[Bug math/14785] New: Incorrect results from fma in round-toward-zero mode when a * b is small and of opposite sign to c

[Bug math/14785] New: Incorrect results from fma in round-toward-zero mode when a * b is small and of opposite sign to c

[jsm28 at gcc dot gnu.org]

http://sourceware.org/bugzilla/show_bug.cgi?id=14785

             Bug #: 14785
           Summary: Incorrect results from fma in round-toward-zero mode
                    when a * b is small and of opposite sign to c
           Product: glibc
           Version: 2.16
            Status: NEW
          Severity: normal
          Priority: P2
         Component: math
        AssignedTo: unassigned@sourceware.org
        ReportedBy: jsm28@gcc.gnu.org
    Classification: Unclassified


fma (a, b, c) gives incorrect results in round-to-zero mode in some cases when
a * b is sufficiently small that the implementation uses a * b + c for the
computation, but a * b has opposite sign to c.  Example, on x86: fmal
(0x1p-16382L, -0x1p-16382L, 0x1p16383L).  In such a case, a * b rounds to zero,
and so the final result returned is equal to c, but correct rounding towards
zero should return the next value after c in the direction of 0.  Using a * b +
c would give the correct results in the other three rounding modes, but not in
FE_TOWARDZERO mode when the signs differ like this.

-- 
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.

[Bug math/14785] Incorrect results from fma in round-toward-zero mode when a * b is small and of opposite sign to c

[jsm28 at gcc dot gnu.org]

http://sourceware.org/bugzilla/show_bug.cgi?id=14785

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #1 from Joseph Myers <jsm28 at gcc dot gnu.org> 2012-11-01 16:49:18 UTC ---
Fixed by:

commit 473611b22d630a5cbb870d2a7edf08071d4d8ace
Author: Joseph Myers <joseph@codesourcery.com>
Date:   Thu Nov 1 16:47:26 2012 +0000

    Fix fma (a, b, c) for small a * b (bugs 14784, 14785).

-- 
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.

[Bug math/14785] Incorrect results from fma in round-toward-zero mode when a * b is small and of opposite sign to c

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=14785

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |security-

-- 
You are receiving this mail because:
You are on the CC list for the bug.