Um Limite Superior para a Complexidade do ShellSort

The worst-case time complexity of the ShellSort algorithm is known only for some specific sequences (a sequence is a parameter of the algorithm). Relating the algorithm to the Frobenius number concept, we present an algorithm for determining the maximum number of comparisons for any sequence and array to be ordered. We apply this method together with the empirical determination of complexity to analyze several sequences whose worst case complexity are known. We show that the empirical approach succeeded in determining the same complexities which are analytically known and presented its results for sequences with unknown worst-case time complexity.