[Cfp-interest] optimization with signaling NaNs

[Jim Thomas]

Here's another proposal for addressing optimization with signaling NaNs. 
Add the following (at places to be determined):

Recommended practice: If the state of FENV_ACCESS is "on", then any 
operation with a signaling NaN input, unless explicitly specified 
otherwise, raises the "invalid" floating-point exception.

The macro

FE_SNANS_ALWAYS_INVALID

is defined if and only if the implementation follows the recommended 
practice above.

Any operation that raises an "invalid" floating-point exception, if 
delivering a floating type result, shall return a quiet NaN.

-Jim

Jim Thomas wrote:

> I'd like to propose the following specification to cover optimization 
> with signaling NaNs:
>
> For signaling NaN inputs, C FP arithmetic operations that return a FP 
> result shall either signal invalid and return a quiet NaN or else 
> return a signaling NaN.
>
> -Jim
>