Parallel Processing Test With n-Queens


We had an assignment where we came up with an algorithm to solve the classic n-Queens puzzle.

(The problem in brief: Given an n x n chessboard, place n queens on it so that none of them can attack each other. For any given n, how many configurations are there where this is true?)

This is obviously a computer problem because by n = 13, for example, there are 73,712 combos, which nobody should be figuring out by hand. By the time n hits the mid-teens, it starts taking a few minutes to calculate on regular computers.

The algorithm we came up with (I worked with classmate Benjamin Zarzycki) could probably use a fair amount of optimization (we didn't take symmetry into account for example, so that would halve the time) but we're using it as is to try out parallel processing through HTML5 web workers.

Web workers are parallel processes you can spawn off through your browser, either to run intensive tasks in the background without interfering with your main browsing experience, or to get a task done faster by splitting it into different parts which can be processed at the same time by different cores on your computer.

Here's our webworker page, feel free to try out different n's and different numbers of web workers on your own. I would start at 12 or 13 and slowly work up. Also most people will see improvement going from 1 to 2 workers, but I think you might need more than 4 cores to do better (I think one is already saved for the main process). I have a Macbook Pro with a Core i5 processor and it doesn't make much difference going from 2 workers to 3 workers and up.

But I'm not totally sure how distribution among cores work - depends on the browser I guess? And how it wants to divvy things up. So let me know if you find out more from your results.

And almost forgot, source code here!