------- Comment #25 from jakub at gcc dot gnu dot org 2008-08-12 14:48 ------- Created an attachment (id=16059) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=16059&action=view) dbr.patch
Patch that should cure this testcase. We weren't handling the dbrs correctly. For the compute_barrier_args_size* calls when setting destination args_size depth for annulled branches we should only consider the insns from target and for normal sequence only insns from !target. And in normal args_size computation we need to ignore the target insns. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36998
