Brancher for tracing
Philippe Morignot
I am writing to you for the following question: I have defined several branchers that execute some source code on a model, following section 8.14 of MPG “Executing code between branchers”. Two of these branchers (lines 2 and 6 below) are for tracing the behavior of the search engine only: printing text and variables on std::cout. And what surprises me is that commenting out these two tracing branchers leads to a slightly different search tree than not commenting them out …
The statistics at the end of a run show that there are more propagations (1682 vs. 1319) and less nodes (26 vs. 37) without tracing branchers than with them.
Would someone have an explanation for this? Here is an excerpt of the source code: (with tracing branchers of lines 2 and 6 commented out)
1 for (int i = 0 ; i < is_performed.size() ; i++) {
2 //branch(*this, lambda_print_begin); // print text and variables on std::cout
3 branch(*this, is_performed[i], Gecode::BOOL_VAL_MAX());
4 branch(*this, lambda_assign_vars1);
5 assign(*this, object_id[i], Gecode::INT_ASSIGN(lambda_select_object));
6 //branch(*this, lambda_print_end); // print text and variables on std::cout
7 }
Philippe
Mikael Zayenz Lagerkvist
The tracing propagators create unary branches in the search tree, so
it is reasonable that there will be more nodes. If you use Gist, they
should be visible in the search tree.
As for propagations, I am not sure what might be the reason. One
reason might be because adding unary branchers does change the depth,
and thus changes where copies are left, which probably has an impact.
Cheers,
Mikael
> You received this message because you are subscribed to the Google Groups "Gecode" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to gecode+un...@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/gecode/714035349.1472937.1784655335958%40mail.yahoo.com.
--
Mikael Zayenz Lagerkvist