mirror of
https://github.com/boostorg/math.git
synced 2026-02-24 16:12:15 +00:00
Optimises which method is chosen. Also correct rescaling in bessel_i_backwards_iterator so we don't get too close to zero and rely on type T having denorms.