Yes that's likely. But the idea is that even if that's the case, it is still better than no PGO. Edit: I'd like to add that if the 10-20% mentioned is measured

Yes that's likely. But the idea is that even if that's the case, it is still better than no PGO. Edit: I'd like to add that if the 10-20% mentioned is measured on the benchmark that was used to do the pgo, then that figure might indeed not be representative of the real gain.