Joachim Schimpf wrote: > Lars Kotthoff wrote: >> Dear list, >> >> is there a general, preferably easy, way of getting the number of search tree >> nodes explored when searching for the solution to a problem? > > A "node" is a place where a nondeterministic choice is made, e.g. in indomain/1 > when you are using finite domains. You need to add a nonlogical counter to > your code, e.g. Another definition of a node is where a unification is performed, regardless of if it was deterministic or not. This can give you some idea of the amount of computation your program may be performing. The basic assumption is that the computation performed in a unification is on average similar for all unifications in your program. This does not work well if some of your reasoning/computation is not performed in Prolog (e.g. if it is done in C, as in eplex or part of the ic solver). ECLiPSe does not provide a direct measurement of this definition of `node'. However, it does provide tools that may be useful to you, e.g. the number of times that various points in a program are reached. See chapter 6 of the tutorial (http://eclipse-clp.org/doc/tutorial/tutorial042.html) for more information on these tools. Cheers, KishReceived on Tue Jan 13 2009 - 15:08:20 CET
This archive was generated by hypermail 2.2.0 : Thu Feb 02 2012 - 02:31:58 CET