https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119786
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Ever confirmed|0 |1
Last reconfirmed| |2025-04-13
--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.
Looks like it has always been a bug:
for (reorg_pass_number = 0;
reorg_pass_number < MAX_REORG_PASSES;
reorg_pass_number++)
{
fill_simple_delay_slots (true);
fill_simple_delay_slots (false);
if (!targetm.no_speculation_in_delay_slots_p ())
fill_eager_delay_slots ();
relax_delay_slots (first);
}
/* If we made an end of function label, indicate that it is now
safe to delete it by undoing our prior adjustment to LABEL_NUSES.
If it is now unused, delete it. */
if (function_return_label && --LABEL_NUSES (function_return_label) == 0)
delete_related_insns (function_return_label);
if (function_simple_return_label
&& --LABEL_NUSES (function_simple_return_label) == 0)
delete_related_insns (function_simple_return_label);
need_return_insns = false;
need_return_insns |= targetm.have_return () && function_return_label != 0;
need_return_insns |= (targetm.have_simple_return ()
&& function_simple_return_label != 0);
if (need_return_insns)
make_return_insns (first);
That is the return insns is filled outside of the normal "reorg_pass_number"
pass. So it was recording the statistics to the wrong element almost always.
Since this is just statistics, I doub't make a difference otherwise.
Note this bug has existed since reorg.c was added to the repo back in 1991
(r0-97-g9c7e297806a27f).
