnsafeMutablePointer to build an UnsafeMutableBufferPointer. The latter type allows subscript indexing (e.g. myBuffer[10]) and iteration using for-in, which we could support by implementing a pass-through subscript() and generate() function.

Then, when using the struct, we'd pass it using inout parameters to ensure that it doesn't get copied.

This wouldn't fix the problem of Arrays being ludicrously unperformant, but it would at least make using the C-style buffer less painful.

The Swift safe version has the benefit of not corrupting memory if you run off the end of the array. In release builds that might account for the small perf difference.

It is not accurate to claim this is comparing the Objective-C version to Swift. It's comparing the C version.

One thing that i want to add is that this whole exercise is actually a straw man. You yourself start your post with

"Problem Statement : Design an algorithm that fills in a buffer of pixel data"

If that is the problem, then do that. Do not fill an array with Pixel object references. It is not the same problem.

I am not denying that Swift has performance issues. I do think your test and proof is artificial: nobody would use the 'Swift Safe' method to manipulate pixel data. This is exactly why the unsafe features exists.

I think what you need to discover is how to use structs, classes and raw buffers. None of those is a golden hammer. If you have performance problems with one of those then that does not mean that it is flawed. It is likely simply a bad choice for the problem at hand.