Function float_min 

pub fn float_min(a: f64, b: f64) -> f64

The language’s min of two reals: a NaN gives way to the other operand, and of two equal values the one with the negative sign, so min(0, -0.0) is -0.0 (IEEE 754-2019 minimumNumber). One answer on every tier and under every optimizer, where f64::min leaves the sign of two equal zeros open.