http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52275
--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2012-02-16 13:46:30 UTC --- > If I apply the pseudo-patch at the end, air still works fine at -O3 or -O2 on > x86_64-linux. So, what exactly is miscompiled with that patch applied? -O2 > -fgraphite-identity? Something else? Please read the beginning of the post: > Revision 184265 fixes the ICE I saw on the 2005 polyhedron test air.f90. > However as reported in comment #9 of pr50561, the test is miscompiled with > '-O2 > -floop-flatten' and higher optimization: Can it be clearer?