stillwater-sc / universal / 29693359164

85%

master: 84%

Build Type

push

github

Committed by web-flow

Commit Message

feat(efloat): give hypot its own math/hypot.hpp header + dedicated regression test (#1158)

efloat's hypot(x,y) already existed and works well -- a native scale-by-max
implementation (sqrt(x^2+y^2) without intermediate over/underflow) on efloat's
own arithmetic, so it is full-precision, not a double-delegating shim. But it
lived inside sqrt.hpp and had no dedicated test. This aligns efloat with the
repository-wide one-function-per-header structure the other number systems use.

- Move hypot from math/sqrt.hpp into new math/hypot.hpp (which includes sqrt.hpp
  for its sqrt dependency). No behavioral change. sqrt.hpp title updated to
  "square root and cube root".
- Wire math/hypot.hpp into mathlib.hpp after sqrt.hpp. Safe: only mathlib.hpp
  and hyperbolic.hpp include sqrt.hpp directly, and hyperbolic needs only sqrt;
  every consumer reaches hypot through the mathlib umbrella.
- Add elastic/efloat/math/hypot.cpp: exact Pythagorean values, symmetry
  (hypot(x,y)==hypot(y,x)==hypot(x,-y)), zero cases, overflow/underflow
  prevention (hypot(3e200,4e200), hypot(3e-200,4e-200)), IEEE special values
  (inf suppresses NaN), and two 512-bit oracle identities (hypot^2==x^2+y^2 and
  hypot(1,1)==sqrt(2), agreeing to < 2^-400).

Existing sqrt_precision.cpp and math.cpp (which also exercise hypot) still build
and PASS on gcc and clang.

Resolves #1121

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Coverage Stats

22 of 22 new or added lines in 1 file covered. (100.0%)

10 existing lines in 1 file now uncovered.

49680 of 58637 relevant lines covered (84.72%)

7039400.13 hits per line