Talk:Sorting algorithms/Cocktail sort with shifting bounds

timings for   cocktail sort   vs.   cocktail sort with shifting bounds

I did some timings   (using the REXX program entries)   for the Rosetta Code

cocktail sort task

versus the

cocktail sort with shifting bounds task.

All timings used the same (repeatable) identical set of random integers   (by using a   seed   for the   random   BIF).

The integers were in the range of   0 ──► 100k,   with every other number being negated,   using   N   numbers.

All timing were performed on a moderately fast air─gap PC.

                            cocktail sort with
                              shifting bounds
                             times faster than
                  N           a cocktail sort (‡)
               ════════   ════════════════════════
                  1k              1.272
                  2k              1.228
                  4k              1.256
                  8k              1.251
                 10k              1.246
                 20k              1.244
               ════════   ════════════════════════
                                                                   
                          (‡)  cocktail sort  ÷  by
                               cocktail sort with shifting bounds

For any amount of   N   integers already in order,   the timings were identical,   and for the most part,   the times used by both sort versions weren't notable.

-- Gerard Schildberger (talk) 21:46, 9 May 2020 (UTC)