https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87047
--- Comment #4 from Alexander Monakov <amonakov at gcc dot gnu.org> ---
This is probably due to logic issue in PR 78120 fix. Its Changelog said,
(noce_process_if_block): Compute an estimate for the original cost when
optimizing for speed, using the minimum of then and else block costs.
but the change was
+ if (else_bb == NULL)
+ if_info->original_cost += then_cost;
+ else if (speed_p)
+ if_info->original_cost += MIN (then_cost, else_cost);
+ else
+ if_info->original_cost += then_cost + else_cost;
so when there's no else_bb, it effectively treats its cost as infinite rather
than zero. As a result GCC wrongly decides that replacing conditionally
executed block with cost 80 by a sequence with cost 72 is fine.
It also seems strange that costs are not scaled by edge probabilities.
