The function returns 0 (zero) when its argument is 0 (zero). When its argument is non-zero, it returns a value of type IlcFloat which is less than its argument.

In order to avoid errors due to rounding, Solver uses the technique of outward rounding when working on intervals. During constraint propagation, when new bounds are computed for the domain of a constrained floating-point variable, the newly computed interval is slightly expanded: its lower bound is decreased a little bit, whereas its upper bound is increased a little bit. This practice avoids making intervals smaller than what they would be with exact computation.

The implementation of this technique conforms to the IEEE 754 standard. In this way, Solver provides for consistent and more nearly accurate results in basic arithmetic operations and thus avoids some of the drawbacks of floating-point arithmetic.

Solver uses the function IlcComputeMin when it is enlarging an interval of floating-point values by slightly decreasing the lower bound of the interval.