Re: puzzle 15
- To: mathgroup at smc.vnet.net
- Subject: [mg24840] Re: puzzle 15
- From: "Ira D. Baxter" <idbaxter at semdesigns.com>
- Date: Tue, 15 Aug 2000 03:04:09 -0400 (EDT)
- References: <8n7qol$1l6@smc.vnet.net>
- Sender: owner-wri-mathgroup at wolfram.com
Being interpretive, Mathematica isn't particularly fast. That explains about a factor of 100x. For exhaustive search, you want a good compiler. For this kind of search, you'd ideally go parellel. That would buy a N-way speedup for N CPUs. We've done this with our own parallel programming language, http://www.semdesigns.com/Products/Parlanse/index.html. We have a 15-puzzle solver coded in Parlanse that finds typically 50-move optimial solutions on an 8-way SMP 200Mhz Pentium Pro in roughly 23 seconds. This language is used to implement a parallel rewriting system, too. -- Ira Baxter, Ph.D., CTO idbaxter at semdesigns.com 512-250-1018x140 Semantic Designs, Inc., www.semdesigns.com FAX 512-250-1191 12636 Research Blvd #C214, Austin, Texas 78759 Borut L. <JustMyName at email.si> wrote in message news:8n7qol$1l6 at smc.vnet.net... > Helo, > > I've written algorithm in Mathematica for a summer homework on Loyd's puzzle > 15. I've used back-tracing, i.e. exhaustive search method for it, because > nothing else came on my mind. It turned out, it's extremlly slow (especially > for boards larger than 3x3). I post here to ask anybody for a good tip on > the subject matter. > > Thanks, Borut L. > > >