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

Chapter 8 - Environmental Control Functions

Controlling the Exception Flags

In MathLib, the following functions control the floating-point exception flags:

feclearexcept	Clears one or more exceptions.
fegetexcept	Saves one or more exception flags.
feraiseexcept	Raises one or more exceptions.
fesetexcept	Restores the state of one or more exception flags.
fetestexcept	Returns the value of one or more exception flags.

The five floating-point exception flags are defined as the constants shown in Table 8-2.

Table 8-2 Floating-point exception flags in MathLib

Exception	Constant
Inexact	FE_INEXACT
Divide-by-zero	FE_DIVBYZERO
Underflow	FE_UNDERFLOW
Overflow	FE_OVERFLOW
Invalid	FE_INVALID

MathLib also defines another constant, FE_ALL_EXCEPT, which is the logical OR of all five exceptions. Using FE_ALL_EXCEPT, you can manipulate all five floating-point exception flags as a single entity. The type fexcept_t also exists so that all
the exception flags may be accessed at once.