The cost of merge sort implemented classically either as a top-down recursive function or a bottom-up iterative with a small local array of pointers is the same: O(N.log(N)). The number of comparisons will vary depending on the actual contents of the array, but by at most a factor of 2.

You can improve this algorithm at a linear cost by adding an initial comparison between the last element of the left slice and the first element of the right slice in the merge phase. If the comparison yields <= then you can skip the merge phase for this pair of slices.

With this modification, a fully sorted array will sort much faster, with a linear complexity, making it the best case, and a partially sorted array will behave better as well.