decode-float

(decode-float float)

Returns three values: the significand as a float in [1/2, 1), the binary exponent, and the sign (1.0 or -1.0), such that significand * 2^exponent * sign is the original number. Zero decodes as 0.0, 0 and its sign. The decomposition scales by two, which is exact in binary floating point, so every backend returns bit-identical values. scale-float is the reverse step.