Routine Name: machineEpsilon

Author: Christopher Johnson

Language: C++. Tested with g++ compiler.

Declared in: “machineEpsilon.h”

Description/Purpose: Computes accuracy of a given C++ type on current machine.

Input: No parameters, but must give a typename T as a template argument.

Output: Returns an int, indicating the number of digits of accuracy of the given type.

Usage/Example:

machineEpsilon() will return the digits of accuracy for the float type. Other types may be used likewise.

#include <iostream>
#include "machineEpsilon.h"

int main (void)
{
	std::cout << "single precision is " << machineEpsilon<float>() << " digits.\n";
	std::cout << "double precision is " << machineEpsilon<double>() << " digits.\n";
}

Output from the lines above:

single precision is 8 digits.
double precision is 16 digits.

Implementation/Code: The following is the code for machineEpsilon()

template <class T>
int machineEpsilon (void)
{
	int digits = 1;
	T val = 1.1;
	while (val>1)
	{
		digits++;
		val = (val-1)/10 + 1;
	}
	return digits;
}

Last Modified: December/2017