Comments on: Computer Algorithms: Jump Search

By: Nikolay Mihaylov

Nikolay Mihaylov — Fri, 22 Jul 2016 18:38:25 +0000

I am currently programming binary search in disk file. I found out that jump search will be better suited, because the OS will cache the jump “points” and jumps will be at no additional “cost”.

Sure this is true for dividing “points” of the binary search too, but they have how to say “priority” (central element is the most used) and OS cache do not perform that well.

Rudimentary benchmark show if bin search doing it for 40 – 50 sec every time, jump search (after first 3-4 searches) is doing it for 3-4 sec.

By: Stoimen

Stoimen — Tue, 13 Dec 2011 10:10:28 +0000

@Jordan Harband – thanks for the remark! fixed!

By: Jordan Harband

Jordan Harband — Mon, 12 Dec 2011 18:35:59 +0000

In your PHP jump search implementation, the line “$step += $step;” means that the step will go from √n -> 2√n -> 4√n -> 8√n, when you want it to go from √n -> 2√n -> 3√n -> 4√n.

You either need a coefficient variable that gets increased by 1 (ie, $step_iteration = $coefficient * $step) or you need a separate $step_iteration variable (ie, $step should never be modified, and each iteration, $step_iteration += $step).