根据结果​​确定排序算法

Question

I am now revising the sorting algorithms. Here is the question:

It is given that this sorting algorithm is written in C, and treats 'a' and 'A' as equal, and after running this sorting algorithm, the results are as follows:

10000 random data -> 0.016 sec
100000 random data -> 0.304 sec

10000 ordered data -> 0.006 sec
100000 ordered data -> 0.108 sec

10000 reversed data -> 0.010 sec
100000 reversed data -> 0.138 sec

Question: In point form briefly state the conclusions that you can draw from the test results above.

What I have done
I know this sorting algorithm is non-stable (as stated in the question), and I can guess it is a quick sort. I know that quick sort has worst case O(n^2) , average and best case O(n log n) , but I have got no idea how to explain from the results, I can't just say oh its because its non-stable and quick sort have bad results in reversed order, so i can determine it's quick sort.

Are there anything specific I can tell from the result? It would be nice if there are maths calculations or some other important observations from the results.

Answer 1

We can tell that this isn't a quadratic-time algorithm like selection sort or insertion sort, since raising the input size by a factor of 10 raised the runtime by a factor of 13-19. This is behavior we'd expect from an O(n*log(n)) average-case algorithm, like mergesort or a good quicksort.

We can tell that the algorithm isn't adaptive, or at least, not very adaptive. An adaptive algorithm would have performed much better on the sorted input, and probably on the reversed input, too. In particular, raising the size of the sorted input by a factor of 10 would have raised the runtime by a factor of about 10. While the algorithm did do better on sorted or reverse-sorted input than random input, this looks more like a result of, say, more effective branch prediction in those cases.

We don't have any information that would indicate whether the sort is stable.

I don't see anything that would distinguish whether this is a quicksort, mergesort, heapsort, or other O(n*log(n)) algorithm. We can exclude certain types of pivot selection for quicksort - for example, a quicksort that always picks the first element as the pivot would run in quadratic time on sorted input - but beyond that, I can't tell.