http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47878
--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-02-24 11:19:57 UTC --- (In reply to comment #3) > (In reply to comment #0) > > I see miscompares with -O3 and -O2 -ffast-math. > > Does the miscompare depend on the -O flags? If so, the patch for PR 47694 > cannot have cause it as it is a libgfortran only change. Looking at the > changelog, it then has to be a middle-end issue. I only extracted information from our autotesters and did not run manual tests yet. That the issue is seen across a wide range of platforms suggests the bug is in libgfortran.
