An Improvement on Tree Selection Sort
ALENEX '02 Revised Papers from the 4th International Workshop on Algorithm Engineering and Experiments
Building a new sort function for a C library
Software—Practice & Experience
Efficient sample sort and the average case analysis or PEsort
Theoretical Computer Science
| Hi-index | 0.00 |
PROPORTION EXTEND SORT is a new sorting algorithm, the basic principle of which is similar to PROPORTION SPLIT SORT. This algorithm sorts a sequence by constructing three subproblems, using a QuickSort-like pivot technique and solving recursively each subproblem. The original problem and three subproblems all are of such a structure: a sorted subsequence followed by an unsorted subsequence. The size of the original problem always equals the size of the third subproblem, but in general, the sorted subsequence of the third subproblem is p+1 times as much as the sorted subsequence of the original, where p is a fixed positive constant. The worst case number of comparisons required by this algorithm is less than 1/log (1+1/(2p2+2p-1))nlog n for $p \geq 1$. Empirical results show that the average number of comparisons is close to n log n-O(n) for some p. From our experiments for sorting integers, when p = 16, this algorithm is yet faster, on average, than PROPORTION SPLIT SORT which is faster than CLEVER QUICKSORT.