stillwater-sc / universal / 29737947253

85%

master: 84%

Build Type

push

github

Committed by web-flow

Commit Message

feat(efloat): implement math/numerics.hpp -- frexp, ldexp (#1169)

efloat was missing frexp/ldexp (ereal already has them). Both only move efloat's
binary exponent, so they are exact at any precision -- no double round-trip:

- ldexp(x, n) = x * 2^n via setexponent(scale() + n).
- frexp(x, &e) returns m in [0.5,1) with x == m * 2^e: e = scale()+1, and m is x
  with its exponent set to -1 (leading bit at 2^-1).

Zero/inf/nan handled per <cmath> (frexp sets *exp = 0 for those). Wired into
mathlib.hpp.

Regression test elastic/efloat/math/numerics.cpp: ldexp/frexp vs std for
double-range values, mantissa-in-[0.5,1), ldexp(frexp(x))==x round-trips
(including a beyond-double 1 + 2^-300 that stays exact), and special values.
PASS on gcc and clang; cppcheck-clean.

Resolves #1164
Relates to #582

Coverage Stats

18 of 18 new or added lines in 1 file covered. (100.0%)

6 existing lines in 1 file now uncovered.

49780 of 58715 relevant lines covered (84.78%)

7035108.17 hits per line