Comments on: Sorting in Linear time http://elliottback.com/wp/sorting-in-linear-time/ Internet & Technology Sun, 22 Nov 2009 02:40:56 -0400 http://wordpress.org/?v=2.8.4 hourly 1 By: Elliott Back http://elliottback.com/wp/sorting-in-linear-time/#comment-2091620 Elliott Back Wed, 24 Sep 2008 23:04:25 +0000 http://elliottback.com/wp/?p=1198#comment-2091620 I totally agree with you! I totally agree with you!

]]> By: Andrasek http://elliottback.com/wp/sorting-in-linear-time/#comment-2090927 Andrasek Wed, 24 Sep 2008 17:38:43 +0000 http://elliottback.com/wp/?p=1198#comment-2090927 I feel like I should mention that you'll never be able to write an algorithm *in Java* that is faster than the in-built .sort() algorithm. The reason for this is due to the overhead from using arrays in Java. For example, it's not possible for an array to have a subscript that is beyond the bounds of the array. For example, you can't have a 100 element array and then reference cell 105. Java throws an error. This simple range checking functionality on every single instance of referencing an array causes a noticeable amount of overhead. Imagine putting an "if" check before every array subscript usage. I count 24 array subscripts in the PartialFlashSort algorithm alone. That's 24 extra checks your code performs that the in-built .sort() routine can ignore. Also, the inbuilt .sort() algorithm uses a "tuned quicksort" which performs significantly better than the standard quicksort on several troublesome data sets. Trying to beat a quicksort is a challenge in itself. If you add to that the additional overhead that the Java language adds, there's no chance that any Java-based sorting algorithm can compete with the internal Java .sort() routine. I feel like I should mention that you’ll never be able to write an algorithm *in Java* that is faster than the in-built .sort() algorithm. The reason for this is due to the overhead from using arrays in Java.

For example, it’s not possible for an array to have a subscript that is beyond the bounds of the array. For example, you can’t have a 100 element array and then reference cell 105. Java throws an error.

This simple range checking functionality on every single instance of referencing an array causes a noticeable amount of overhead. Imagine putting an “if” check before every array subscript usage. I count 24 array subscripts in the PartialFlashSort algorithm alone. That’s 24 extra checks your code performs that the in-built .sort() routine can ignore.

Also, the inbuilt .sort() algorithm uses a “tuned quicksort” which performs significantly better than the standard quicksort on several troublesome data sets. Trying to beat a quicksort is a challenge in itself. If you add to that the additional overhead that the Java language adds, there’s no chance that any Java-based sorting algorithm can compete with the internal Java .sort() routine.

]]>