Re: Adding isinfl() and isnanl()

On 09/02/2026 15:45, Corinna Vinschen wrote:
> On Feb  9 08:49, Joel Sherrill wrote:
>> I went looking for a permissively licensed implementation over the weekend
>> and found this:
>>
>> https://github.com/JuliaMath/openlibm/blob/master/src/s_isinf.c
>>
>> Is that OK to add to libm/common?
> 
> It's 2-Clause BSD, sure.
> 
> But on second thought, maybe I was overcomplicating things.
> 
> Look at libm/common/sl_finite.c.  We have a precedent for using
> the builtin funcs.
> 
> For those targets not supporting FP in HW, maybe we just don't support
> long double?  In that case, the equivalent to the finitel function
> would be sufficient.
> 

Well AArch64 has HW floating point; but not for long double.  That is emulated.

>> If that's OK, would the cygwin guard on isinf*() in math.h just get deleted?
> 
> Yes.
> 
> Corinna
> 

R.