http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48509
Richard Guenther <rguenth at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |WAITING Last reconfirmed| |2011.04.08 10:05:33 Ever Confirmed|0 |1 --- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-08 10:05:33 UTC --- Works for me with just -O3 -ffast-math (on x86_64-linux). If you run on an Atom then -mtune=native will disable vectorization of doubles because it's slow. So, please provide more details.