them we are executing the original algorithm without modifications, while the sorting
of the two remaining elements is trivial at the end. We have already observed that the
use of two heaps reduces the depth of the heap trees and thus it reduces the cost of
heapify. This leads us to a refinement of our algorithm trying to make this happen for
more iterations. The ideal would be for n – 2 iterations and in order to achieve this, ot
would require the use of at least two heaps constantly without merging them into one
in any step. This is the main idea of our approach.