I appreciate that this writeup takes care to call out use cases when they help with understanding!

I do have a semi-unrelated question though: does using the recursive approach prevent it from being calculated efficiently on the GPU/compute shaders? Not that it matters; plenty of value in a CPU-bound version of a solution and especially one that is easy to understand when recursive. I was just wondering why the prominent examples used a non-recursive approach, but then I was like "oh, because they expect you to use them on the GPU". ...and then I was like "wait, is that why?"