[Bug target/88856] [8/9 Regression] gfortran producing wrong code with -funroll-loops

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88856

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
So, to me this looks like a backend bug, using dead_or_set_p in a splitter when
the split passes don't really compute the note problem.  Seems s390 is the only
backend that does this, other backends use dead_or_set_p either only in
peephole2s (which is fine, peephole2 pass starts with
  df_set_flags (DF_LR_RUN_DCE);
  df_note_add_problem ();
  df_analyze ();
and even when many targets don't use df_or_set_p, they do use peep2_dead*), or
(cris) in delayed branch scheduling (I believe that doesn't guarantee that
either).  Can't what you are doing in the splitters be done in define_peephole2
instead?