https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83189
--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
The fix at updating time is to promote to non-guessed 0 when it is demanded.
Index: tree-ssa-loop-manip.c
===================================================================
--- tree-ssa-loop-manip.c (revision 256419)
+++ tree-ssa-loop-manip.c (working copy)
@@ -1378,7 +1378,8 @@ tree_transform_and_unroll_loop (struct l
{
/* Avoid dropping loop body profile counter to 0 because of zero count
in loop's preheader. */
- freq_e = freq_e.force_nonzero ();
+ if (freq_h.nonzero_p () && !(freq_e == profile_count::zero ()))
+ freq_e = freq_e.force_nonzero ();
scale_loop_frequencies (loop, freq_e.probability_in (freq_h));
}
What is odd however is that predictive comonning is trying to do something
about loop executed 0 times. This should be fixed incrementally though. I will
test this change.