# Number Comparison Tolerance

 Name: NUMBER_COMPARISON_TOLERANCE  Type: double Default Value: 1.e-6 Range: $$[0 – 1]$$

This tolerance is used to determine whether two floating point numbers are the same. Its use in the solver is context sensitive, as outward rounding is always used to ensure that solutions are not incorrectly eliminated.

Making this tolerance smaller can accelerate convergence, as the solver becomes more aggressive in domain reduction. This, however, increases the risk of eliminating otherwise acceptable solutions. Relaxing this tolerance slows down convergence, as the solver adopts a more conservative approach.