Function fast_math::exp_raw

pub fn exp_raw(x: f32) -> f32

Compute a fast approximation to e^x for -104 ≤ x ≤ 104.

This will return unspecified nonsense if x does not satisfy those requirements. Use exp if correct handling is required (at the expense of some speed).

The maximum relative error for inputs for which the result is normal (x ≥ -128 ln(2) ≈ -88.7) is less than 0.011. For x < -128 ln(2), the relative error in the (subnormal) result can be as large as 1.