Numeric Namespace Reference

Floating-point approximations. More...

Functions
double	GetAbsoluteDifference (double a, double b)
	Returns the absolute value of the difference between a and b. More...
double	GetLogDifference (double a, double b)
	Returns the difference of the logarithm; input values are truncated at the minimum positive value. More...
double	GetRelativeDifference (double a, double b)
	Returns the safe relative difference, which is 2(|a-b|)/(|a|+|b|) except in special cases. More...

Detailed Description

Floating-point approximations.

Function Documentation

GetAbsoluteDifference()

double Numeric::GetAbsoluteDifference	(	double	a,
		double	b
	)

Returns the absolute value of the difference between a and b.

Definition at line 23 of file Numeric.cpp.

{
    return std::abs(a - b);
}

Referenced by SimDataPair::absoluteDifference().

GetLogDifference()

double Numeric::GetLogDifference	(	double	a,
		double	b
	)

Returns the difference of the logarithm; input values are truncated at the minimum positive value.

Definition at line 41 of file Numeric.cpp.

{
    double a_t = std::max(a, std::numeric_limits<double>::min());
    double b_t = std::max(b, std::numeric_limits<double>::min());
    return std::abs(std::log(a_t) - std::log(b_t));
}

GetRelativeDifference()

double Numeric::GetRelativeDifference	(	double	a,
		double	b
	)

Returns the safe relative difference, which is 2(|a-b|)/(|a|+|b|) except in special cases.

Definition at line 29 of file Numeric.cpp.

{
    constexpr double eps = std::numeric_limits<double>::epsilon();
    const double avg_abs = (std::abs(a) + std::abs(b)) / 2.0;
    // return 0.0 if relative error smaller than epsilon
    if (std::abs(a - b) <= eps * avg_abs)
        return 0.0;
    return std::abs(a - b) / avg_abs;
}

Referenced by DataUtils::createRelativeDifferenceData(), DataUtils::relativeDataDifference(), SimDataPair::relativeDifference(), HistoUtils::RelativeDifference(), and IHistogram::relativeDifferenceHistogram().