[Bug 294719] lib/msun: Added fmaximum_mag_num

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D294719

Robert Clausecker <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|16.0-CURRENT                |CURRENT
             Status|New                         |Open

--- Comment #4 from Robert Clausecker <[email protected]> ---
(In reply to Jes=C3=BAs Cipriano Bl=C3=A1zquez Mart=C3=ADnez from comment #=
3)

Ok, I guess I was unclear about what I meant by "Not sure if the void cast =
to
avoid the unused warning is actually needed".

I meant that perhaps the whole (void)force_except; statement can be removed.
If it is present, then yes, you should cast (void) to avoid a =E2=80=9Cstat=
ement
without effect=E2=80=9D warning.  But I think it can be removed entirely.

I'll check which case it is and copy-edit your patch before landing it.

I do also wonder if we can do better in the final tie break.  How about this
for fmaximum_mag*()?

    return (signbit(x) ? y : x);

and conversely for fminimum_mag*()?

    return (signbit(x) ? x : y);

This would also apply to the versions with _num() and would allow us to avo=
id
unpacking the datum entirely.

Unfortunately our signbit() macro resolves in such a way that the compiler =
does
not recognise it as such and generates really bad code.  That should be fix=
ed
long term.  Maybe short term the current approach is thus better.

kargl, what do you think?

--=20
You are receiving this mail because:
You are the assignee for the bug.=