integer-decode-float

(integer-decode-float float)

Returns three values: the significand as an integer, the binary exponent, and the sign (1 or -1, an integer -- a float sign is decode-float's business only), such that significand * 2^exponent * sign is the original number. The significand is scaled to exactly (float-digits float) bits -- 53 for a normal double, fewer for a subnormal. Zero decodes as 0, 0 and its sign. The decomposition scales by two, which is exact in binary floating point, so every backend returns bit-identical values. decode-float is the same decomposition with a float significand.