http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55939
--- Comment #17 from Aldy Hernandez <aldyh at gcc dot gnu.org> 2013-01-29 21:20:53 UTC --- Both -fexcess-precision=standard and -ffloat-store fix the testcase. This is perhaps a duplicate of PR323. Although, Jakub mentions that according to: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=323#c127 ...-fexcess-precision=standard might not be 100% reliable on m68k, so you may have to go for -ffloat-store. Mikael, can you check the GMP testsuite with -ffloat-store? And perhaps see if there is a floating point EQ/NE compare in the GMP test assembly? Thanks.
