relative tolerance is not always enough, especially near 0. This classic article gives a nice overview of absolute, relative, ulp-based and other methods for co

relative tolerance is not always enough, especially near 0. This classic article gives a nice overview of absolute, relative, ulp-based and other methods for comparison of float numbers: https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/ https://randomascii.wordpress.com/2012/02/25/comparing-float...