Yes, when opt_userstat enabled the unlikely() case becomes likely(). However, some overhead from enabling userstat is expected, and is probably much larger than branch mispredictions overhead. In other words, we are optimizing for the "common" case here, with the common case being userstat disabled.
Yes, when opt_userstat enabled the unlikely() case becomes likely(). However, some overhead from enabling userstat is expected, and is probably much larger than branch mispredictions overhead. In other words, we are optimizing for the "common" case here, with the common case being userstat disabled.