Module float_ops 

Float atoms outside the big verbatim-upstream Float/FloatCore bundles: bit-pattern access, ULP stepping, IEEE 754-2019 min/max, round-half-to-even, the algebraic_* fast-math family and the libm-backed special functions.

This is the first slice of the planned Float decomposition: everything here that touches only bits and comparisons is a c0nst trait with const impls on nightly — unlike the transcendental bundle, which waits on std/libm.

CT tiers: FloatBits, NextUp/NextDown and Algebraic are Tier A; Minimum/Maximum are Tier B-ish (comparison ladders, but on floats CT rarely applies); RoundTiesEven, Erf and Gamma are Tier C (data-dependent branching / libm).

Traits

Algebraic

Arithmetic with an “algebraic” license: the compiler may reassociate, use reciprocal shortcuts, or contract into fused operations.

Erf

The error function and its complement.

FloatBits

Raw IEEE 754 bit-pattern access.

Gamma

The gamma function and the natural log of its absolute value.

Maximum

IEEE 754-2019 maximum (NaN-propagating, unlike max).

Minimum

IEEE 754-2019 minimum (NaN-propagating, unlike min).

NextDown

Steps one ULP towards negative infinity.

NextUp

Steps one ULP towards positive infinity.

RoundTiesEven

Rounds half-way cases to the nearest even integer (banker’s rounding).