Page 1 of 1
Developing WLS
Posted: October 7th, 2011, 8:45 pm
by David
I think WLS should be upgraded. Please post here ideas or upgraded WLS.
My ideas are here.
1. More speed. Do you have any better algorithm?
2. Be searchable for any cellular automaton. Load ruletable to use.
Do you have other ideas?
Re: Developing WLS
Posted: October 24th, 2011, 6:59 pm
by p46beth
David wrote:1. More speed. Do you have any better algorithm?
If we are writing a new searching algorithm, then we are basically writing an entire new search program
I'm more interested in a hexagonal version then in another square version.
Re: Developing WLS
Posted: October 30th, 2011, 2:31 am
by David
137ben wrote:David wrote:1. More speed. Do you have any better algorithm?
If we are writing a new searching algorithm, then we are basically writing an entire new search program
I think so, too... Does WLS uses the best algorithm, currently?
Re: Developing WLS
Posted: October 30th, 2011, 10:23 pm
by Jason Summers
David wrote:I think so, too... Does WLS uses the best algorithm, currently?
I'm not aware of any breakthroughs in this type of searching. Of course, there are many ways to optimize it. But fundamentally, it's all just brute force: try all possibilities, and eliminate the ones that lead to a contradiction.
If someone wants to work on WLS or a similar program, my wishlist of features for speeding it up would start with:
1. Hashing (like gfind/ofind), to prevent it from redoing the same search over and over after making an irrelevant change.
2. Multithreading. Run one search thread per processor core. I think this is possible. For example, if a thread has a choice of setting a cell OFF or ON, and another thread is idle, it assigns the OFF state (and the search tree associated with it) to that other thread.
Re: Developing WLS
Posted: November 6th, 2011, 6:50 pm
by David
A new idea:
Limit with population, instead of bounding box.