pub fn erfc(x: f64) -> f64Expand description
Complementary error function, erfc(x) = 1 - erf(x).
Chebyshev-coefficient form (Numerical Recipes, 3rd ed., section 6.2.2, the Erf
struct’s erfccheb), valid for any nonnegative z via the substitution `t = 2 / (2
- z)
, which maps[0, infinity)onto(0, 1]-- so the same 28-term expansion covers the whole range without a separate small-x/large-xsplit. Negative arguments use the exact identityerfc(-x) = 2 - erfc(x)`.
Accurate to a couple of ULP of the true double-precision result across the whole
range (see this module’s tests: the Abramowitz & Stegun Table 7.1 values, a denser
libm cross-check out to x = 6, and the far tail at x = 6/sqrt(2) used by
normal_cdf(-6.0)). erfc(NAN) is NAN; erfc(INFINITY) is 0.0; erfc(NEG_INFINITY)
is 2.0.