tgmath.h

Remembering to write sqrtf for float, sqrt for double, and sqrtl for long double is tedious and error-prone — forget the suffix and you silently get the wrong precision. tgmath.h (C99) provides type-generic wrappers that dispatch to the right variant based on the argument type, so you can just write sqrt and get the correct function.

#include <tgmath.h>
 
float       f  = 2.0f;
double      d  = 2.0;
long double ld = 2.0L;
 
sqrt(f)    // calls sqrtf
sqrt(d)    // calls sqrt (double)
sqrt(ld)   // calls sqrtl

The dispatch happens at compile time via _Generic (C11) or compiler-specific mechanisms. All math functions from <math.h> that have f and l variants are covered, plus complex versions:

double complex z = 1.0 + 1.0 * I;
sqrt(z)    // dispatches to csqrt
sin(z)     // dispatches to csin

<tgmath.h> includes both <math.h> and <complex.h>, so all three headers' declarations become available when you include it.

The tradeoff is reduced explicitness. If you accidentally mix types in an expression, the dispatch may silently select a lower-precision variant. Precision-sensitive code — embedded DSP, signal processing, anything where you care about the exact FP operations — often prefers the explicit suffixes to make the precision choice visible in the source.

// compile: gcc -o tgtest tgtest.c -lm
// run: ./tgtest
// description: show that tgmath dispatches to different functions by type
 
#include <tgmath.h>
#include <stdio.h>
 
int main(void) {
    float       f  = 2.0f;
    double      d  = 2.0;
    long double ld = 2.0L;
 
    printf("sqrt(float)       = %.10f\n",  (double)sqrt(f));
    printf("sqrt(double)      = %.10f\n",  sqrt(d));
    printf("sqrt(long double) = %.10Lf\n", sqrt(ld));
 
    return 0;
}

All three produce approximately 1.4142135624, but the internal computation uses float, double, and long double precision respectively. To confirm the dispatch, compile with -S and check the assembly — you will see calls to sqrtf, sqrt, and sqrtl in the generated code.