mach.math.floats

This package contains functions for reading and manipulating floating point values efficiently and in a format-agnostic way.

mach.math.floats.extract

This module provides functionality for extracting the individual components of floating point values.

float zero = 0;
assert(zero.fextractsgn == 0); // Sign
assert(zero.fextractexp == 0); // Unbiased exponent
assert(zero.fextractsexp == -126); // Biased exponent
assert(zero.fextractsig == 0); // Raw significand
assert(zero.fextractnsig == 0); // Normalized significand

mach.math.floats.inject

Complement to the extract module, this one provides implementations for setting the components of floating point values.

auto x = float(0).finjectsgn(1); // Sign
assert(x.fextractsgn == 1);
auto y = float(0).finjectexp(100); // Exponent
assert(y.fextractexp == 100);
auto z = float(0).finjectsig(1000); // Significand
assert(z.fextractsig == 1000);

mach.math.floats.neighbors

Each float has a successor and a predecessor - defined as the least greater and greatest lesser values, respectively. This module provides functions for retrieving those neighbors given a floating point value.

assert(float.max.fsuccessor == float.infinity);
assert(float.infinity.fpredecessor == float.max);
assert(float(0).fsuccessor.fpredecessor == 0);

mach.math.floats.properties

Defines properties for checking various states a floating point value can occupy, including being NaN, infinity, normal, subnormal (or denormal), and unnormal.

assert(float.nan.fisnan);
assert(float.infinity.fisinf);
assert(float(0).fiszero);
assert(float(0).fisnormal);