Inside Macintosh: PowerPC Numerics / Part 2 - The PowerPC Numerics C Implementation

Chapter 10 - Transcendental Functions

Comparison Functions

MathLib provides four functions that perform comparisons between two floating-point arguments:

$fdim(x,y)$	Returns the positive difference x - y or 0.
$fmax(x,y)$	Returns the maximum of x or y.
$fmin(x,y)$	Returns the minimum of x or y.
$relation(x,y)$	Returns the relationship between x and y.

These functions take advantage of the rule from the IEEE standard that all values except NaNs have an order:

$-$ < all negative real numbers < $-0$ = +0 < all positive real numbers < +

These functions also make special cases of NaNs so that they raise no floating-point exceptions.