[Bug math/13304] fma, fmaf, fmal produce wrong results

["jsm28 at gcc dot gnu.org" <sourceware-bugzilla at sourceware dot org>]

http://sourceware.org/bugzilla/show_bug.cgi?id=13304

Joseph Myers <jsm28 at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|jakub at redhat dot com     |unassigned at sourceware
                   |                            |dot org

--- Comment #34 from Joseph Myers <jsm28 at gcc dot gnu.org> 2012-02-26 17:45:29 UTC ---
I think we should go with something like this version (but of course under
LGPLv2.1+ for glibc) for now - with the source file set up so the same file
(apart from license notices and space/tab differences) can be used in both
glibc and gnulib.

A soft-fp version might be faster - but the one in bug 3268 is not a clean
implementation using soft-fp (it relies on soft-fp knowing about a wider "quad"
floating-point type, rather than having an internal _FP_FMA operation that uses
more precision internally but without requiring a wider type).  So I think we
should use the portable implementation for math/s_fma{,f,l}.c for now.  If in
future there is a clean soft-fp implementation, that can then be used instead
(the portable version would still be useful for ldbl-128ibm - as I understand,
it will work as if that format uses a fixed 106 bit mantissa - unless and until
there is an implementation specific to that format).

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.