usr/src/sys/tahoemath/FP.h

/*
 *      @(#)FP.h        5.1 (Berkeley) 11/3/86
 *
 * General definitions of the floating point stuff on Power 6/32.
 * The floating point format definition is:
 *
 *              S    (exp-128)
 *          (-1)  * 2          * F
 *
 *      Where exp is the exponent field and F is the binary
 *      mantissa following it, including the hidden bit.
 *      The hidden bit actually is 1/2, so F is known to
 *      satisfy the range:
 *              1/2 <= F < 1
 */

typedef struct {
        unsigned        sign:1;
        unsigned        exponent:8;
        unsigned        mantissa:23;
} sp_format;

typedef struct {
        unsigned        sign:1;
        unsigned        exponent:8;
        unsigned        mantissa:23;
        unsigned        mantissa_lst;
} dp_format;

#define EXP_BIAS        128             /* Exponent bias */
#define SIGN_BIT        0x80000000      /* S bit mask */