------- Comment #2 from jakub at gcc dot gnu dot org 2009-03-13 12:38 ------- Created an attachment (id=17457) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17457&action=view) gcc44-pr39454-2.patch
Alternative simpler patch, in rtlanal.c just in case shift counts aren't canonicalized avoids assuming anything from them. As we should try to have the shift counts canonicalized, this handles the correctness part and could be just a small missed optimization if we fail to canonicalize them somewhere. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39454
