The <tgmath.h> header supplies type-generic mathematics macros. An invocation such as fabs(x), sin(x), or pow(x, n) dispatches to a numerical function selected from the argument type, with a result of the corresponding type. section 8.3
Using fabs in an iterative calculation
#include <tgmath.h> constexpr double epsilon = 1E-9; double const a = 34.0; double x = 0.5; while (fabs(1.0 - a*x) >= epsilon) { x *= (2.0 - a*x); } Here fabs supports the termination test without spelling out a separate floating-point variant. The surrounding Heron iteration repeatedly improves an estimate of 1/a until a*x is sufficiently close to 1. section 3.2
The facility covers a broad range of floating-point operations, including trigonometric, logarithmic, rounding, classification, scaling, and complex-number-related functions. Prefer these established interfaces over handcrafted replacements: implementations may use processor-specific instructions and controlled numerical precision. section 8.3
Table 8.6 Numerical functions that work with floating-point types. Names followed by [f|l] describe three functions for double, float, and long double arguments. The others are type-generic macros that adapt to the concrete type of their arguments.
acosh |
Hyperbolic arc cosine | |
acos, acospi |
Arc cosine (divided by \(\pi\)) | C23 |
asinh |
Hyperbolic arc sine | |
asin, asinpi |
Arc sine (divided by \(\pi\)) | C23 |
atan2, atan2pi |
Arc tangent, two arguments (divided by \(\pi\)) | C23 |
atanh |
Hyperbolic arc tangent | |
atan, atanpi |
Arc tangent (divided by \(\pi\)) | C23 |
canonicalize[f|l] |
Canonicalizes a floating-point value | |
cbrt |
\(\sqrt[3]{x}\) | |
ceil |
\(\lceil x \rceil\) | |
compoundn |
\((1+x)^n\) | C23 |
copysign |
Copies the sign from \(y\) to \(x\) | |
cosh |
Hyperbolic cosine | |
cos, cospi |
Cosine function, \(\cos x\) (\(\cos\pi x\)) | C23 |
erfc |
Complementary error function, \(1 - \frac{2}{\sqrt{\pi}} \int_{0}^{x} e^{-t^2} dt\) | |
erf |
Error function, \(\frac{2}{\sqrt{\pi}} \int_{0}^{x} e^{-t^2} dt\) | |
exp2 |
\(2^x\) | |
expm1 |
\(e^x-1\) | |
exp |
\(e^x\) | |
fabs |
\(|x|\) for floating point | |
fadd, dadd |
Rounded addition to float or double |
C23 |
fdim |
Positive difference | |
fdiv, ddiv |
Rounded division to float or double |
C23 |
floor |
\(\lfloor x \rfloor\) | |
fmaximum_mag[_num] |
Floating-point value of maximal magnitude | C23 |
fmax, fmaximum[_num] |
Floating-point maximum | C23 |
fma |
\(x \cdot y + z\) | |
fminimum_mag[_num] |
Floating-point value of minimal magnitude | C23 |
fmin, fminimum[_num] |
Floating-point minimum | C23 |
fmod |
Remainder of floating-point division | |
fmul, dmul |
Rounded multiplication to float or double |
C23 |
fpclassify |
Classifies a floating-point value | |
frexp |
Significand and exponent | |
fromfp[f|l] |
Round to signed integer value with specific width | C23 |
fsub, dsub |
Rounded subtraction to float or double |
C23 |
hypot |
\(\sqrt[2]{x^2+y^2}\) |
Browse the numerical-function table to distinguish type-generic macros from families with explicit double, float, and long-double forms, and to survey available operations. section 8.3
