| |||
| Home > Floating-point support > Using C99 signalling NaNs provided by mathlib (_WANT_SNAN) | |||
If you want to use signalling NaNs, you must indicate this
to the compiler by defining the macro _WANT_SNAN in
your application. This macro must be defined before you include
any standard C headers. If your application is comprised of two
or more translation units, either all or none of them must define _WANT_SNAN.
That is, the definition must be consistent for any given application.
You must also use the relevant command-line option when you
compile your source code. This is associated with the predefined
macro, __SUPPORT_SNAN__.
Compiler Reference: