http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49374
--- Comment #2 from Uros Bizjak <ubizjak at gmail dot com> 2011-06-12 11:45:39 UTC --- (In reply to comment #0) > Revision 174635 is OK. It may be caused by revision 174951: > > http://gcc.gnu.org/ml/gcc-cvs/2011-06/msg00440.html Huh? Unless these testcases depend on totally broken vec_dupv4sf pattern, I sincerely doubt that my patch could trigger these errors. And we don't vectorize at -O2.
