doubledouble: A C++ repository from WarrenWeckesser

DoubleDouble is a C++ class that implements a double-double floating point type.

The DoubleDouble class defines:

arithmetic operators: +, -, *, /
inplace operators: +=, -=, *=, /=
comparison operators: ==, != , <, <=, >, >=
the functions: abs, sqrt, powi, exp, expm1, log, log1p, hypot
several constants: dd_e (base of natural log), dd_pi (π), dd_sqrt2 (sqrt(2)), and more.

C++17 is required to use the DoubleDouble class.

Example

Here is a small program to compute the volume of a cone with base radius 2.0 and height 7.5.

#include <cstdio>
#include <charconv>
#include <system_error>
#include "doubledouble.h"

using namespace std;

// For DoubleDouble and dd_pi.
using namespace doubledouble;

void print_double(const double x)
{
    char buf[24];
    const to_chars_result res = to_chars(buf, buf + sizeof(buf), x);
    if (res.ec == errc{}) {
        printf("%.*s", static_cast<int>(res.ptr - buf), buf);
    }
    else {
        printf("<to_chars() failed!>");
    }
}

int main()
{
    // Compute the volume of a cone with DoubleDouble inputs.

    // Radius of the base of the cone.
    DoubleDouble r{2.0};

    // Height of the cone.
    DoubleDouble h{7.5};

    // Volume is pi*r^2*h/3.
    auto volume = dd_pi * r.powi(2) * h / 3;

    printf("Volume of cone is ");
    print_double(volume.upper);
    printf(" + ");
    print_double(volume.lower);
    printf("\n");
}

When run, the program prints

Volume of cone is 31.41592653589793 + 1.2246467991473535e-15

For comparison, here is how that result could be computed in Python with the mpmath library:

from mpmath import mp

mp.dps = 50
r = mp.mpf(2.0)
h = mp.mpf(7.5)
vol = mp.pi*r**2*h/3
print(float(vol), float(vol - float(vol)))

The Python script prints

31.41592653589793 1.2246467991473533e-15

WarrenWeckesser/doubledouble

Example