[Bug c++/31789] New: odd roundoff behavior (may be due to 80-bit vs. 64-bit behavior?

["dkruger at stevens dot edu" <gcc-bugzilla at gcc dot gnu dot org>]

Adding +x to -x should result in a hardware zero. But in this small gaussian
integration routine, .6^3 + -.6^3 is not zero.
When the same computation is done inline, it works.
This may be due to the hardware rounding properties, but given that the
integration routine is inline, I wouldn't see why it would behave differently
than the physically inline code. It's at the very least weird. My apologies if
this isn't a bug.


-- 
           Summary: odd roundoff behavior (may be due to 80-bit vs. 64-bit
                    behavior?
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dkruger at stevens dot edu
  GCC host triplet: i686-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31789