Grounding gets too slow with large constant numbers.

[see-mike-out]

Hello.

To summarize my situation, I'm having the following problem set.

cost(C) = C = #sum { W, X: cost_item(X, W) }.
#minimize { C : cost(C) }.

cost_item(W,X) :- cause_cost(X), cost_value(X,W).

% repeating the follow:
cost_value(a, 1000).
cost_value(b, 1001).
cost_value(c, 1002).
cost_value(d, 1003).
cost_value(e, 1004).
cost_value(f, 1005).
cost_value(g, 1006).
cost_value(h, 1007).
cost_value(i, 1008).
% ...
cost_value(Zxa, 2120).

Then, I run different problems with the following format:

cause_cost(a).
cause_cost(g).
% ...

It took a really long time, so I changed values in cost_value to be in the range of [0, 15], which made it faster.

Is it a known behavior of Clingo?

Thank you, and please redirect me to the original issue if its a duplicate question.

[rkaminsk]

Assignment aggregates should ideally only be used over domain predicates and otherwise avoided.

I suppose that the cause_cost predicate in your example is meant to be derived by some other rules.

In the program above, it is easy to remove the assignment aggregate:

#minimize { W,X: cost_item(X,W) }.

[see-mike-out]

I appreciate your response, yet I don't think the main issue was the assignment aggregate. When I rescaled the constants to lower values, it got faster. I mean, because I solved that way, it's not a big deal anymore, but I am wondering if that is the expected behavior.