Names specified for `<fenv.h>`, “Floating-point environment”
Name	Description				Source
`FE_ALL_EXCEPT`	All floating-point exceptions		M		`<fenv.h>`	C99	C11
`FE_DFL_ENV`	Default floating-point environment		M		`<fenv.h>`	C99	C11
`FE_DIVBYZERO`	Floating-point exception: division by zero	?	M		`<fenv.h>`	C99	C11
`FE_DOWNWARD`	Floating-point rounding mode: downward	?	M		`<fenv.h>`	C99	C11
`FE_INEXACT`	Floating-point exception: inexact computation	?	M		`<fenv.h>`	C99	C11
`FE_INVALID`	Floating-point exception: invalid representation	?	M		`<fenv.h>`	C99	C11
`FE_OVERFLOW`	Floating-point exception: overflow	?	M		`<fenv.h>`	C99	C11
`FE_TONEAREST`	Floating-point rounding mode: to nearest	?	M		`<fenv.h>`	C99	C11
`FE_TOWARDZERO`	Floating-point rounding mode: toward zero	?	M		`<fenv.h>`	C99	C11
`FE_UNDERFLOW`	Floating-point exception: underflow	?	M		`<fenv.h>`	C99	C11
`FE_UPWARD`	Floating-point rounding mode: upward	?	M		`<fenv.h>`	C99	C11
`FENV_ACCESS`	Warn of access to floating-point environment			P	`<fenv.h>`	C99	C11
`feclearexcept()`	Clear floating-point exceptions			(·)	`<fenv.h>`	C99	C11
`fegetenv()`	Get Get floating-point environment			(·)	`<fenv.h>`	C99	C11
`fegetexceptflag()`	Get floating-point exception flags			(·)	`<fenv.h>`	C99	C11
`fegetround()`	Get floating-point rounding direction			(·)	`<fenv.h>`	C99	C11
`feholdexcept()`	Save floating-point environment			(·)	`<fenv.h>`	C99	C11
`fenv_t`	Floating-point status			T	`<fenv.h>`	C99	C11
`feraiseexcept()`	Raise floating-point exceptions			(·)	`<fenv.h>`	C99	C11
`fesetenv()`	Set Get floating-point environment			(·)	`<fenv.h>`	C99	C11
`fesetexceptflag()`	Set floating-point exception flags			(·)	`<fenv.h>`	C99	C11
`fesetround()`	Set floating-point rounding direction			(·)	`<fenv.h>`	C99	C11
`fetestexcept()`	Test floating-point exceptions			(·)	`<fenv.h>`	C99	C11
`feupdateenv()`	Restore Get floating-point environment			(·)	`<fenv.h>`	C99	C11
`fexcept_t`	Floating-point exception status			T	`<fenv.h>`	C99	C11

This header is available in C99 and C11.

[

: This still needs more consideration.]

The floating-point environment is a collection of:

status flags that record when each kind of floating-point exception has occurred as a result of a mathematical computation; and
control modes that influence how future mathematical computations will proceed, including rounding, and whether exceptions should raise a signal [such as SIGFPE?].

The FP environment has thread-local storage, and all functions that operate on it access the environment of the current thread. Furthermore, by convention, unless documented to do so:

no function should alter the caller's control modes;
no function should clear any exception flags (but it can set them);
no function should depend on exception flags;
every function should assume that the caller is using default control modes at the point of invocation;

These conventions are designed to allow a programmer to be unconcerned with the FP environment if not conscious of it.

Functions that access the status flags include:

Functions that access the environment in general include:

fesetround
fegetround
fegetenv
feholdexcept
feupdateenv

The header <fenv.h> declares types, macros and functions for accessing the FP environment. fenv_t is a type used to store the entire FP environment, while fexcept_t can store only the state of the exception flags.

#pragma STDC FENV_ACCESS on-off-switch

on-off-switch: ON; OFF; DEFAULT

The pragma FENV_ACCESS must be enabled before using functions from <fenv.h>, or the behaviour of those functions is undefined. The effect of the pragma can have block scope or file scope. Within a block, it must occur at the very beginning; multiple such pragmas may occur together at the beginning, but what would be the point, except in generated code?

[

: List exactly which functions are covered by


        FENV_ACCESS

, if not all. Some might not operate on the FP environment, just on the caller's arguments. Actually, it does seem to be all of them in <fenv.h>.]

Missing

FE_DFL_ENV
feupdateenv
fesetenv
fegetenv
feholdexcept