FLINT3.7.0-dev
/

_dfloat_rad_mul_slow, _dfloat_rad_mul, _dfloat_rad_div, _dfloat_rad_finish, _dfloat_lower

Synopsis#include <flint/dfloat.h>
double _dfloat_rad_mul_slow(double a, double b, double p)source
double _dfloat_rad_mul(double a, double b)source
double _dfloat_rad_div(double a, double b)source
double _dfloat_rad_finish(double r)source
double _dfloat_lower(double r)source

Radius arithmetic on nonnegative doubles: an upper bound (up to the final slack factor) for abab and for a/ba/b (b>0b > 0), with products and quotients that may have underflowed bumped up by 2−10702^{-1070} and with 0⋅∞=00 \cdot \infty = 0 (_dfloat_rad_mul_slow is the underflow path of _dfloat_rad_mul, given p=fl⁡(ab)p = \operatorname{fl}(ab)); the final upper bound r(1+2−45)r(1 + 2^{-45}); and a lower bound r(1−2−45)r(1 - 2^{-45}) for a positive quantity computed with a few roundings (0 for r<2−1000r < 2^{-1000}).

Cite this page

The FLINT team, “_dfloat_rad_mul_slow, _dfloat_rad_mul, _dfloat_rad_div, _dfloat_rad_finish, _dfloat_lower (dfloat.h),” FLINT: Fast Library for Number Theory, documentation, version 3.7.0-dev, 2026. https://flintlib.org/doc/dfloat/_dfloat_rad_mul_slow.html (generated 2026-11-02).