Error: Registry: Constraint float_log2 not found

[Andrew Gill]

Hi - Hoping there is an obvious (although not to me :-() fix. 

int: K=5;

set of int: K_set=1..K;

array[K_set] of var int: N;

constraint ceil(log2(int2float(N[K]))+0.001)=4;

I thought int2float would coerce N[K] to a float, which log2 takes and returns a float, which I could add to the float 0.001 (yielding a float), which ceil takes and returns an int, which I could constrain to equal an int?

Andrew

[Jip Dekker]

Hi Andrew,

The problem seems to be that Gecode does not implement the "float_log2" predicate, which is used when compiling your model. I think the simplest fix is to stick to integers, and use the fact that constraints are relational to use "pow" instead.

You could replace your constraint by:

constraint pow(2, 4) = N[K];

Or you could even define an integer log2 function and use that:

function var int: int_log2(var int: x) = let {
var 0..ub(x) div 2: y;
constraint pow(2, y) = x;
} in y;
constraint int_log2(N[K]) = 4;

Note that this might quickly lead to integer overflow, so you will have to put some bound on 'N'. For example, the bounds '0..50' would work.

[Andrew Gill]

Hi Jip - Thanks for the suggestion, but I maybe should have provided a bit more context. The constraint I was after was to set "the smallest power of 2 strictly greater than N[K]" to a given integer, for which I thought  pow(2,ceil(log2(N[K])+0.001))=16
would work, i.e., N[K]=15 makes LHS=16 but N[K]=16 makes LHS=32 (which I why I add the 0.001 to the ceil() argument). I can't see how I can achieve this without using the float version of log2().

Andrew