Re: How does SICStus form its internal constraint queue?
- From: "Mats" <matsc@xxxxxxx>
- Date: 10 Nov 2006 05:20:38 -0800
Tomas wrote:
I am conducting some random CSP experiment with binary (allowed
tupples) and the SICStus built in non-binary constraint (up to arity
5). I work in the area of over-constrained CSP (never have solutions).
Constraint layout: all binary constraint posted first followed with a
mix of the non-binary constraint.
When I start removing some non-binary constraint the search result
(resumption, backtrack) stays the same. It seemed that the non-binary
constraint was redundant. I changed the order layout of the constraint
(posted binary constraint last) but the search effort was nevertheless
the same.
My question is; does SICStus forms its own queue, places all the binary
constraints at the head, regardless of the order in which we post them?
SICStus maintains a suspension queue with a simple priority scheme, as
follows. Binary vs. non-binary does not affect the priority. Priority
levels:
1. Indexicals.
2. Globals suspended on value event.
3. Globals suspended on bound change event.
4. Globals suspended on domain value event.
Each level is a mix of queue and stack.
Feel free to email me.
--Mats
.
- References:
- Prev by Date: Re: problema coi frames
- Next by Date: Student-safe executables with SWI-Prolog?
- Previous by thread: How does SICStus form its internal constraint queue?
- Next by thread: problema coi frames
- Index(es):