2. Numerical Helpers

These help converting to and from floating point and real numbers in scheme.

2.1. Real to Float

Converts a real value (represented as an s9_cell) to an internal GFFLT value.

The full function name is monolith_scheme_float_to_real, but can also be utilized via the macro definition REAL2FLOAT.

#define REAL2FLOAT monolith_scheme_real_to_float

GFFLT monolith_scheme_real_to_float(s9_cell real);

Bignum architecture is still something of a dark art for me, but from what I gathered reading the s9 source code and some trial and error is that the architecture uses base 10 to store values. Converting the value is a matter of extracting the mantissa and exponent components from the cell, and then plugging those values into the pow equation seen below.

GFFLT monolith_scheme_real_to_float(s9_cell real)
{
    GFFLT val;
    int mant;
    int expt;

    if (s9_integer_p(real)) {
        return integer_value(NULL, real);
    }

    mant = s9_bignum_to_int(s9_real_mantissa(real));
    expt = s9_real_exponent(real);

    val = mant * pow(10, expt);

    return val;
}

2.2. Float to Real

Converts a floating point value to a real value. The decimal precision needs to be explicitly specified.

#define FLOAT2REAL monolith_scheme_float_to_real

s9_cell monolith_scheme_float_to_real(GFFLT val, int p);

s9_cell monolith_scheme_float_to_real(GFFLT val, int p)
{
    GFFLT c;
    unsigned int ival;
    int sign;

    c = pow(10, p);

    sign = val < 0 ? -1 : 1;

    val = fabs(val);

    ival = floor(val);
    val = floor(c * (ival + (val - ival)));
    return s9_make_real(sign, -p, s9_make_integer(val));
}

---

prev | home | next